The Innovation Crowd with Helen Dawson

Most buildings heat, cool and ventilate to a timetable.Steve Burch's company teaches them to read the room.The technology was the easier piece. Getting the six or so people who can veto it to say yes took longer.Steve is co-founder and Chief Operating Officer of Envelo, which uses building data to save energy and keep indoor air healthy. He built consulting businesses before co-founding Envelo during the pandemic. We talk about what changed when he moved into a product business, and what Envelo found at King's Cross.We get into:- Trusting your gut, long-term relationships, and the clarity a startup forces on you- How a pandemic air-quality idea turned into an energy business- Integration, and why a pilot on its own dashboard never scales- About six decision-makers per building, and finding the ones who want a better way- King's Cross: 31% less ventilation energy on top of an already efficient building, independently validated- Where AI fits: fine-tuning inside guardrails, and chatting with your building- Five people, five different meetings, and why alignment sets the pace- Difficult conversations lead to an easy lifeConnect with Steve: https://www.linkedin.com/in/steve-burch-b581591/Envelo: https://envelo.solutions/The King's Cross case study (Havas, 3 Pancras Square): https://envelo.solutions/case-study/Connect with Helen: https://www.linkedin.com/in/helen-dawson/The Hard Part newsletter: helen-dawson.kit.com/the-hard-partAI Confidence Snapshot (4 mins): scorecard.helen-dawson.com

What happens when a virtual human becomes one of your most effective communicators?Shell needed ten training films that would help its global sales team talk confidently about lubricants for off-highway machinery. The budget was limited, the presenter was a real engineer who isn’t a natural on camera, and the audience speaks many languages. Shell’s in-house film agency, Content Engine Studios, answered with Project Elevate: an AI digital twin of the engineer, Rob, which presents the series in languages Rob doesn’t speak.Jane Sayers, Global Discipline Lead for Film and Photography at Shell, explains how the films were made and what it took to get them approved. Filming all ten episodes took a day and a half. The approvals added four to six months.In this episode:Building a digital twin from a 3D scan and 20 minutes in a studioWhy they kept a trace of Rob’s Canadian accent in the Mandarin versionWhat consent, data security and the right to withdraw look like when your own employee’s likeness is the dataDisclosing the AI on camera instead of with a disclaimerWriting the script after filming, and why that changes productionHow doing it once opened the door for other teams at ShellJane’s advice to any team thinking of doing the same: start with the problem, not the toolThe project has won six awards, including the Grand Prix at the inaugural AI Communications Awards.Best watched on Spotify video as Jane shares behind-the-scenes footage and clips from the finished films and seeing is believing.Links:Shell Content Engine Studios on LinkedIn: https://www.linkedin.com/company/shell-ces/home/Jane Sayers on LinkedIn: https://www.linkedin.com/in/janesayers/Get in touch with HelenEmail: [email protected]: helen-dawson.comLinkedIn: linkedin.com/in/helen-dawsonNewsletter, The Hard Part: helen-dawson.kit.com/the-hard-partThis episode was first broadcast as a LinkedinLive on 16 October 2026.

The tell tale signs of fantasy business casesHelen Dawson breaks down why oversized TAM slides and polished strategy decks often hide a lack of real market diagnosis. This episode is a sharp rant about replacing fantasy sizing with fieldwork, real users, and a clear go to market wedge that can actually move.Key topicsFantasyland business cases are usually built on meaningless TAM numbersWhy TAM, SAM, and SOM can still miss the pointDesktop research is not strategyMarket sizing should name real humans and job rolesThe real question is where the wedge of value isFieldwork beats armchair assumptionsBring customers into the room earlyYou are upgrading the user, not the technologyFantasy business cases may look impressive, but they avoid the real work

Helen Dawson explains how corporate leaders can build a stronger AI adoption team by looking inside the organization rather than rushing to hire externally. The episode focuses on three often-misunderstood employee types that can become key drivers of AI success: shortcut-seeking people, risk-aware skeptics, and unconventional thinkers.Key topicsIn this episode: Helen frames AI adoption as a team-building challenge, especially for leaders asking who should help drive AI in their organization.The first overlooked group is the "lazy" people - the ones who love shortcuts and friction reduction. Helen argues they are often the best at spotting practical AI use cases because they naturally look for ways to save time.Shortcut-minded employees can become power users - Helen notes these people may spend a weekend building something with AI if it saves them an hour every day.The second group is the "awkward" people - the ones who raise objections, point out governance concerns, and explain why things might fail. Helen reframes them as risk managers rather than pessimists.Leadership’s job is to coach the awkward voices so they can present risks and red flags in a more constructive, politically workable way.The third group is the "wacky" people - unconventional thinkers who come up with off-the-wall ideas and spot blind spots that others miss.Wacky thinkers help break groupthink - Helen contrasts their originality with the limits of using an LLM for brainstorming, which can stay too close to common knowledge.These three types together form a practical AI adoption team - shortcuts, risk management, and original thinking create a balanced internal capability for scaling AI.

For years, customer service teams benefited from an invisible ally: Customer effort.Many consumers simply gave up. The complaint took too long. The form was too complicated. The process was too frustrating. That safety net is disappearing. In this episode, Helen explores why consumer AI agents are changing the economics of complaints and what this means for customer service leaders.Key discussion points:• Why complaint volumes are likely to rise• How AI removes effort as a barrier• Why customers will become harder to dismiss• The rise of AI-generated complaint correspondence• Why "bot fighting bot" is the wrong strategy• The enduring importance of service recovery• Why fixing root causes beats handling more volume• The principle of "subtract, don't add"If customers now have a tireless digital advocate, what should organisations do differently?This solo episode explores five practical truths every customer service leader should understand.

Failure is not an option when aircraft need to stay the right way up.So how does the most safety-obsessed industry on earth adopt AI?Carefully. And more ambitiously than you'd think.Jonathan Burnip is Head of Advisory Services at Marshall, the Cambridge aerospace and defence firm, where he's spent his whole career from placement student to CTO to advisory. His argument: the regulators aren't the blocker, the data isn't what outsiders assume, and the real line to hold is that AI mustaid decisions without ever undermining the industry's ability to learn from what goes wrong.We get into:- The two things outsiders get wrong about AI in aerospace — regulators, and data- The sparse-data paradox: measuring everything, designing to one-in-a-billion failure rates- How data sharing really works in defence — assured systems, trusted people, and why two harmless numbers can add up to a classified one- eVTOL and why this feels like another golden age of aviation- Just culture: the safety principle that should set the terms for AI adoption everywhere- The sticky note on the steering wheel — how aids quietly become decision-makers- Coach, don't show: why an AI that prompts you beats one that demonstrates — and what the FAA's human-factors guidance says about skills fade- What AI does to the early-careers engineering pathway, and the skills that remain- Jonathan's hard-way lesson: you don't have to know everythingConnect with Jonathan:https://www.linkedin.com/in/jonathan-burnip-42373920b/Connect with Helen: https://www.linkedin.com/in/helen-dawson/The Hard Part newsletter: helen-dawson.kit.com/the-hard-partAI Confidence Snapshot (4 mins): scorecard.helen-dawson.com

Mark Zuckerberg recently published a lengthy vision for a future of "superintelligence in every pocket".But before business leaders accept that vision, there's a more important question to ask:Who marked the homework?In this solo episode, I take a sceptical look at one of the most influential AI manifestos of the year and explore what happens when technology vendors define the problem, propose the solution, and assess their own success.I agree with some of Zuckerberg's thinking, particularly the idea that AI should expand human invention rather than simply automate existing work. But I challenge the assumptions around safety, governance, organisational accountability and the claim that making powerful AI available to everyone automatically makes it safer.Along the way we'll examine:Why access to AI tools is not the same as creating business valueThe growing problem of shadow IT and unmanaged AI systemsWhat AI governance actually looks like inside large organisationsWhy vendor incentives matter more than vendor promisesHow to spot technology marketing disguised as philosophyThe difference between open access and responsible deploymentWhat leaders should be asking before approving major AI investmentsMost importantly, I share the three questions I believe every executive team should ask whenever they are presented with an AI strategy, manifesto or sales pitch:If we do everything this recommends, who gets paid?Which part of this costs the author something?Who set the homework, and who marked it?If you're responsible for AI strategy, digital transformation, technology investment or organisational change, these questions may save you from making expensive mistakes.Because in AI, as in school, nobody should be allowed to mark their own homework.

Engineers can't sign their name to a guess.That's why generative AI stalled at the door of the control room.This episode is about what got through instead.Andy Webster is Senior Director for Digital and AI at KBR's Sustainable Technology Solutions division. His team spent the last three years going from an organisation curious about everything to one focused on a handful of bets: culling projects on a six-week clock, then backing physics-based AI hard enough to invest in it. His argument: language models predict the next word, physics-based AI predicts the next state of your asset, and that difference is what earns an engineer's trust.We get into:- The "difficult, if not traumatic" project cull — and why stopping things created the velocity- The six-week rule: every experiment defaults to stop unless it earns another six- Physics-based AI explained properly: prediction constrained by gravity, thermodynamics, reality- Why KBR went from reseller instinct to investor in Applied Computing, and what an ecosystem play looks like in engineering- "What if your plant could think?" Talking to a physical asset like a colleague- Souls on steam trains: what the history of tools says about the fear of this one- Will professional bodies end up approving the data sets we trust? The next fight over "data is the new oil"- Andy's hard-way lesson: the person saying no is as smart as youConnect with Andy: linkedin.com/in/andywebster1Connect with Helen: linkedin.com/in/helen-dawsonSign up to The Hard Part newsletter: helen-dawson.kit.com/the-hard-partAI Confidence Snapshot (4 mins): scorecard.helen-dawson.com

Everyone tells you to start your AI journey with a use case. I'm going to tell you to start with a notebook, a coffee, and a darkened room.A solo episode: the two questions I give leaders to kick-start their AI thinking. No tools, no vendors, no pilots. Twenty minutes of honest thinking before you spend a penny.We get into:Question one: what's getting in the way right now — for you, your peers, and your team? Be honest. That list is your leadership to-do list.Why "no use case is too small" is bad advice: AI should move the needle on your business, not do your conditional formattingQuestion two: wouldn't it be cool if? Start with a problem you care about solving, and think six months ahead of today's capabilityWhy the quick win is a trap, and what to aim for instead: a shortcut to alignment on the big ambitionAI is an ingredient, it's not a mealConnect with Helen: linkedin.com/in/helen-dawsonThe Hard Part newsletter: helen-dawson.kit.com/the-hard-partAI Confidence Snapshot (4 mins): scorecard.helen-dawson.com

Nuclear is about as unforgiving as engineering gets. Safety critical. Heavily regulated. So why does Graham Openshaw think it should be innovating faster, not slower?Graham is Operations Director, Digital at Assystem, the second-largest independent nuclear consultancy in the world. His argument: proper governance and sponsorship aren't the brakes. Done well, they're what lets you manage riskand move faster.We get into:The myth that people resist change, and why the frustration sits in how organisations are wired, not in the individualsThe power of no: how every approval step hands somebody a veto they learn to enjoy, and how to make no the expensive answerWhy a pickup truck explains everything wrong with corporate procurementThere is no digital strategy. There is just strategy. And often the missing piece of the puzzle is defining outcomesTechnical debt is compromise with bad brandingWhy internal projects deserve the same rigour as a client tenderAI as a lever for quality improvement, because the expert becomes the checker, not the authorGraham's hard-learned lesson: thinking is workingConnect with Graham: linkedin.com/in/grahamopenshawConnect with Helen: linkedin.com/in/helen-dawsonSign up to The Hard Part newsletter: helen-dawson.kit.com/the-hard-part