Runpoint: AI Business Transformation Podcast

Giving every employee an AI license feels fair and safe. Mostly it multiplies whatever judgment was already there.By Sam Gaddis. From Issue Two of Runpoint Magazine, Half-Lives. Read by an AI narrator.Read the piece at runpointmagazine.com/issue-two/arm-the-few, and order the printed issue at runpointmagazine.com.

American factories had electric motors for forty years before the productivity showed up. The wait was mostly about the buildings they were bolted into.By Matthew Hall. From Issue Two of Runpoint Magazine, Half-Lives. Read by an AI narrator.Read the piece at runpointmagazine.com/issue-two/group-drive, and order the printed issue at runpointmagazine.com.

Most of what executives track about AI goes stale inside a quarter. The useful skill is sorting what you know by how fast it decays, and checking each pile on its own clock.By Sam Gaddis. From Issue Two of Runpoint Magazine, Half-Lives. Read by an AI narrator.Read the piece at runpointmagazine.com/issue-two/half-lives, and order the printed issue at runpointmagazine.com.

This year we gave up on quarterly goals. Then we started a quarterly magazine.By Sam Gaddis. From Issue Two of Runpoint Magazine, Half-Lives. Read by an AI narrator.Read the piece at runpointmagazine.com/issue-two/the-wrong-unit, and order the printed issue at runpointmagazine.com.

Sam Gaddis sits down with the largest Runpoint Podcast panel yet: Matthew Hall and Ryan Mish from the Runpoint team, Thomas McNally of Zaelab, and returning guest Thanh Pham.The conversation starts with the Theo video that stirred up the "local models are overrated" argument and goes deep from there. Thomas walks through what it actually looks like to deploy open weight models across a firm, including the cost math, the hardware, and why non-technical employees are lining up to join the program. Thanh brings the hobbyist-turned-practitioner view from his home lab and where local genuinely holds up. Along the way: the four-way split between frontier, local-on-prem, local-in-cloud, and rented open weight inference, plus a straight look at the two fears CEOs raise most often.If you run a mid-market company and you're trying to figure out where these models fit, this one is for you.Guests:Thomas McNally, technology lead at ZaelabThanh Pham, Managing Director of Asian EfficiencyMatthew Hall and Ryan Mish, RunpointTimestamps:0:00 Intro and the panel0:53 The Theo video and the local vs open weight debate1:22 Deploying open weight models across a firm: the cost story4:41 Which models and hardware Zaelab landed on (Qwen, RTX 6000, VRAM)6:41 Why employees want in on an "inferior" model program8:00 The interface problem and Anthropic's third-party inference feature9:28 Pay-per-use to all-you-can-eat, and protecting your IP9:56 The subscription gravy train is ending10:21 AI sovereignty and why the conversation shifted12:11 Thanh's home lab and the hybrid local setup13:41 What "most people" can actually run on their hardware16:03 The routing trifecta: frontier, cloud infrastructure, and on-device18:32 What each panelist actually uses day to day19:30 Model agnosticism and avoiding vendor lock-in21:40 How to build an intelligent routing layer24:14 Evaluating routing solutions without a lab24:54 Myth busting: Chinese models and training on your data27:18 Where IP really lives for mid-market companies28:10 The fourth option: rented open weight inference (Fireworks, GLM)30:00 Clearing up the terminology confusion32:52 Does fine-tuning transfer, or is it a one-time cost?35:38 The maintenance burden nobody talks about38:22 Tempering expectations if you're coming from frontier39:19 Wrap up

AI agents are no longer just standalone demos. They are moving directly into the systems where work already happens: Google Workspace, SAP, Salesforce, email, calendars, files, CRMs, and enterprise systems of record.In this episode of the Runpoint Podcast, Sam Gaddis, Matthew, Ryan Mish, and guest Harrison Wells of Dodo Digital discuss the shift from flashy AI tools to operational AI workflows. They cover Google Spark and Omni, the SAP + Anthropic partnership, Salesforce’s reported engineering productivity gains, Pi agents, Codex 5.5, Claude, and why many companies are increasing AI budgets faster than they are building true AI readiness.The big question: will enterprises actually transform around AI, or will they just bolt chatbots onto old systems and call it innovation?Timestamps00:00 — Intro: agents are moving into real work surfaces02:18 — Pi agents, Codex 5.5, and custom AI workflows06:00 — Google Spark, Omni, and the Workspace distribution advantage12:26 — Could Google own the enterprise AI surface?15:19 — Agents making purchases and real-world decisions21:03 — SAP + Anthropic: Claude inside systems of record23:25 — Why chatbot-style enterprise AI often fails25:30 — Daily AI reports, heartbeat workflows, and what actually works31:44 — Salesforce’s AI engineering productivity claims37:21 — AI budgets are rising, but readiness is lagging39:48 — How companies should build real AI adoption42:31 — Why engineering is ahead of the rest of the enterprise43:21 — Closing

Runpoint Podcast E11: Code with Claude, the SaaSpocalypse, and AI Consulting Goes MainstreamBack after a couple months off, with a bigger crew this time. Sam and Matthew are joined by Ryan Mish, one of Runpoint's operator engineers, and Thanh Pham of Signal Advisory.Matthew just got back from Anthropic's Code with Claude conference and gives a side-by-side of the vibe at Anthropic vs. OpenAI (he hit both in the same week). From there we get into the stuff that's actually on our minds:The flood of new "AI consulting firms" charging $20K to install Claude Code and call it a deployment, and what separates real work from theaterWhether the SaaSpocalypse is real, when it makes sense to rip out HubSpot, and where SaaS still wins (network effects, niche infrastructure, the school district that no VC will ever touch)Why mid-market companies often only use 10% of the software they pay for, and what changes when the data finally talks to itselfGetting good design out of LLMs: Claude's house style vs. Codex 5.5 with image-gen mockups, and why Claude Design is still a prototyping toolClosing round of what each of us is using right now: managed agents, Pi as a harness for steering Claude and Codex together, a PNPM supply chain attack PSA, HTML artifacts instead of Markdown, and a $30 conference swag computer turned into a reading game for Matthew's daughterChapters00:00 Intro and new faces01:30 Matthew's report from Code with Claude05:30 Anthropic vs. OpenAI, in person10:15 The AI consulting gold rush14:30 Change management and company size30:00 Is the SaaSpocalypse real?40:30 Replacing software you only use 10% of45:00 Getting good design out of LLMs52:00 What we're using right now: managed agents, Pi, PNPM, HTML artifacts, and a tiny computerGuestsThanh Pham, Signal Advisory

Happy New Year! In this episode, Matthew and Sam break down what actually happened in AI during 2025 and make predictions for what's coming in 2026.We cover the biggest winners and losers, the apps and tools that changed how we work, who to follow for smart AI takes, what surprised us most, and where we think things are headed.Key topics:• Why companies that stayed curious won 2025• OpenAI's fall from dominance to a three-horse race• Claude Code and why it changed everything• The tools we actually use daily (Whisper Flow, Granola, Cursor)• Why vibe coded apps are passing enterprise code reviews• Google's image generation breakthrough• Our custom CRM build and the future of back office AI• Why product managers are the winners of 2026• The coming downfall of SaaS and low-skill tradesPlus: We launch our new Run Point Magazine and Sam accidentally emails hundreds of people.🔗 Get the Run Point Magazine: [link]🔗 Subscribe to our newsletter: [link]---## Chapters0:00 - Intro & Happy New Year0:39 - Biggest Winner of 2025: Curious Companies2:16 - Claude Code Changed Everything4:04 - Biggest Loser of 2025: AI Resisters5:25 - Why OpenAI Lost Ground in 20257:07 - Best App of 2025: Claude Code & Whisper Flow9:17 - Granola, Transcripts & the Transcript-to-Action Pattern10:29 - Why Cursor Won the IDE Wars11:33 - Best Thinkers to Follow: Tyler Cowen & Twitter Curation14:40 - The Run Point Magazine Launch (and Email Disaster)17:17 - Biggest Surprise: Enterprise-Grade Vibe Coded Apps20:35 - Google's Image Generation Breakthrough22:53 - Best Thing We Built: The AI-Powered CRM27:08 - Worst Things We Built: Complex RAG Systems28:52 - 2026 Predictions: Product Managers Win Big30:37 - 2026 Loser Prediction: Half of SaaS Dies32:28 - Why Low-Skill Home Services Are in Trouble35:04 - AGI Is Basically Here (Contrarian Take)38:10 - Why "Vibe Coding" Will Get Rebranded38:38 - 2026 Resolutions: Back Office AI & Continuous Learning41:01 - Wrap Up

Learn how to save on your AI initiatives. On the Runpoint podcast, hosts Matthew Hall and Sam Gaddis welcome Ari Salafia, CEO and founder of TaxTaker. We explore how operator-engineers can claim R&D tax credits to ship working AI systems and significantly reduce development costs.Chapters:00:00 Welcome Ari Salafia of TaxTaker01:00 Understanding R&D Tax Credits: TaxTaker's Mission02:30 Why Businesses Should Care About R&D Tax Credits Now04:30 What You Stand to Gain: Key Expense Buckets08:50 The Difference: Dollar-for-Dollar vs. Percentage10:00 Qualified Research Expenses (QREs) Explained11:00 Project Qualification: The Four-Part Test14:00 Internal Use Software: Additional Requirements15:45 Defining "Innovative" for Tax Credits16:30 Case Study: Custom CRM Development19:15 Custom Configuration vs. New Development20:00 Documenting Projects for R&D Credits21:50 Structuring Contracts with AI Consultants (like Runpoint)25:00 Case Study: AI for Business Intelligence Tools27:10 Case Study: AI for Resource Allocation28:15 Case Study: AI-Powered New Service Offerings30:45 Advice for Employees Seeking Project Approval35:40 Addressing Global Talent in R&D Claims38:00 Connect with TaxTakerKey Takeaways:Measurable Impact: Claim up to 10% of your AI development spend back in R&D tax credits.Pilots to Production: Understand how to qualify your projects, from custom builds to new AI service offerings.No BS: Learn direct strategies for contract structuring (IP retention, financial risk, US-based work) to maximize your credit.Value Creation: For early-stage companies, credits can offset payroll taxes; for profitable companies, income tax.Forward-Deploy: Empower internal champions to make a strong business case for AI initiatives by highlighting potential savings.Connect with Runpoint: We partner with executives but sit with end users, moving clients from pilots to production with measurable impact.See a 2-week pilot plan for your next AI project: https://runpoint.ai/Learn more about R&D tax credits and TaxTaker: https://www.taxtaker.com/Tags:#AIOps #AIinProduction #RandDTaxCredits #TaxTaker #Runpoint #AISavings #BusinessFinance #Innovation #TechTax #AIImplementation #OperatorEngineers #AIStrategy

Matthew Hall and Sam Gaddis break down the latest in AI, challenging the "95% failure" narrative with new ROI data and dissecting ChatGPT's "Atlas" browser launch. They codify emerging best practices for AI workflow automation, advise on navigating the crowded AI coding assistant market, and celebrate AI's power to enable entirely new work. The episode culminates in a crucial discussion on the diminishing value of software in acquisitions and what truly constitutes a moat in the age of AI.Chapters:00:00 AI's Evolving Landscape: Successes and Failures04:00 ChatGPT's New Browser: A Game Changer?07:33 Best Practices for AI Workflow Automation12:00 Navigating the AI Coding Assistant Market16:55 New Opportunities: AI Empowering New Work21:28 Valuing Software in the Age of AIKey Takeaways:AI ROI is proving positive for most firms, with a new study challenging older "failure" statistics.ChatGPT's new "Atlas" browser shows potential but needs deeper context integration to become a true game-changer.Effective AI automation prioritizes breaking workflows into atomic steps, automating deterministic parts, and using single agents for judgment with human oversight.Distribution, data, and brand are the new moats; the value of software itself is rapidly diminishing in the age of AI.AI empowers "net new" work, allowing individuals to tackle tasks and projects they previously wouldn't have attempted.Tags:#AI #ChatGPT #Automation #EnterpriseAI #AICoding #SoftwareValuation #Podcast #RunpointPodcast