M4 · PROTOTYPE TO PRODUCTION · P10

The Prototyping Loop Playbook

Thesis to decision in 5 days. Test the thesis before you scope the project.

10 SLIDES · ~10 MIN · PLAYBOOK OVERVIEW
Key Takeaways
  • Test the thesis. Then scope the project.
  • 53% — of enterprise AI pilots stall before reaching production.
  • 6–9 mo — typical pilot-to-production timeline at mid-market firms.
  • 5 days — from thesis to kill-or-scale decision in the prototyping loop.
  • Name one repetitive internal task that has clear inputs and outputs — write the one-sentence thesis.
MODULE 4 · PLAYBOOK 10

The Prototyping Loop Playbook

Thesis to decision in 5 days. Test the thesis before you scope the project.

BuildClub Academy
01 · THE HOOK

Most AI pilots stall. A five-day loop tells you why before you spend.

53%
of enterprise AI pilots stall before reaching production.
McKinsey AI deployment research
6–9 mo
typical pilot-to-production timeline at mid-market firms.
BuildClub field interviews · 2026
5 days
from thesis to kill-or-scale decision in the prototyping loop.
THE THESIS

Test the thesis.
Then scope the project.
Not the other way around.

The Thesis
02 · THE PROTOTYPING LOOP

Five steps from thesis to kill-or-scale — one working week.

1
2
3
4
5
1
State the thesis
In one sentence. If you cannot, the project is not ready to be scoped, let alone built.
2
Build in 5 days
A working prototype against a real data sample. Not a deck. Not a flowchart. Running code.
3
Wire to one flow
Connect it to one workflow that exists today. No greenfield.
4
Measure against one metric
One named number moves or it does not. If the metric does not move, kill it.
5
Kill or scale
The Friday decision. Prototype cost under $5K and five days either way.
03 · WHY PILOTS STALL

Four failure modes that turn a six-week pilot into a six-month sinkhole.

1
Scope before thesis
Committee writes the scope before anyone tests whether the idea works. The pilot inherits assumptions nobody owns.
2
Synthetic data
Built and demoed against clean, curated, or fake data. Real-world drift kills it the week it reaches production.
3
IT owns it
Routed to IT before a business owner is named. Becomes a technical project with no outcome metric.
4
No kill criterion
Nobody defined what failure looks like. The pilot never gets killed. It just runs forever.
04 · THE COST OF BEING WRONG

Thesis-first prototyping drops the cost of being wrong by 100×.

Takeaways
  • 1 Scope-first pilot — $500K / 6 months / 53% stall rate. Scope written before the thesis is tested.
  • 2 Thesis-first prototype — under $5K / 5 days / kill-or-scale by Friday. Real data, one workflow, one named metric.
  • 3 Cost of being wrong drops 100×. Speed of being right increases 36×.
05 · THE OPERATING PATTERN

One week, four checkpoints.

Mon
Tue
Wed
Thu
Fri
Mon
State the thesis
One sentence. Business owner plus one engineer only.
Tue
Build against real data
Perplexity prototype. Real data sample. No synthetic.
Wed
Wire to one workflow
The actual live workflow. Real system access.
Thu
Measure against the one metric
Run the comparison. Collect the output.
Fri
Kill or scale
Present to the sponsor. Make the call. If kill, document why.

No slides Monday–Thursday. The Friday meeting is the first presentation.

06 · THE BUILD HIERARCHY

Pick the tool by what it has to do — not by what it is.

1
1 — Perplexity
First 5 days. Idea to working artifact in one window. No infrastructure required.
2
2 — Replit
Needs auth, schedule, or multi-user. Cloud-resident prototype.
3
3 — Local PC
Faster, free, private. Runs on the machine in front of you.
4
4 — Extension / native
When it needs to live on your screen. Browser extension or desktop app.

Try Perplexity first. Move up the hierarchy only when the prototype proves the thesis.

07 · THE AI-ON-AI PATTERN

Two windows. One architects. One builds. Copy-paste when they disagree.

1
Architect window
Receives the goal. Produces the spec. Thinks about structure, edge cases, data model. Stays out of the code.
2
Builder window
Receives the spec. Produces the code. Heads-down implementer. Does not question the goal. Turns the spec into running code.
3
When they disagree
Copy-paste back. The architect wins. The second system is the voice of reason.
MONDAY-MORNING ACTIONS

Five moves to ship a prototype this week.

Monday Morning
  1. 1
    Name one repetitive internal task that has clear inputs and outputs — write the one-sentence thesis.
  2. 2
    Open Perplexity, state the thesis, build against real data — start today.
  3. 3
    Wire it to one workflow that exists today — no greenfield, no synthetic data.
  4. 4
    Name the one metric it must move by Friday — write it on the calendar invite.
  5. 5
    Make the kill-or-scale decision on Friday — no extensions.
P10 · Questions CEOs Ask

Frequently Asked Questions

What is the core idea of The Prototyping Loop Playbook?
Test the thesis. Then scope the project.
What does the data say a CEO should pay attention to?
53% of enterprise AI pilots stall before reaching production. 6–9 mo typical pilot-to-production timeline at mid-market firms.
What should a CEO do Monday morning after reading The Prototyping Loop Playbook?
Start here: Name one repetitive internal task that has clear inputs and outputs — write the one-sentence thesis; Open Perplexity, state the thesis, build against real data — start today; Wire it to one workflow that exists today — no greenfield, no synthetic data.
What are the steps in The Prototyping Loop Playbook?
1) State the thesis; 2) Build in 5 days; 3) Wire to one flow; 4) Measure against one metric; 5) Kill or scale.
Where do the claims in The Prototyping Loop Playbook come from?
The playbook cites McKinsey AI deployment research; MIT NANDA — 95% pilot-stall figure; YPO Technology Network AI Daily Brief "Stop Prompting. Start Briefing." (Mar 7, 2026); YPO Technology Network AI Daily Brief #20 (Rokt 135 apps in 24 hours).
P10 · Sources

All Sources in This Playbook

00 / 10