The AI Forward Method
Find the bottleneck.
Then apply AI.
A company moves at the speed of its slowest step. AI anywhere else is a cost, not a lever. The method is four steps, repeated every time the bottleneck moves.
The AI Forward Method has four steps. Find: map where the business leaks time and money and name the bottleneck. Fix: fix the process, then build and run AI at that step. Own: keep a named person accountable and a decision ledger the company owns. Teach: pass the method to the team, so the next bottleneck is found faster.
01 / The four steps
Find. Fix.
Own. Teach.
- Find
Map the leaks, name the bottleneck
We trace the revenue process step by step, from lead to contact, qualify, close, deliver and collect, and measure each step. The slowest one is the bottleneck. Everything else waits. This is the AI Forward Diagnostic.
- Fix
Process first, then AI at that step
A broken process run by AI just breaks faster. We fix the process, then build the AI system at the bottleneck and run it ourselves, with tests built from real cases before it goes live.
- Own
A named owner and a ledger you keep
Every system logs its decisions: the inputs, the output, any human override and the outcome. A named person owns the overrides. The ledger, accounts, data and code sit in your systems, with a written exit plan.
- Teach
Make the team faster at the next one
We teach the team to read the ledger and spot the next bottleneck, through working sessions built on your own workflows.
02 / Why owners can't do this alone
Three assumptions
we take away.
Owners don't skip AI because they doubt it. They skip it because rebuilding competes with running today's business for the same time and cash. The method is built to remove that conflict.
- 01
“It will take my time.”
We diagnose, build and run the system. You give decisions, not hours.
- 02
“It has to change everything at once.”
We start at one bottleneck only, and move to the next once the first one breaks. The diagnostic shows the whole sequence, so the first fix is the start of a roadmap, not a pilot.
- 03
“We'll be locked in.”
You own the accounts, the data, the code and the decision ledger. If we leave, the system and its history stay with you.
03 / The bottleneck moves
Re-find it
every cycle.
Classic Theory of Constraints assumed a few stable, physical bottlenecks. AI changes that. Once it clears one step, the bottleneck moves somewhere AI can't reach yet.
- 01
Where it moves
To checking the work, to judgment and decision rights, to trust and compliance, and to the data the system needs. That is why every system ships with tests and a ledger: they make the next move visible.
- 02
Loops everywhere, judgment at the bottleneck
Cheap AI loops can improve every part of the company. Your people's judgment is scarce, so it goes only where the bottleneck is. The loops elsewhere are what make you fast when the bottleneck moves.
- 03
One measure
We measure what the business produces at the bottleneck: more sales closed, more jobs delivered, more cash collected. Not hours saved, and not tools adopted.
04 / Teach
Talks and
team sessions.
Shake gives talks and runs working sessions on bottleneck-first AI for leadership and operations teams, using the team's own workflows as the material. A recent example: a live session on working with Claude for Canadian travel advisors (September 2026).
- 01
Leadership session
Find your own bottleneck in a working session, and leave with the first fix scoped.
- 02
Team training
How to work with AI at the bottleneck: prompts, tests, overrides and the ledger.
- 03
Talks
The AI Forward thesis, with what we have learned running AI on a live sales floor.
Questions
Straight
answers.
Why not start by giving everyone ChatGPT?
Do-it-yourself AI tools mostly get pointed at the easy, visible work (posts, email replies), which is rarely the bottleneck. They help people, but the business doesn't produce more. Start at the bottleneck, then let tools spread.
What if AI isn't the answer to our bottleneck?
Then we say so. Sometimes the fix is a process change, a hire or a pricing decision. We never sell AI that isn't on your bottleneck without saying so plainly.
What is the Theory of Constraints?
Eliyahu Goldratt's idea that in a chain of dependent steps, output equals the capacity of the slowest step. Improving any other step leaves output unchanged until it becomes the slowest. The AI Forward Method applies it to AI, and extends it: with AI, the slowest step keeps moving.
Why does every system need tests and a ledger?
An AI system is only as good as the feedback it gets. Tests built from real cases catch failures before customers do, and the ledger turns every decision into material for the next model.
Insights
In the
news.
Start here
Find the
bottleneck.
The diagnostic is the Find step, done properly. The scorecard is the five-minute version.