AI adoption happens between people. Measure it there.

Hi there,Welcome back to Untangled, a newsletter and community for people who refuse the story that technology just happens to us. It’s stewarded by me, ​​​​​​Charley Johnson​​​​​​, and enriched by ​​​​​​members​​​​​​ like you.This week I’m sharing my conversation with Paula Goldman about her new book, ​Manage the Machine​, and why rolling out AI one employee at a time misses where work gets done: between people.As always, please send me feedback on today’s post by replying to this email. I read and respond to every note.On to the show!Untangled HQAMA: On November 11 at 10:00 am PT, I’m hosting an AMA on organizational AI adoption. Are things going sideways inside your org? ​Join us!​New Tool: This week’s organizational tool is “​Buying AI: The AI Procurement Card​” — it walks you through what to ask and what to listen for when you’re paying for a subscription from a major provider.Deep DiveAI adoption happens between people. Measure it there.I’ve ​argued before​ that “human in the loop” functions more like a permission structure than an actual safeguard. On the podcast this week, ​Paula Goldman​—Salesforce’s first Chief Ethical and Humane Use Officer and author of the great new book ​Manage the Machine​—explained how the phrase got so watered down. Born from a Cold War-era tradition of thinking about human accountability in complex systems, the concept has recently been flattened into something Paula dryly describes as: “AI drafts something, and I hit yes.”She prefers “human at the helm.” You set the guardrails the system can’t cross, anomalies escalate to your desk, and the tool sharpens your judgment. Then she added an important caveat: “Or sometimes humans, plural, at the helm.”This post is about the plural. Almost every organization I work with is rolling out AI one isolated employee at a time—encouraging individuals to ​experiment​ and counting how many people used the tool this week. Paula sees the exact same pattern: Here’s a tool. Go use it. We want to measure your individual output.I think this is a massive category error. Because as Paula puts it, “work happens between people.” A functional team is a fundamentally different organism than a pile of individuals who just got 20% more efficient. And a dysfunctional team forms when we focus on individual output while ignoring the group patterns and dynamics individuals create.Multiplayer Alignment?Ethan Mollick isn’t so concerned. In his recent post, ​“The Dot and the Swarm,”​ he argued that we don’t need to worry about how AI work gets organized. Initially, Mollick assumed we’d need to design complex management structures for AI agents. Now, he thinks he was wrong: “It turns out that organizing work is just one more thing AI can learn to do.”Mollick contends that your agent (e.g. SpaceX’s Grok Bot, Meta’s Muse , OpenAI’s dots, etc.) can now coordinate other agents to solve your problems. Not unlike the ‘swarms’ OpenAI launched to produce the proof for the Navier-Stokes equations by exchanging 2.7 million messages over 88 hours with remarkably thin oversight.How this coordination plays out in organizations hinges on the principal-agent problem. This is the idea that people have personal goals that don’t always align with the company’s goals. “A lot of what we call management,” Mollick writes, “exists to solve problems that come from organizations being made of people.” Managers, in short, help align individual goals to the team’s goals, and from the team’s goals to the company.Now, Mollick concedes the principal-agent problem doesn’t vanish; it just moves to the space between the swarm and us. But he’s confident that it is manageable “as long as humans are guiding them in the right direction” and so long as the agents are “properly aligned to our needs.” But who is “our”? And what does “aligned” even mean in this context?I’ve ​argued previously​ that “AI alignment” is a myth. The systems AI touches are complex and unordered; they change as we act on them, and so do our goals. In those environments, what we actually need is an ongoing, collective sense-making process across diverse perspectives.I still think alignment is an incoherent concept, but when I critiqued it three years ago, I was effectively writing about one principal and one agent. Now that game is becoming multiplayer in both directions -- multiple people managing multiple agents in service, ostensibly, of a common goal. Imagine a foundation team where everyone has their own agent coordinating still other agents. The program officer wants to fund a grantee she’s championed for two years, so her agent drafts a glowing recommendation. The finance lead needs to keep next year’s budget flat, so her agent flags the grant as too expensive. The comms lead promised the board a splashy announcement, so his agent rewrites the summary to exaggerate the grantee’s reach. Each agent perfectly and faithfully serves its owner’s immediate needs. The result? An incoherent recommendation no one on the team actually wrote, born from invisible negotiations no one had, serving a collective goal the team never explicitly discussed. That is the principal-agent problem gone multiplayer that Mollick doesn’t address.RelationslippingThis is easier said than done, since ‘the team’ is constantly evolving. A healthy team operates on what the late social psychologist ​Dan Wegner​ called a transactive memory system: a shared map of who knows what. You don’t need to hold all the expertise, because you know which teammate to ask. New research suggests individual AI adoption is undermining that map.In a recent ​Harvard Business Review ​study, researchers surveyed over 1,000 U.S. desk workers. They found that as AI usage increased, so did the perception of teammates’ specialized knowledge. But the one metric that flatlined? How well people actually knew where that expertise lived.When you skip a five-minute chat with a colleague to get a quick answer from an AI, you never update your mental map of their expertise. As one respondent noted, “I have absolutely no idea what [my colleagues] are thinking or doing because the generative AI does most of their work.”The authors argue that over time, this leads to “relationslipping” (a clunky academic term for a very real problem): the slow decay of working relationships as we swap out our peers for prompts. And they found that on teams where this is the norm, coordination and trust drop over 20%.The authors also found that people with low psychological safety on their team were more than twice as likely to turn to AI instead of a teammate. As a regulatory manager in pharma put it, “I use AI when I do not want to be thought of as an idiot or unintelligent.” On a team where asking feels risky, that’s a reasonable choice. Multiply it across a team, though, and the shared map disappears.Designing for the TeamIn Manage the Machine, Paula tells a story about social workers in an English town who trialed an AI note-taker and surprisingly rejected it. The bot worked just fine but they enjoyed their afternoon walks together, where they traded notes, vented, and worked through hard cases together. In Wegner’s terms, those ‘inefficient’ afternoons were the glue holding their team’s memory together. Leaders need to identify these so called critical inefficiencies before they automate them away.The team is the unit that makes sense of an unpredictable world, and it does so through messy, inefficient, sometimes hard conversations. When I asked Paula about this, she pointed toward a different architecture: design the agent to shape and support the group’s dynamic. Some companies she studied programmed a team-level agent to play devil’s advocate, or to interrupt and say, “Here are the criteria we agreed on last week—are we skipping them?”That agent works for the team. The agents aren’t coordinating among themselves to support the individual; they’re working to support the coordination and collaboration of the team. I’m into it!But notice th… This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit untangled.substack.com

Hi there,Welcome back to Untangled, a newsletter and community for people who refuse the story that technology just happens to us. It’s stewarded by me, ​​​​​Charley Johnson​​​​​, and enriched by ​​​​​members​​​​​ like you.This week, I’m sharing my conversation with ​Afua Bruce​, an engineer, technology executive, and the founder and CEO of A&B Advisory. Afua also wrote The Tech That Comes Next, which I highly recommend.As always, please send me feedback on today’s post by replying to this email. I read and respond to every note.On to the show!Untangled HQ* AMA: On November 11 at 10:00 am PT, I’m hosting an AMA on AI adoption inside your organization. ​Join us!​* New Tool: I created a new tool to help you create an AI values statement for your organization. Most AI values statements sound nice and change nothing. This tool helps your team figure out what you believe about AI and what you won’t give up, then turn that into a short set of principles, each with a reason, an owner, and clear examples of what you’ll say yes and no to. For those keeping track, members now have access to four tools: ​AI Readiness Assessment​, ​Systems Change Strategy Assessment,​ ​Strategic Imagination​, and now the ​AI Values Statement.​ Become a member to get access to all of them!Deep DiveNo shortcut worth taking in the long-termWhen a foundation or nonprofit asks me for help with AI, the first conversation is usually about what its AI strategy should be. It turns out that Afua and I handle that question the same way. We acknowledge it, set it aside, and ask leaders to describe the future they’re working toward. Because if the first question is about AI, every problem starts to look like a problem AI can solve.Most organizations add AI to their existing ways of doing things. This is the well-worn path of automation that feels good in the short term — efficiency! builder mindset! experiment! — but it mistakes the shortest route for the right one.The right route is longer and more circuitous. It means redesigning workflows around the outcomes an organization wants, configuring AI tools to fit those workflows, and protecting the human work that depends on relationships and context or that teaches people judgment. My conversation with Afua is a practical deep dive into that path, and it prompted Afua to remember a line her dad often used: “There’s no shortcut worth taking in the long term.”Along the way, we discuss:* The questions Afua asks before she asks anything about AI: what the organization’s vision is, how it manages change, and who inside it is pushing for AI in the first place.* Why the work that needs to stay human is so hard to name. Much of it is tacit, like knowing how to read a room.* Afua’s approach to redesigning a workflow. She starts with what needs to get done, how staff and clients should feel, and what data the interaction should produce, and only then looks for places a tool might help.* Why so few organizations make it through the messy middle of that redesign, and what leaders have to protect for it to happen (time, mostly, plus permission for a small rollout to fail).* Why teams tend to converge on an AI-generated summary instead of debating it, and the facilitation moves Afua uses to protect independent thinking.* What organizations lose when they automate the junior work that builds judgment, and Afua’s point that Gen AI regresses to the mean while the communities organizations serve keep changing.* My worry that “human in the loop” works more like a permission structure than a safeguard, and Afua’s alternative: put AI in the decision-making loop and hold leaders accountable for the decisions it informs, the same way they’re accountable for the decisions their teams make.* The feedback loop that forms when an organization loads its whole archive into an LLM and treats the output as guidance, so that past practices end up steering future ones.* Why Afua thinks sensemaking, judgment, and systems thinking become more important as an organization adopts more AI.* What getting this right looks like ten years from now, from organizations comfortable saying no to bad use cases to strategic plans with no separate “AI plan.”We wrapped up with Afua’s new picture book, ​Bessie the Inventor: The STEMinist Who Forged Her Own Path​, which comes out in January. Bessie was a real inventor, born in 1914, who also worked as a nurse, a physical therapist, a handwriting expert, and a business owner. Afua wrote the book to show kids that technology belongs inside a community, built by people trying to help the folks around them. ​Pre-order a copy here.​I hope you enjoy my conversation with Afua Bruce.Until next time,CharleyWork With Me​Here are 3 ways I can help:* ​Advising: I can help you navigate uncertainty, make sense of AI, and steward change in your system.* ​Training: Everything you and your team need to cut through the tech-hype and implement strategies that catalyze true systems change.* ​​​​​​​​​1:1 Leadership Coaching:​​​​​​​​ I can help you facilitate change — in yourself, your organization, and the system you work within. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit untangled.substack.com

Welcome back to Untangled, a newsletter and community for people who refuse the story that technology just happens to us. It’s stewarded by me, ​​​​Charley Johnson​​​​, and enriched by ​​​​members​​​​ like you.This week’s podcast guest is ​Ami Fields-Meyer​. Ami was a senior policy advisor in the Biden White House, and he wrote ​On Courage: How to Be a Dissident in an Age of Fear ​with Julia Angwin. Ami and I talk about dissent, AI inevitability, and why every movement needs both a “no” and a “yes.”As always, please send me feedback on today’s post by replying to this email. I read and respond to every note.On to the show!Untangled HQLast week, I rolled out two free tools — an ​AI Readiness Assessment​ and ​Strategy for Uncertainty Assessment​. This week I’m rolling out a new tool -- ​Strategic Imagination​ -- for ​Untangled members​Imagination is one of the most important strategic practices in the context of AI, and most organizations are pretty bad at it. They get stuck in grooves they can’t see: inherited imaginaries about what’s possible, what’s realistic, and what the future is supposed to look like, many of them written by someone else. ​Strategic Imagination​ helps you break out of those grooves of the past, describe the future you actually want, and work backward to what it would take to get there.Deep DiveYour AI use policy needs a right to say no.In the episode, Ami and I discuss:* The ​White House meeting​ where Sam Altman couldn’t say what he’d do with his technology* The “regulate us… no, not like that” cycle* What the passive voice in AI discourse hides.* Data centers as leverage for renegotiating our relationship with big tech* The four-part framework for dissent in On Courage* Hungary, negative coalitions, and why movements need both a “no” and a “yes”The conversation left me thinking about the feeling of inevitability that is pervasive in AI. You know the one. The meeting transcriber is on by default. The all-staff memo announces that the organization is now “AI-first,” and the only question left for you is how fast you’ll get on board. Every news article quotes a Silicon Valley leader telling you what the future will look like, as if it has already been determined. And your agency is, well, no where to be found.Ami thinks this feeling of inevitability is the defining feature of this moment — and I couldn’t agree more. Authoritarian politics and the tech industry, he told me, are both “characterized by a profound and heavy sense of inevitability.” Even the Pope’s ​document on AI​, which Ami called “interesting and often beautiful,” accepts the premise “that this is here and there’s nothing we can do about the fact that this is here.” Ami’s antidote is just as clear: “puncturing that inevitability is part of what gives people the motivation and the strength to push back.” As he put it, “a democratic society is all about open horizons.”So how do you puncture it? Ami pointed to Dr. Maria J. Stephan, the political scientist who runs ​The Horizons Project​. She has studied the massive coalitions that have unseated authoritarian leaders, and argues “Grounding this work in an affirmative vision of the kind of inclusive democracy that prioritizes human dignity” is key. Every movement, in short, needs to say no to something and yes to something. And the ‘no’ needs to come first, according to Ami.“It’s really important to be able to build a coalition that’s big enough to reject the political repression. And that usually does not happen through fine-grained, narrow policy prescriptions because nobody can agree on those.”In other words, you don’t need everyone to agree on what comes next before you can agree on what has to stop. And when it comes to AI, there’s one place where people are already agreeing on what has to stop: the data center. It’s physical, it draws on your water, and it needs a zoning hearing. In ​a new paper,​ Faisal Saeed Malik, a researcher at the Karlsruhe Institute of Technology calls this “infrastructural resistance,” and Ami wants to use that local no as leverage to “renegotiate our contract with big tech writ large.” That’s the sequence Malik describes, from refusal to “reclamation,” where people reorganize the conditions under which technology gets governed. And as Malik insists, resistance “is not inherently emancipatory.” Some no’s are about democratic voice and some are about property values. Normalizing no doesn’t make every no right. It makes no a legitimate answer in the first place.What does saying no to AI look like inside an organization?At some level, what a ‘no’ looks like is clearer at the community or movement level. Data centers are physical and their presence directly impacts cities and towns. But what might a ‘no’ look like in organizations? As Malik ​argues​, “treating resistance primarily as a problem of acceptance or adoption risks overlooking the possibility that contestation may identify conditions under which… institutional dependence has become difficult to question.” In other words, whether refusal is an option tells you something about the organizational system around it. It might show you that a decision can no longer be challenged, or that opting out has become so costly that almost nobody does it anymore.Picture a nonprofit that rolls out an AI note-taker for every meeting. The policy says it’s optional. One program manager switches it off in her calls with community partners, because they’ve told her they won’t speak candidly while being recorded. So her meetings stop showing up in the organization knowledge management system. Colleagues prepping for a site visit search the system for background on those partners and find nothing. Some ping her for a recap. Others skip it and walk in cold. At her mid-year review, her manager tells her that her work “isn’t very visible.” The note-taker is still optional. She’s just paying for the choice in time and reputation. As Malik ​explains​, “a right to refusal has limited substance when non-participation effectively excludes individuals from essential institutions or opportunities.”The point is, a yes doesn’t mean much if you can’t say no. If an employee knows that declining the tool will show up in her performance review, a yes is simply compliance.Why AI use policies should include the right to opt outWhich brings us to AI use policies. Most of the organizational use policies I’ve seen govern how people use AI. They cover which tools are approved, what data can’t go into them, and when to disclose that AI was involved. Those are reasonable things to govern. But, importantly, these policies almost never ask whether. So the policy, which is the organization’s one formal statement about its relationship to AI, ends up reinforcing the felt sense of inevitability.It doesn’t have to. As Ami’s friend and former colleague Alondra Nelson likes to say, “nothing is inevitable,” and “we can steer machines.” An AI use policy is one of the few places where an organization can do that steering on purpose, by putting “no” back on the table. It could do that in a few ways:* It could give staff an explicit right to decline AI tools for their own work without penalty, so that no isn’t career-limiting.* It could guarantee the people an organization serves (applicants, grantees, patients, clients) a non-AI path and a human who can review a decision.* It could treat opt-outs as data, in Malik’s diagnostic sense. When people decline a tool, someone reviews where and why -- and uses that information to adapt the policy.* It could name who can pause a tool, and under what conditions.None of this is anti-AI per say. A policy like this makes every yes in the organization count for more — because it was freely given!So here’s what I’d suggest. Let’s retire “How do we get people to adopt AI responsibly?” It assumes everyone has already said yes. A more honest place to start is: where can people still say no, and what does it cost them? If you’re drafting an AI use policy right now, I’d start wi… This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit untangled.substack.com

Welcome back to Untangled, a newsletter and community for people who refuse the story that technology just happens to us. It’s stewarded by me, ​​​Charley Johnson​​​, and enriched by ​​​members​​​ like you.This week I’m writing about what David Robinson’s exit from OpenAI and Maggie Jackson’s work on uncertainty reveal about pacing AI in the workplace.As always, please send me feedback on today’s post by replying to this email. I read and respond to every note.On to the show!Untangled HQOver the next few months, I’ll roll out one new tool per week. The first two — an ​AI Readiness Assessment​ and ​Strategy for Uncertainty Assessment​ — are free and accessible to anyone. All subsequent tools will be paywalled for those who sign up for ​the new Untangled membership.​Deep DiveWe Have No Time to Rush: Why Organizations Should Slow Down AI AdoptionThis week I’m sharing my conversation with Maggie Jackson, author of ​Uncertain: The Wisdom and Wonder of Being Unsure​, a book I’d hand to anyone leading an organization through AI right now. We discuss:* How the early AI field treated uncertainty as an obstacle, and why AI that can say “I don’t know” matters.* What happens to our curiosity when AI search summaries hand us one tidy answer.* Whether tolerance of uncertainty is a skill people build or something their circumstances allow.* Why people grow more confident in AI’s answers while losing confidence in their own abilities.* How organizations can slow down AI adoption, and what surgeons can teach them about pausing.The conversation is well worth the listen, and it got me thinking about what organizations lose when they adopt AI in a hurry. If AI companies are unwilling to slow down, we can.A while back, Maggie gave a short talk on uncertainty to a small group of chief technology officers from Fortune 100 companies. When she finished, the first response wasn’t a question. One CTO told her, in effect, “We don’t have time for uncertainty. We need answers immediately.”This sums up the status quo. Adopt AI as fast as humanly possible. Ask questions later, or preferably, never. Speed is scarce, and uncertainty eats time, so whatever you do, don’t slow down. Imagining alternative futures AI use might enable or constrain? Too much time! Redesigning workflows in a way that intentionally distinguishes between what must stay human and what AI does well? Not enough time. Attending to the organizational dynamics AI use might exacerbate? You get the point.This entire way of thinking is premised on the assumption that uncertainty slows us down, speed is a virtue, and pausing is a luxury we simply don’t have.Why AI companies won’t slow downThe companies building AI work from the same premise. Earlier this month, ​David Robinson quit OpenAI​, where he led the writing of the safety reports the company publishes alongside new models. I got to know David a bit when I was at Data & Society Research Institute. He’s deeply smart and thoughtful, and he’s not quick to hyperbole. He’s as even-keeled as they come. So I was keen to read ​his essay​ in The Atlantic, and listen to ​his conversation​ with Ezra Klein.I’d encourage you to read/listen to both. Basically, Robinson argues that the whole AI industry has an unsafe culture, and he described how fast things now move inside it. New releases used to come a few times a year. Now companies keep adding new training, tools and features to existing models, and,​ he said​, “we’re shipping new capability and risk every Tuesday.” Klein asked whether safety testing can keep up at that speed. Robinson answered a narrower question, how much time there is to kick the tires: “And the answer is not a ton.” There are brakes, ​he said​, and some launches and training runs have been stopped, but the pauses are “very carefully scoped.”Robinson also explained why he didn’t see the danger sooner, which narrowed in on incentives and time. He joined OpenAI in May 2023, and “it’s been one Slack ping after another ever since then,” ​he said​. “Only in stepping back from my operational responsibilities over the last few weeks have I started to have the time to really reflect on where we are.” The pace of the work, in short, kept him from seeing what the work was doing. (That should sound familiar to anyone running adopting AI in their organization.)He concluded that OpenAI “is such a machine and it is moving so fast” that the change he thought was needed couldn’t come from inside. As he left, he ended a note to colleagues with this line: “We have to make good choices. We have no time to rush.”How uncertainty helps leaders make better decisionsRobinson wrote that line for his colleagues at OpenAI, but it applies just as well to the companies and organizations adopting AI — and to the presumably exhausted CTO Maggie encountered. The CTO’s comment rests on a common assumption: leaders who feel unsure hesitate, and leaders who feel certain act. Maggie pointed me to research from Nils Plambeck and Klaus Weber who ​tested​ that assumption and surveyed 104 German CEOs about the European Union’s 2004 expansion, a change that opened new markets and brought in new competitors. The upshot? The CEOs who were ambivalent going into this challenge navigated it most skillfully. Those who saw it as clearly good or bad for their company — those who felt most certain — did almost nothing. They got stuck. Feeling ambivalent and uncertain, in other words, pushes leaders to look beyond their usual playbook for answers.So where does that ambivalence come from? In a ​follow-up study, Plambeck and Weber found that it depends on two big things. First, CEOs whose companies pursued new opportunities while also protecting what they already had were more likely to see both sides than CEOs whose companies did only one or the other. The second was the company’s sense of control. CEOs who thought their company had little control over the expansion didn’t look closely, and CEOs who thought it had a great deal grew overconfident and fell back into their routine. The CEOs in between were the most likely to feel torn and embrace uncertainty.How to pace AI adoption in your organizationIt’s hard not to see current AI adoption status quo in both of those groups. Some leaders see AI adoption as inevitable. It is happening to them, so the vendors speed sets the pace and the market sets the timeline. Their only job is to keep up. These leaders look a lot like the low-control CEOs who stop looking closely. By contrast, the leaders who treat AI just like another IT rollout — who look for stability in past routines — look like the high-control CEOs. The leaders most likely to sit with the question long enough to respond well are the ones in between: they believe their choices shape how AI changes their organization, without believing they control the outcome.Slowing down doesn’t mean doing nothing. It means moving with intention or as Jen Briselli ​suggested​, never acting faster than you can learn. Maggie once spent time embedded in operating rooms, and some of the surgeons understood that to operate well in a crisis, that “you just need to spend a couple of minutes.” The surgeons used those minutes to access the reflective, curious side of themselves before they acted. The pause was intentional and purposeful, albeit brief. An organization’s pause needs a purpose too. It needs to be able to answer the question Robinson and Maggie keep coming back to: is your organization clear on what must stay human, ready to say what it doesn’t know, and attend to the organizational patterns AI-use generates?That’s the question behind the ​AI Readiness Assessment​ I mentioned up top. It’s free, and it’s a good place to start if you want your organization’s pause to be purposeful. Take it, share it with your team, and compare answers. We have no time to rush.Until next time,CharleyWork With Me​Here are 3 ways I can help:* ​Advising: I can help you navigate uncertainty, make sense of AI, and steward change in your system.* ​… This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit untangled.substack.com

Welcome back to Untangled. It’s written by me, ​​Charley Johnson​​, and valued by ​​members​​ like you.This week I talked with ​Jen Briselli​, co-founder of ​Topology​ and an adjunct professor at Massachusetts College of Art and Design. Jen is one of my favorite thinkers on complexity, and she’s written two essays in the last six months that I keep returning to: one on ​what hockey taught her about complexity​, and one arguing that ​“uncertainty tolerance”​ doesn’t live inside a person.We discuss:* Why “skate to where the puck is going” is worse advice than it sounds.* What a triangle of three forwards can teach an organization about rules.* Why “you’re not in traffic, you are traffic” is the hardest lesson in complexity work.* Jen’s image of organizations amputating healthy limbs to make room for a prosthetic -- and why it’s an apt way to think about AI adoption.As always, please send me feedback on today’s post by replying to this email. I read and respond to every note.On to the show!Untangled HQIn our next ​Facilitators' Workshop​, Kate and I are hosting an '​Ask Us Anything.​' ​Bring the meeting that went sideways. Bring the group that won't stop talking, or the one that won't start. Bring the facilitation question you've been slightly embarrassed to ask out loud. We'll dig into it all together.Deep DiveUncertainty tolerance is not a personality traitJen wrote ​her essay on uncertainty tolerance​ because she was fed up with ​headlines​ telling leaders that their teams can’t handle ambiguity. She argues that people actually love uncertainty. We gamble, we play sports, and we get mad at anyone who spoils the end of a movie. What people resist is uncertainty that comes with high consequences and no influence over the outcome. If those consequences reach someone’s salary, Jen said, “I don’t blame someone for being a little bit uptight about it.”The point, as Jen argues, is to stop treating uncertainty tolerance as a trait that a person has or lacks. Tolerance emerges from the entanglement between a person and their context. The same person who bluffs happily at a poker table on Saturday can freeze during the announced reorg on Monday. Nothing inside that person changed over the weekend. The stakes changed, and so did how much say they have.Why employees hesitate to experiment with AINow map this distinction on to being instructed by your boss to “go experiment with AI.” The generous read might be that they don’t know how to strategically incorporate it into a workflow, and they want to involve you in the process. A li’l experimentation can’t hurt, right?The less generous interpretation is that management is asking staff to figure out what parts of their job can be automated — and they’re being asked to fork over that information with no sense of how it’s going to be used. Read this way, your hesitation makes sense. Why would you experiment with AI if success means you may have just shown your employer how to automate part of your job? When leadership calls that hesitation “resistance,” they locate inside you something that you and they produced together.How individual AI experimentation becomes a team problemThe other issue with this instruction is that “go experiment with AI” asks you to act. It says nothing about whether you’ll ever learn what your action set in motion — as an individual or a team. This cuts against one of the core ideas in ​Jen’s hockey essay​: never act faster than you learn. Take actions that are, in her words, high in learning potential and low in disaster potential, and sense for what those actions changed. So if your experimenting faster than your learning, slow down.Now comes the hard part: doing this as a team. Because teams that don’t sense and learn together develop problematic patterns. It’s rarely because someone has bad intentions or doesn’t care about their impact on the group. It’s because everyone is moving fast and trying new things. It feels like they’re saving time and being more efficient. But it turns out that each individual, working alone, is making what seems like a small accommodation: deferring to a confident output instead of forming their own view, starting from the AI’s framing instead of their own, letting the time it saves get absorbed as more volume. Within teams, these small accommodations snowball. They build and build until they become a patterned dynamic — judgment that no longer gets exercised, disagreements that no longer get voiced, accountability no longer exercised. And no one notices, because action outpaced learning and the action felt productive.What hockey systems teach leaders about AI adoptionHockey has something to offer here, too: hockey teams operate in systems, not plays. As Jen explained, 1-2-2 forecheck tells almost no one what to do. It tells the first forward to pressure the puck, the two behind her which lanes to hold, and the two behind them what they own if the puck gets through. Everyone can see where everyone else is, and the decision belongs to whoever is closest to the puck. The first forward can skate hard at a risky puck because two layers behind her make losing it recoverable. Her uncertainty is tolerable because of where her teammates are standing. “Go experiment with AI” is the opposite of a system. It says nothing about where anyone stands in relation to anyone else or what layer is responsible for covering the person who takes the risk.So what’s a better approach than ‘go experiment’? First, say plainly what will and won’t happen to someone’s role if they find a way to automate part of it, and keep your word. Second, help people see the system: protected time where people from different teams compare what they tried, what they learned, what they’re sensing and observing, what they would change, etc.This isn’t a moment for individual experimentation in a vacuum. It’s a moment for collective sense-making and learning. The technology is moving faster than any one person can grapple with one their own. But we can put in place team and organizational processes that attend to emerging patterns and allow groups to adapt together.So the next time a leader says “go experiment with AI,” the right response is to ask, where do you want me to stand, and who’s behind me?Until next time,CharleyWork With MeHere are 3 ways I can help:* ​​​​​​​​​Advising:​​​​​​​​ I can help you navigate uncertainty, make sense of AI, and steward change in your system.* ​​​​​​​​​Organizational Training:​​​​​​​​ Everything you and your team need to cut through the tech-hype and implement strategies that catalyze true systems change. (For either Stewarding AI or Systems Change for Tech & Society Leaders)* ​​​​​​​​​1:1 Leadership Coaching:​​​​​​​​ I can help you facilitate change — in yourself, your organization, and the system you work within. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit untangled.substack.com

Welcome back to Untangled. It’s written by me, ​Charley Johnson​, and valued by ​members​ like you.This week I’m sharing my conversation with ​Nick Pyati​, a former senior strategy leader at Microsoft and the founder of ​Belari AI​. Nick spent nine years doing strategy for a company whose ground shifted every few months, and then all at once when ChatGPT landed. We disagreed about how much of the work AI will eventually do but on what strategy is and why most companies are getting it wrong, we landed in the same place. I hope you enjoy it — I certainly did!Untangled HQGet on ​the waitlist​ for the second cohort of Stewarding AI: How to Build Responsible Principles, Workflows, and Practices. Here's what one participant had to say about the first cohort:Deep DiveWhat is an ‘AI strategy’?Nick and I talk about a lot of things:* What Nick learned about setting strategy at Microsoft in the months after ChatGPT, when planning horizons compressed from a year to a few months.* The three approaches to AI Nick sees right now, and why the companies that have done nothing about AI may be best positioned.* Facilitation as a strategic practice, and what the human conversations do that most analytical approaches to strategy can’t.* What AI is doing to knowledge work, and how companies are commodifying what once set them apart.But the question we kept circling back to was this: when a company’s context changes, how does it rebuild the link between that context and the way it works? Every team works inside a context — the people it serves, the competitors it faces, the technology available to it, and all of the interpersonal dynamics and ways of doing things that amount to culture. And every team has a set of habits, beliefs, and capabilities that fit that context.But then the context shifts. A new competitor arrives, customers start wanting something else, technology changes, etc. The habits, beliefs, and capabilities that made the team successful stop making sense. Yet they keep doing the same thing! Why? Because, as Nick argues, it served them well for years. Re-establishing the link means first recognizing that the old way of doing things is no longer serving you. High performing teams across companies, non-profits, and government will paradoxically have the hardest time because they have so many positive examples validating their current approach.And here we are. AI changed the context for everyone everywhere all at once. Your staff are working differently. The people you serve, and the board you answer to, expect something different from you. Somewhere a competitor you haven’t heard of is rebuilding the work from scratch. A context that felt durable now feels more like quicksand. As Nick notes, every company in every industry now has the problem Microsoft had in the months after ChatGPT, when planning horizons shrank from a year to a few months and the historical trend line stopped being a guide to, well, anything. Strategy, then, is whatever it takes to re-establish the link between how a team works and the context it now sits in.But strategy is also a big bet. Yes, Microsoft is data-rich and analytical, but Nick says almost none of that data decided the big questions. When your context is regularly changing, forecasting or trend analysis isn’t much use. Strategy stops being an analytical question because the future can’t be known. Instead, strategy becomes more about imagination, conviction, and sense-making; about what the team can believe in, and how the team adapts together without fragmenting. I’ve written recently about ​imagination​ and ​sense-making​, so what makes for a good big bet?Nick thinks of it as a Venn diagram. The bet has to be plausible — supported by what the market is doing, with a real story for why this team has a comparative advantage. The team has to be able to get conviction around it. They have to believe in it. And the bet has to be big enough to be worth the risk, because of the fifteen good ideas on the table, maybe three will ever pay off at a scale that justifies the investment of time and resources.Companies and organizations aren’t making big bets mapped to the future they want to bring about. They’re focused on the same question every company has asked of every IT rollout before this one: where can we plug this in? What can we automate? And when you ask these questions of a technology that can reshape how the whole business works, you get a scattering of small bets with no idea underneath them. I wrote about the mechanism behind this in ​The Copycat Economy​. When a technology is poorly understood, organizations copy each other rather than think for themselves — DiMaggio and Powell called it mimetic isomorphism. Every company is doing what its peers are doing, which is why their ‘AI strategies’ all look roughly the same.When your ‘AI strategy’ maps the technology to existing work, another differentiation problem pops up. The models are the same for everyone. Unless a company is using them in some thoughtful, tailored way, whatever it produces with them is essentially what everyone else can produce with them. That model is what a company turns itself into when it automates the work its people were doing. The idiosyncratic humans on staff, with their peculiarities, big unconventional ideas, and their weird accumulated knowledge of how things get done here, were the entire source of anything novel the company had to offer. As I argued in an essay on ​why data-driven organizations are the least prepared for AI​, this knowledge is tacit and relational — it lives in bodies, in trusted relationships, in a team’s accumulated sense of how decisions get made versus how they’re supposed to get made. The knowledge isn’t in the text, so it isn’t in the model.Every time a CEO or Executive Director highlights the ‘efficiencies’ generated by cutting their workforce, Nick said you should interpret that as: you lacked the imagination to know what to do with those people now that you have the capability. I agree! It means that they started from the question “what is our AI strategy” and simply bolted the technology on to what they already do, as the context changed beneath their feet. The better question is this: what future are we trying to create, what do the people we already have make possible, and where — if anywhere — does AI help? Answer that and you have a bet.More soon,CharleyWork With Me​Here are 3 ways I can help:* ​​​​​​​​​Advising:​​​​​​​​ I can help you navigate uncertainty, make sense of AI, and steward change in your system.* ​​​​​​​​​Organizational Training:​​​​​​​​ Everything you and your team need to cut through the tech-hype and implement strategies that catalyze true systems change. (For either Stewarding AI or Systems Change for Tech & Society Leaders)* ​​​​​​​​​1:1 Leadership Coaching:​​​​​​​​ I can help you facilitate change — in yourself, your organization, and the system you work within. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit untangled.substack.com

Hi there,Welcome back to Untangled. It’s written by me, ​​​​​​​​​​​Charley Johnson​​​​​​​​​​​, and valued​​​​​​​​​​​​​​​​​​​​ by ​​​​​​​​members​​​​​​​​ like you. ​​​​​​​Help me make it better?​​​​This week I’m sharing my conversation with Marina Nitze — former CTO of the Department of Veterans Affairs, senior White House technology advisor, and co-author (with Mikey Dickerson and Matthew Weaver) of Crisis Engineering: Time-Tested Tools for Turning Chaos into Clarity. We talk about the book, sense-making practices, and what actually helps a system change.As always, please send me feedback on today’s post by replying to this email. I read and respond to every note.On to the show!Untangled HQ* Stewarding AI: The next cohort of Stewarding AI: How to Build Responsible Principles, Workflows, and Practices is coming up. ​Get on the waitlist and be the first to hear when it opens. ​* Ask Me Anything: Ask me about the knots in your own systems change work (”my funder wants a five-year plan but my system doesn’t work like that — what do I do?”). Ask me about stewarding AI inside your organization (”my board/boss keeps asking what our AI strategy is -- where do I start?”). Ask me about something in the news you want untangled. Or ask me something personal — what I’m reading, what I’ve changed my mind about this year, or what my high-school self — Dashboard Confessional blasting in his headphones — would make of this newsletter. No question is too small or too “I should already know this.” Just use this form.Deep DiveYou can’t make sense of the world alone with a chatbot.Marina and I talk about a lot of things (watch it here or on YouTube), including:* How the Healthcare.gov crisis ended on a shared story rather than any change in the metrics, and what Marina took away from living through it.* Why nobody in a large org holds an accurate picture of the whole system, and what it takes for a group to see it together.* Why people can’t be argued out of a comfortable story with facts, and what actually lets a group take up a truer, stranger one.* How crises loosen what was stuck, why most orgs waste the opening, and what leadership looks like inside that window.* What AI adoption would look like if it took the author’s account of why organizational change usually fails seriously.* How a leader should weigh automating against sustaining the human adaptive capacity that automation tends to engineer away.* What AI does to an org’s ability to know itself when AI summarizes the reports and watches the metrics, and where sensemaking has to stay human.You should give the entire episode a listen / watch, but I wanted to zoom in on one thread we kept coming back to: what actually allows systems to change amidst a crisis?When Marina was CTO of the VA, the disability claims backlog kept growing, and the only tool the agency had was mandatory overtime. So she flew around the country and watched claims get processed from start to finish — which, at the time, nobody owned. People owned steps of the process. Nobody owned the process. And nobody had traced the entire thing.She sat in on medical exams. Doctors would examine a veteran head to toe and then write an eight-page medical essay. When she asked them why, their answer was consistent: I’m a medical expert, writing for an even better medical expert, who will read every detail and make a nuanced decision about this veteran’s compensation. The longer the essay, the thinking went, the better served the veteran.Except the essay didn’t go to a better medical expert. It went to a claim specialist with no medical training and a very specific form — a form that asked, say, for the number of degrees the veteran’s elbow could bend. Often the number wasn’t anywhere in the eight pages. So the claim got kicked back, and the veteran was sent in for another head-to-toe exam. Not great.The doctors believed they were serving veterans as thoroughly as possible. So did the claim specialists. Yet together they were grinding veterans through repeat exams and growing the backlog. What fixed this system wasn’t a fact-filled memo, it was sense-making. Marina brought everyone into one room — the doctors, the claim specialists, etc. — and had them demonstrate their part of the process to the other. They left understanding how their well-intentioned actions were contributing to the overall system.Sensemaking is about how we assemble what’s happening around us into a story we can act on. Marina and her co-authors lean on organizational theorist Karl Weick in the book, who contends that sensemaking runs on plausibility, not accuracy. Given the choice, we’ll take a familiar story that fits some of the facts over a strange or complicated story that fits all of them. The book uses the accident at Three Mile Island as an example: nearly fifty years on, there are still two incompatible stories of that accident — “we were lied to,” and “honest mistakes were made, no harm done.” Both stories are alive and well, and no fact is coming along to settle it.The sensemaking process didn’t simply offer new information to well-intentioned actors. It was a collision with reality that went deeper than the workflow. Right, the doctors didn’t care about writing the eight-page document. They were protecting their own identity, an image of themselves as the person who serves veterans thoroughly. And now they’re in a room showing them that their contributions were an obstacle to change. A new shared story can’t change the system unless the participant’s identity shifts alongside it. From I’m the expert who serves veterans to I contribute to a process that serves veterans. Turns out, facts are downstream from the shared stories we tell. And all of that is downstream from who we think we are.Near the end of our conversation, almost in passing, Marina said something I haven’t stopped chewing on since: “You’re not going to be able to sense-make with yourself and your LLM. That’s actually fairly dangerous.” A chatbot simulates a conversation but it doesn’t offer the friction required for you to update your picture of the world, or yourself. Your chatbot doesn’t hold a different map than you do; after a few exchanges it mostly holds yours because its incentives are set to find your story plausible. And there is no room for it to read — no raised eyebrows or dismissive smirks, no stake in the disagreement. Whereas sensemaking is fundamentally social. It travels through tone and body language and power dynamics, which is why Marina insists you cannot work a crisis over Zoom — everyone has to be in the same physical room, looking at the same reality, revealing and picking up on social cues, interpersonal issues, and all the messy things about being a human that cannot be automated away.Organizations are concerned with the ‘AI risks’ they hear about in the media. How can we minimize ‘hallucinations?’ How can we ‘align’ it with our values? If your a long-time Untangled subscriber, you know I’ve been poking holes in these frames for a long time. Because the real danger of making sense of the world with a chatbot was never that it tells you false things or contradicts your values. It’s how smooth your own story starts to feel.More soon,CharleyWork With MeHere are 3 ways I can help:* ​​​​​​​​​Advising:​​​​​​​​ I can help you navigate uncertainty, make sense of AI, and steward change in your system.* ​​​​​​​​​Organizational Training:​​​​​​​​ Everything you and your team need to cut through the tech-hype and implement strategies that catalyze true systems change. (For either Stewarding AI or Systems Change for Tech & Society Leaders)* ​​​​​​​​​1:1 Leadership Coaching:​​​​​​​​ I can help you facilitate change — in yourself, your organization, and the system you work within. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit untangled.substack.com

Hi there,Welcome back to Untangled. It’s written by me, ​​​​​​​​​​​Charley Johnson​​​​​​​​​​​, and valued​​​​​​​​​​​​​​​​​​​​ by ​​​​​​​​members​​​​​​​​ like you. ​​​​​​​Help me make it better?​​​​This week: my conversation with Helen and Dave Edwards of The Artificiality Institute on how AI affects your thinking, cognitive sovereignty, and staying the author of your own mind.As always, please send me feedback on today’s post by replying to this email. I read and respond to every note.On to the show!Deep DiveToday, I’m sharing my conversation with Dave and Helen Edwards, co-founders of The ​Artificiality Institute​, which helps people stay human in the age of AI. They also happen to be happily married — a fact that mattered in our conversation. Because it starts with what Helen calls an unhappily married moment. She had spent hours generating frameworks with Claude, feeling productive and energized the whole time — until Dave asked where she’d gotten to, and she couldn’t explain any of it. What happened next anchors our entire conversation.Along the way, we cover a lot of territory:* Helen offers three changes to watch for in yourself: blending, bonding, and bending. Blending is when you can’t quite tell which ideas are yours. Bonding is when the tool starts shaping your identity. And bending is when your frameworks for meaning-making shift, when what you believe you can become starts to change. None of these is bad on its own, she argues. But bending governs the other two because when your meaning-making shifts, it starts deciding what you notice and choose before you’re aware anything happened.* Cognitive sovereignty, in their telling, is three practices: noticing, choosing, and showing up. Noticing is real hard — “we’re wired to automate,” as Helen put it, and metacognition is “a devil to teach.” Her practical entry point, ​which resonates with my ​own, is to notice which role you’ve put the AI in: are you asking it to frame, to catalyze, to collaborate, to take something off your plate entirely? Choosing is about keeping the stakes in view — Helen’s point is that we choose more actively when the stakes stay front of mind, and drift when they fade -- and showing up is about remaining accountable to one another.* People with intuition for each other can do what the machine cannot. Dave’s side of the opening story is my favorite part of it. He looked at Helen’s tangle of slides and saw, in minutes, that she’d already figured it out — three dimensions, each with a high and a low, “a two by two by two” — she just hadn’t assembled it. He could see that because he knows how her mind works, the context of the presentation, and the purpose of what they were there to do. The machine had none of that context.* Frictionlessness erodes collective intelligence. There’s nothing wrong with friction. As Helen reminds us, synthesis is thesis plus antithesis — it requires struggle, turn-taking, and what she calls “mental maps of who knows what.” Dave adds the organizational history: decades of believing all the answers were in the data created organizations that ​value data and technology and devalue fundamentally human work.​ Put individual AI silos on top of that culture and you get summarization dressed as synthesis. Not great.* The line that matters is reducible versus irreducible. A language model works on data that is already a representation of what humans do, so the real question is what parts of work can’t be turned into a representation without losing the thing itself. So they created a tool -- ​The Irreducible Complexity Index​ -- to help measure exactly this. I’ve been playing with it this weekend and have found it super interesting -- ​give it a look​!For all the frameworks in this conversation — and Dave and Helen are wonderful framework-makers — every one of them resolves, in the end, into other people. Helen’s story of losing her thinking begins with a machine but turns on Dave connecting the dots. The machine had the content; Dave had the context. Even their diagnosis of what’s breaking in organizations comes down to the same thing — synthesis needs people willing to struggle with each other’s ideas, not ten tabs of AI output waiting to be summarized. The risk may live in each of our heads but the safeguard, it seems, lives between us.The other thing I appreciated is the spirit of their whole project: none of this is inevitable. So much AI commentary asks you to pick a lane — the machines are coming for everything, or the machines will save us all — and both lanes have the same flaw, which is that there’s nothing left for you to do in them. Dave and Helen are building a third thing: vocabulary, frameworks, and a community for people who want to stay the authors of their own minds while working with these tools, not despite them.If you want somewhere to start, borrow the test from Helen’s story: the next time you finish a stretch of work with AI and you’re feeling productive, try explaining what you made to someone close to you. Whatever happens next will tell you whether you were doing the thinking — and noticing the difference is the whole practice, in miniature.Helen’s book, ​Stay Human: Authoring Your Mind in the Age of AI​, is free, you just have to register. And if you want to do this thinking in a room full of people doing the same,​ their summit ​is October 22–24 in Bend, Oregon. I went last year and it’s the good kind of mind-bending.Until next time,CharleyWork With Me​Here are 3 ways I can help:* ​​​​​​​​​Advising:​​​​​​​​ I can help you navigate uncertainty, make sense of AI, and steward change in your system.* ​​​​​​​​​Organizational Training:​​​​​​​​ Everything you and your team need to cut through the tech-hype and implement strategies that catalyze true systems change. (For either Stewarding AI or Systems Change for Tech & Society Leaders)* ​​​​​​​​​1:1 Leadership Coaching:​​​​​​​​ I can help you facilitate change — in yourself, your organization, and the system you work within. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit untangled.substack.com

Hi there,Welcome back to Untangled. It’s written by me, ​​​​​​​Charley Johnson​​​​​​​, and valued​​​​​​​​​​​​ by ​​​​members​​​​ like you. ​​​Help me make it better?​​​​Today I’m sharing my conversation with Angelica Quicksey, Managing Director of New_Public, about the rise of the agentic interface era, and how we might shape it.As always, please send me feedback on today’s post by replying to this email. I read and respond to every note.On to the show!Untangled HQ​Big update: I’m getting married next weekend! So I’m going to take a short break from Untangled, and I’ll be back in your inbox on June 21.In the meantime, don’t forget to sign up for the next Untangled community event on See the System — a one-hour workshop where we start not with the tool but with the system it would enter. Bring a specific use case you’re weighing, and leave with a map, a vision statement, and a Proceed/Pause/Decline decision you can actually defend.Deep Dive​We Don’t Have to Build the Filter Bubble of OneThis week I spoke with Angelica Quicksey, Managing Director of New_ Public, about their new report After the Feed: Trust, connection, and the next era of social technology — which argues that we’ve crossed into a new era of social technology, as consequential a shift as the move from newspaper editors to algorithmic feeds was fifteen years ago: the agentic interface era. Let’s dig in.New_Public’s whole orientation comes from urban planning — what physical public space can teach us about the digital kind — and early in our conversation Angelica described what algorithmic social media actually feels like: you wake up every morning in Times Square. Bright, loud, and engineered to separate you from your money and your attention. Even people who enjoy visiting Times Square don’t want to live there! And yet that’s the only public space the last fifteen years built for us — one deafening square, optimized to keep us standing in it as long as possible.The argument in After the Feed is that we’re being pulled out of the square, whether we like it or not. A few forces are doing the pulling at once.The first is that the feed is no longer where our social lives or our information diet actually live. People will still scroll — parasocial entertainment isn’t going anywhere — but the place we go to figure out what’s happening, what to think, what to do, is increasingly a chat with an agent. Think about that handoff for a second. It used to be Walter Cronkite. Then it was the algorithmically ranked feed. Now it’s a chat window built just for you, and nobody else.The second is that the big platforms are quietly falling apart anyway — not because anyone reformed them, but because AI broke the things holding them together. Harassment is happening at industrial scale. The genuine back-and-forth between people is drying up. Machine-generated slop is everywhere, and bots already make up the majority of internet traffic. The gardens are still walled, but the walls are crumbling from the inside.And the third is that, as engagement gets cheaper to fake, the metrics that used to signal real human attention stop meaning much of anything. Likes, followers, reviews — all gameable. So the scarce thing is no longer attention; it’s trust. New_Public has a nice term for what trust looks like once you try to make it operational: thick reputation. Not “10K followers,” but “contributed thoughtfully to this community for two years.” Not “verified,” but “vouched for by people I trust.”But being pulled out of Times Square is not the same as arriving somewhere good. Angelica named the failure mode hiding underneath the whole promise: the filter bubble of one. We leave the deafening square and we don’t get the online equivalent of parks and libraries; we each get an information world drawn so tightly around us that nothing is held in common anymore. The old filter bubble at least had other people in it. This one wouldn’t. And it’s the default outcome, not the worst case, if nobody designs against it.So the real question the report is asking isn’t what’s replacing the feed? It’s what do we want to build in the space the feed is vacating — before the defaults get set for us?And the hopeful part of New_Public’s answer is that the raw materials are suddenly cheap. The cost of building software has fallen off a cliff: a community platform for 500 people used to cost millions, and now you can stand one up for a few hundred dollars a month. The old logic that said a platform needs billions of users to be worth building simply stops applying. A neighborhood, a hobbyist group, a mutual aid network, a book club — each can finally have software built just for it. Thousands of small, purpose-built spaces, instead of one square for everyone.Which sounds lovely until you try to run one! Healthy communities don’t tend themselves; they’re held together by stewards — the people who set norms, welcome newcomers, manage conflict, keep the shared memory. It’s real labor, usually unpaid, and burnout is the most common reason these spaces collapse. So the obvious move is to hand the routine moderation work to an AI agent and free the human steward up for the hard stuff that really requires care.Perhaps, but Angelica pointed to research on call centers that complicates the whole thing. When you route the easy tickets to self-service and leave the humans only the hard ones, the humans burn out faster. It turns out the easy work wasn’t filler. It was rhythm. It was rest. Strip it away and you don’t always get a more strategic steward; you get an exhausted one.This is the question I keep finding underneath every “what can we automate?” conversation, and it’s the thread that ties the whole report together for me. We treat routine as fungible — the part we can safely lift out — when sometimes it’s exactly where judgment gets built, where a steward comes to know the texture of her own community. The friction wasn’t always a cost to be eliminated. Sometimes it was doing the work. So maybe the better question isn’t what can we hand off? It’s what is the rhythm quietly doing that we haven’t named yet?That, in the end, is what I admire about After the Feed. It isn’t a promise that things will work out. It’s that Angelica and her colleagues are doing the thing tech criticism has mostly refused to do for fifteen years: describing, in concrete terms, what it would look like if we got it right. Many small spaces built for actual communities, owned by their members, connected through open protocols so you can carry your history with you. AI working quietly in the background as a kind of shared memory, rather than running the show out front. Stewards supported, paid, and designed for. Parks and plazas and libraries — not one more Times Square.That’s a long way from where we are. But it’s worth knowing someone’s building toward it.Until next time,CharleyWork With Me​Here are 3 ways I can help:* ​​​​​​​​​Advising:​​​​​​​​ I can help you navigate uncertainty, make sense of AI, and steward change in your system.* ​​​​​​​​​Organizational Training:​​​​​​​​ Everything you and your team need to cut through the tech-hype and implement strategies that catalyze true systems change. (For either Stewarding AI or Systems Change for Tech & Society Leaders)* ​​​​​​​​​1:1 Leadership Coaching:​​​​​​​​ I can help you facilitate change — in yourself, your organization, and the system you work within.​​ This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit untangled.substack.com

Hi there,This week I’m sharing a conversation I had with ​Bo Young Lee​, CEO of ​AI4All​ about Silicon Valley imaginaries, rational refusal, and the futures we haven’t been offered. As always, please send me feedback on today’s post by replying to this email. I read and respond to every note.On to the show!Untangled HQ* Wednesday, May 5: I’m hosting a ​workshop on how to trace what must stay human ​when implementing AI responsibly. It will double as a preview of ​my new course on stewarding AI. ​* Thursday, May 6: As part of ​The Facilitators’ Workshop​, Kate and I are hosting a ​workshop on how to turn stuck meetings into breakthrough moments. ​* Tuesday, May 12: Aarn and I are hosting a workshop on the discipline of holding tension: how to name tension without personalizing it, slow the moment without stalling the meeting, and protect the disagreement that actually matters. Join us!Deep DiveThe World They’re Building TowardStart with the bunkers.In the last several years, a number of Silicon Valley’s most powerful technologists have been quietly building survival infrastructure. ​Bunkers in New Zealand.​ ​Fortified compounds in remote locations.​ Escape hatches from the civilization their products are shaping.Bo Young Lee noticed this before most people were talking about it, and she asked the obvious question: if these are the imaginaries — the foundational visions of the future — animating the people building our most consequential technologies, what does that tell us about the products they’re building? And how does their imaginary constrain our imagination?An imaginary is not a fantasy. It’s the operative picture of the future that structures present decisions — the unstated assumptions about where the world is going that determine what problems are worth solving, what risks are worth taking, and what populations are worth designing for. Imaginaries are embedded. They show up in product decisions, in hiring, in what gets funded and what gets ignored.Bo argues that the dominant Silicon Valley imaginary is, at its core, a story about inevitability and survival. Civilization is fragile. Disruption is coming. The question isn’t whether things collapse but who gets to build what comes next. If that’s the picture of the future you’re working from — even unconsciously — you’re not going to prioritize safety, privacy, or good governance in the present. Those things just get in the way!As Bo explains, the products that follow are predictable. Why design for women when women don’t figure prominently in survival scenarios? Why prioritize people with disabilities when they’re among the first casualties of disaster-oriented futures? Why hold yourself accountable to the communities your technology harms when they’re not in the imaginary?This isn’t hyperbole. Bo is describing a logical coherence between worldview and product — a through-line from the bunker to the algorithm that becomes visible once you start looking for it. Take the supposed ‘​AI gender gap.​‘ The narrative goes something like this: women are underrepresented in AI adoption because they lack confidence, access, or awareness. All we need to close the gap is a li’l education, outreach, and encouragement! Bo argues that women’s skepticism about AI is rational. Not because women don’t understand the technology, but because they understand it clearly enough to recognize that it wasn’t built for them, doesn’t work as well for them, and in specific contexts actively harms them.Right, women face ​systematically harsher​ professional consequences than men for identical workplace errors — a well-documented asymmetry researchers call the “​tighter world​” phenomenon. Women are more likely to be fired for mistakes and less likely to find subsequent employment. When a high error rate tool like generative AI enters that context, the risks land differently. Men’s mistakes get absorbed as the cost of experimentation. Women’s mistakes land on a narrower margin. A woman who understands this and proceeds with caution is doing the math. Calling that a confidence problem is its own kind of imaginary!The “AI for good” movement is similarly trapped by the Silicon Valley imaginary, but they don’t see through it in the same way. As Bo argues, the AI for good world has largely accepted the imaginaries it inherited. Its animating question is how to reduce harm within the existing AI paradigm — how to make the technology that’s been built safer, fairer, less biased. For example, Bo describes a philanthropy that funded three separate organizations — at seven-figure grants each — to build AI agents that would coach and tutor low-income, first-generation college students. The goal was equity. But research shows that when you train LLMs to eliminate overt racism, the covert bias doesn’t disappear — it actually increases. Show the same model two pieces of writing, one in standard English and one in African American Vernacular English (AAVE), and the LLM will rate the AAVE writer as less intelligent and less educated. A coaching agent built on that model, deployed to help first-generation students many of whom communicate in AAVE, may well steer those students toward easier majors and less rigorous courses — without anyone noticing, without anyone intending it.This example starts from a present-tense imagination of what AI is and what it’s for, and works forward from there. To free ourselves from these constraints, we have to separate refusal of this AI from refusal of AI altogether. Because when we do that, we can ask the more generative question that rarely gets asked: what futures do we actually want — and what would it take to build toward them?Bo’s organization offers one path forward. AI4All trains the next generation of AI practitioners from underrepresented communities, asking them from the beginning to identify social problems they want to address and work backward to the role AI might play. Because changing the imaginaries requires changing who builds the technology and who gets to define what it’s for. A more diverse AI workforce is an epistemic necessity — different people imagining different futures producing genuinely different technology.We were not given these imaginaries. We don’t have to keep them.Tools for WeaversMy conversation with Bo inspired me to distill a number of the articles I've written about ​imagination​, ​building alternative AI futures​, and ​mapping backwards from the future​ -- and turn them into a tool!Your strategy documents already contain a picture of the future. You probably haven’t named it. It’s embedded in your metrics, your hiring plans, your roadmaps — quietly nudging you toward a particular kind of future without anyone actively choosing it.Imagining Otherwise is a practice for naming that picture — and then building a different one. Backcasting, futures in plural, and the question most teams skip: what are we willing to stop?Working canvas included. The last page will make sense when you get there.“Remember to imagine and craft the worlds you cannot live without, just as you dismantle the ones you cannot live within.” - Ruha BenjaminWork With MeHere are 3 ways I can help:* ​​​​​​​​​Advising:​​​​​​​​ I can help you navigate uncertainty, make sense of AI, and steward change in your system.* ​​​​​​​​​Organizational Training:​​​​​​​​ Everything you and your team need to cut through the tech-hype and implement strategies that catalyze true systems change. (For either Stewarding AI or Systems Change for Tech & Society Leaders)* ​​​​​​​​​1:1 Leadership Coaching:​​​​​​​​ I can help you facilitate change — in yourself, your organization, and the system you work within. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit untangled.substack.com