The Forward Slash Podcast

What happens when your biggest customer isn’t a person — it’s an AI agent? Tommy Cotter has spent five years building Benzinga’s API business, turning market data and news into products others build on. He and James Carman get into designing data for machines, not dashboards: exposing feeds through an MCP server, structuring tools so agents don’t get lost, and the tokenomics of attaching a sentiment score instead of making an agent re-run inference across a hundred articles. Plus the obvious test: can you just tell an agent to make your trades?

Everyone has the same AI now — so what’s left to compete on? UX researcher Colleen Pellissier and James Carman dig in on this week’s Forward Slash. Her read, after eight years studying what people do versus what they say: AI nails the themes but parks on the safe center line — a sea of same — and never reaches the ideas that earn real trust. The bigger trap is speed: racing from a good question to a locked answer, skipping the messy middle where insight lives. Her case — sit in the uncertainty a little longer before you bet the roadmap on the average.

What happens when your engineers stop writing code and just tell the machines what to build? That’s a software factory, and in 2026 every IT leader is being pushed to stand one up. Adam Barrett returns to The Forward Slash for a record third time and calls it what it is: vibe coding that finally got respectable. He and James Carman get into how far you can let the agents run — and the governance gap nobody’s closed: who answers for a system that writes, ships, and spends on its own? Nobody sues the model when it breaks.

Your AI budget keeps going up. The business feels about the same. Michael Schank has a theory about why: AI trained on the whole internet, and it still knows nothing about your company. He calls that the context gap, and it's where most enterprise AI quietly stalls before it ever ships. We get into the three places AI actually lives at work — running tasks, running processes, running the organization — and why most teams never make it past the first.

Your engineers are shipping code faster than ever with AI — so why hasn't your time-to-market moved? Thanos Diacakis's answer: the bottleneck just moved with it. The fractional CTO who led Uber's JUMP bikes integration joins us to discuss why faster code doesn't mean a faster business — the real constraint jumps to deciding what to build and whether customers even want it — and what the teams actually seeing returns do differently: they redesign how they work, not just which tool they buy.

Loving a technology and trusting it are two very different things — and Tamara Laine would know. She's the founder of MPWR, where her team builds policy-bound, audit-ready AI for highly regulated financial-services work. She joins James Carman to get specific: on the highest-stakes calls — like whether someone gets a loan — AI should never decide. It does the grunt work and hands a human the file. She also flips the script on regulation, making the case that firm rules aren't the enemy of innovation but the guardrails that let it move fast. Her bottom line for any leader deploying agents right now: let AI do the work — just not the deciding.

Your AI costs keep climbing, and the power bill climbs with them. The usual answer is just "buy more compute." But what if the real ceiling is the 80-year-old design inside every chip? Steve Brunker, a 30-year enterprise CIO now CEO of Brain CA Technologies, joins us to explain why a fruit fly runs on a fraction of a watt while our data centers strain the grid, and why the next leap in AI may come from rethinking the chip it runs on.

Your AI bill went up this year, so why does it feel like you're paying to ask the same question twice? Ilan Man's fix: stop using AI everywhere and start using it on purpose. The founder of Paradox Machines and former data lead at Squarespace joins James Carman with a test any leader can run. Let AI move fast where being wrong is cheap, keep a human where it's costly, and once you know the answer you want, put it in code instead of paying for it on repeat.

Who's actually reviewing the open source code your business runs on? Rich Bowen — Apache Software Foundation member since the late '90s, six terms on its board, now an open source strategist at AWS — tells host James Carman that maintainers are drowning in AI-generated pull requests that fix problems nobody has. He calls them "pizzas I didn't order." James and Rich cover what AI slop means for the dependencies buried in your stack, the software supply chain regulation Log4Shell set in motion, and how to judge whether an AI foundation's governance is real or pay-to-play.

Nearly every company is "using AI" now — but for most, it still isn't showing up in the P&L, and boards have noticed. Todd James says that gap is a leadership problem, not a technology one. The former Chief Data & Technology Officer at 84.51° and head of AI at Kroger — now advising leaders through Aurora Insights — joins James Carman to unpack why AI stalls after the first pilot, why the fix starts with the P&L, what to build versus rent, and what you should never hand to an outside model. Plus the Coast Guard lesson behind his bottom line: waiting can be as risky as moving too fast.