Matt Smith

Was thrilled to host a recent OhioX webinar on Agentic AI in Production with leaders from Foundry for AI by Rackspace (FAIR™). here is a link to the webinar - https://lnkd.in/gcVGxwsF . My 3 biggest takeaways from our discussion: 1. The Infrastructure Reality Check Our panelists made it clear: agentic AI requires fundamentally different plumbing. We're moving from point-to-point APIs to protocols like Model Context Protocol (MCP) that let agents communicate across systems seamlessly. As one panelist noted, it's like giving an employee access to all the tools they need—but we must carefully orchestrate that access. 2. The "Policy as Code" Imperative What fascinated me most was hearing about organizations writing separate rollback agents for every production agent they deploy. The message was clear: policies that exist in documents must become executable code with built-in guardrails. We can't scale agentic systems without this foundation. 3. The Human Factor Surprise Perhaps the most eye-opening moment: our internal CTO revealed that human resistance was a bigger challenge than any technical hurdle. People hold onto familiar manual processes. Success requires intentional change management—not just technical deployment. One statistic that stopped me: 77% of AI leaders already have agentic AI in production vs. much lower adoption elsewhere. The leaders are pulling away fast. Thanks to Simon Bennett, Nirmal Ranganathan, and Brian Lichtle for sharing their authentic experiences. And to everyone who joined from across the globe (we had 4 continents represented)—your questions made this conversation richer. Thanks to Chris Berry for your partnership The future isn't about automating what we do—it's about reimagining what's possible.

21

Dashboards don’t create accountability. They create meetings. Charts get shared. Everyone nods. Action items get “noted”. Same problems. Next quarter. Accountability doesn’t come from visibility. It comes from systems — where ownership is clear and decisions can’t be ignored. I wrote a short note on why dashboards feel productive but rarely change behaviour — and what actually does. If your org has great dashboards but the same issues every review… this will sound familiar. Dashboards Don’t Create Accountability — Systems Do: https://lnkd.in/gCGpm-VU

7

1 Comment

AI Builders are shifting 100% of the risk to the rest of us. It is time for a new deal. Rujuta Singh's post about the Air Canada ruling hits home for anyone actually trying to deploy AI in the real world. As practitioners, we are being put in an impossible position. We want to use these powerful tools, but the current "terms of service" are basically a giant "good luck" from the builders. If we want AI to be more than a risky experiment, we have to address three practical gaps: 1. The Warranty Gap: Right now, if an AI "hallucinates" and costs your company $100k, you pay. The company that sold you the API pays nothing. We wouldn't accept this for a database or a cloud server—why are we accepting it for AI? 2. The "Black Box" Liability: It is hard to build safety guardrails when the model builders won't show us how the engine actually works. We are being asked to "trust the process" while carrying all the legal weight when that process fails. 3. Cleaning Up the "Pollution": Every time a model generates a false policy or bad advice, it’s an informational mess. The builders make the profit from the "generation," but the practitioner spends the hours (and the legal fees) cleaning up the mess. The Bottom Line: We need to move away from "use at your own risk." We need builders to take on a "Duty of Care"—meaning they share the responsibility for the outputs their models produce. Until the people building these systems have real skin in the game, practitioners will always be the ones left holding the bag. #AI #TechImplementation #BusinessRisk #AIStrategy #PractitionerPerspectives

6

3 Comments

🧵 Stitch Fix: Elevating the Art + Science of Styling with Knowledge Graphs For years, Stitch Fix has been the industry benchmark for blending human intuition with data science. But in a world of rapidly evolving AI, how do you stay ahead of the curve and continue to inspire individual style journeys? I had the pleasure of sitting down with Meg Paulosky, Director of Data and Analytics at Stitch Fix, at the last Open Data Science Conference (ODSC) to discuss how they are using graphs to move beyond traditional data models. "Exploring not only AI, but knowledge graphs has really helped us reimagine what [the style journey] looks like in a way that works for everybody... " Three Takeaways from our Conversation: 1️⃣ Architectural Innovation: Shifting the focus from basic data points to sophisticated Agent Design and systems modeling. 2️⃣ Universal Personalization: How graphs allow Stitch Fix to scale style inspiration beyond their existing clients and services. 3️⃣ The ODSC Pulse: Why it's essential to meet other innovators and see tangible, cross-industry AI applications in person. It was incredible to get a few minutes with such a forward-thinking leader at one of the best data science events. (Next ODSC is in April!) 🎧 Listen to the Graph Chat here: https://lnkd.in/etUqtVWW 🎬 Watch here: https://lnkd.in/ePxSzutJ

42

2 Comments

Just kicked off a new paper serialization: FallacyTag — on surfacing fallacies and reasoning structure in the wild. Two posts are live: 1. Why I’m Writing About Fallacies in Public 2. The Architecture of Avoidance (The problem statement — and a whiff of solution.) Feedback welcome. ToC: https://lnkd.in/ge-H3Dcy

18

1 Comment

Smart glasses just went from sci‑fi to SKU. A new roundup compares the most compelling pairs you can actually buy now—think Ray‑Ban Meta, Amazon Echo Frames, Xreal Air 2, Rokid, Solos—and the trade‑offs driving adoption. Why this matters: we’re watching the next computing surface area take shape. Cameras vs. displays, battery vs. comfort, on‑device vs. cloud AI. The winners won’t just be the best hardware—they’ll be the best workflows for hands‑free, glanceable computing. - 💡 What this means for founders: Build “jobs-to-be-done” use cases (capture/recall, translation, real‑time guidance, remote assist). Latency and battery are your north stars. - 📌 For product teams: Design voice‑first and glanceable UX with subtle haptics. Plan for on‑device models and privacy by default—ambient computing demands trust. - 📉 Market signal: Distribution + fashion matter. Meta/Luxottica are pairing brand and channels with AI features; AR displays are improving but comfort and price remain barriers. If you were building one killer feature for smart glasses today, what would it be—and who’s the first ICP you’d target? #StartupNews #SmartGlasses #Wearables #SpatialComputing #AI #AR #Founders #ProductManagement #VentureCapital

2

Are you using Loom to record product content for users? 🎬 If so, quick question: How are you deleting snippets that shouldn't make the final cut mid-recording (when the recording is paused)? Or are you editing mistakes after recording the full clip? (Asking for a friend) 🙃 Told this to my teammate who responded with: "yeah sometimes after a couple of takes I want to throw my phone across the room" Thankfully, laptop's still in one piece... To my Ora users, the tutorials you see in product were done in one take ( I swear) 😉

19

3 Comments