I just watched Kipp Bodnar and Kieran Flanagan make a claim most AI fans won’t like: the best model won’t make you a winner. As someone who has shipped products for decades, I agree. The real game is not a model race. It’s a work redesign race.
Here’s the truth many leaders dodge. Model quality is already “good enough” for most use cases. What’s missing is the hard part—changing how teams operate. That’s why firms keep buying shiny tools while results lag.
The Chart Everyone Misreads
Kipp and Kieran pointed to a viral chart showing two things: what AI could do in theory across fields like coding, finance, legal, and admin work, and what AI is actually doing on the ground. The gap is massive. That gap is not a research problem. It’s an operations problem.
“The best AI models are not going to dictate who wins and who loses in AI.”
People see the blue “theoretical” coverage and panic about jobs. They’re missing the red “observed” coverage, which shows how little has changed so far. I’ve seen this before in crypto, social, and online business. Hype peaks. Adoption stalls. Winners roll up their sleeves and rebuild the workflow.
It’s Not Tech—It’s Teams
“The bottleneck is humans.”
Kieran and Kipp are right. Most workers aren’t using AI daily. A study they cited showed 84% of normal users hadn’t tried AI. Only about 8.6% of companies have any AI agent in production. Meanwhile, top users are racing ahead. The gap between the top 5% and everyone else is huge.
That’s not because GPT-5.4 can’t code. It’s because companies still run steam-era layouts with electric motors. Kieran’s story nails it. Factories swapped steam for electricity but kept the old floor plan. Adoption stayed under 5% until they redesigned the plant for the new power source. Most firms are swapping “human steps” for AI steps without rethinking the process. That’s not transformation. That’s theater.
“You cannot walk into the future if you are looking back at the past.”
There’s also the money angle. Model costs are heavily subsidized right now. If you tried to close that entire gap at true cost, compute and inference would sting—think tens of thousands, not hundreds. Random tool trials won’t cut it. You need design, not dabbling.
The Work Redesign Playbook
Here’s where the show got practical. They outlined a simple, tough framework to rebuild teams as AI-native. I’d use this tomorrow inside any marketing, sales, support, or ops org.
The “Rapid Five” sequence gives you a clear arc. Think of it as a loop you can run every quarter.
- Reveal: Map real workflows. Mark where AI helps or hurts right now.
- Architect: Design the target operating model. Show before/after flows and pick the stack.
- Proof: Run two-week pilots on real work. Measure efficiency, capability, and step-change wins.
- Ingrain: Shift habits. Use peer learning, AI-first defaults, and performance reviews to lock it in.
- Dynamize: Reassess every 90 days as tools and costs change.
That list is simple. Implementation is not. You’ll need inputs few teams have documented. Role profiles, the 5–10 core workflows, current AI usage, goals, systems access, and risk limits. My advice: record Looms of how work actually gets done. Transcribe them. Feed that into your design process. You’ll find waste, handoffs, and repeatable tasks screaming for automation.
What Most Leaders Get Wrong
They hire for prompts, not for change. The major labs are filling “forward-deployed” roles—engineers and operators who land inside client teams and refactor workflows. That’s the signal. If you can’t hire that profile, develop it in-house. Pair a strong operator with an engineer and give them ownership of one workflow at a time.
Another miss: chasing models for bragging rights. Leaders swap model A for model B and celebrate. Results don’t move. Why? Because the email approval chain still has six people in it. The sales team still copy-pastes notes. Support still triages by gut. Tools don’t fix broken flows. Design does.
My Take—and Your Next Moves
I’ve shipped enough products to know momentum beats theory. If you run a team, stop worshiping benchmarks. Start shipping redesigned workflows that shrink cycle time and raise quality. Do it in small, measured sprints. Then lock in the gains with culture and comp.
“There’s a real gap for a lot of businesses from where they are to where they could be.”
Here’s where I’d start this week:
- Pick one workflow with high volume and clear handoffs.
- Document it with a 20-minute Loom and a checklist.
- Run a two-week pilot with a single AI-enabled path.
- Track three numbers: time saved, error rate, and output quality.
- If it works, standardize it and train the team. Then move to the next workflow.
Final thought: This isn’t an AI race—it’s an operating model race. The firms that redesign how work gets done will lap the field while others argue about scores. Choose to be on the right side of that gap. Start with one workflow. Make it real. Then keep going.
