"The competitive edge isn't AI tools. It's how fast your team learns to use them."
The Bottleneck Has Moved
"This isn't a technology observation. It's a structural opportunity. The teams that figure this out first will operate at a different level."
⬅ Before
Expertise bottleneck: "Can the SA answer this?"
Output throttled by human availability
Time spent: writing, translating, chasing context
Scale = headcount
After ➡
Judgment bottleneck: "Is this answer worth acting on?"
Output limited only by quality of thinking
Time spent: framing problems, challenging assumptions
Scale = system design
This Has Been Done Before
Fiona Fung's approach at Anthropic is the closest reference model we have.
01
Start with the team, not the tools
Capability uplift before workflow deployment. Trust before automation.
02
Build the system, not the hero
Document everything. The knowledge should live in the org, not in one person.
03
Prove it small, then scale
Two pilots. Real work. Honest measurement. Then expand.
"The principle isn't new. The application to our context is."
90 Days to Signal
Phase 1
Days 1–14 Foundation
Map current workflow bottlenecks
Identify 2 pilot volunteers
Define what "better" looks like (metrics)
Phase 2
Days 15–45 Build
Stand up lightweight AI-assisted workflows for pilots
Weekly check-ins, no forced adoption
Document what works and what doesn't
Phase 3
Days 46–75 Learn
Debrief with pilots
Identify what's repeatable
Build first internal playbook draft
Phase 4
Days 76–90 Report
Present signal back to leadership
Make go/no-go recommendation
No sunk cost — honest read
I Want Your Feedback, Not Your Approval
"I want to run two pilots in Q3. I want honest feedback from you today — where you see the gaps, what I'm missing, what would make you more confident. I'm not asking you to bet on this. I'm asking you to let me prove it."
🟢 What I'm asking for
Permission to run 2 pilots, access to early adopters
🟡 What I'm NOT asking for
Budget, headcount, org changes
⚪ What you get
A 90-day honest read, decision-ready data
"If it doesn't work, we'll know in 90 days and lose nothing. If it does, we'll have built something durable."