wilcot.co Book a call

The Academy / Free

Learn to build this. Free.

How to build and run an AI-operated business, for yourself or for clients.
Three modules are live now, pulled straight from the real installs in the cases.
Free, no gatekeeping, no upsell at the end.

WILCOT / Academy Modules 01 to 03 live

What is live now

Module 01 / The map before the code
  • Two working sessions with the owner, no forms
  • The operating map, with a template you can copy
  • Why tool-first installs fail
Open module 01 Module 02 / Inside Picky
  • A named assistant in the owner’s own Telegram
  • The routines and the content loop it runs
  • Boundaries, and why the name drives adoption
Open module 02 Module 03 / The QA gates
  • The five gates we run on every piece of work
  • Banned patterns, our em dash rule included
  • The loop when a gate fails: fix, rerun, ship
Open module 03

Module 01 / Live

The map before the code

This is what we run before a single line of code exists.
Two working sessions with the owner. No forms, no questionnaire, no discovery deck.

What gets asked

01 Where money actually comes in Not the org chart answer. The real channels, and who touches each one.
02 What promises the business makes To customers, staff and suppliers. Written down nowhere, kept daily.
03 Which questions customers repeat Word for word. A repeated question is an unbuilt system.
04 Where the owner’s hours actually go The real week, by activity. It never matches what the owner guesses.
05 Which errors keep happening The ones everyone stopped mentioning. That silence is the best input.
Anti-lesson

Starting from tools is how installs fail. A tool hunting for a problem automates the wrong thing well. Start from the map and the tools pick themselves.

The output: an operating map

OPERATING MAP: [business]
Built in two owner sessions. No forms.

ENTITIES
  People: [roles, not titles]
  Systems: [every tool money or data touches]
  Outside: [customers, suppliers, platforms]

FLOWS
  Money in: [channel] > [who touches it] > [where it lands]
  Promises: [promise] > [who keeps it]
  Questions: [what customers repeat, word for word]
  Hours: [the owner’s real week, by activity]
  Errors: [what keeps going wrong, who catches it today]

DECISIONS
  [decision] | owner: [person or system] | automate: [yes or no]

STAYS HUMAN
  [explicit list, chosen by the owner, kept on purpose]

Every decision labeled with an owner, and what stays human made explicit. Fill it in before you open a single tool.

Module 02 / Live

Inside Picky, the assistant that runs a sales desk

How ProPick’s assistant is actually built.
Not a chatbot demo: a named assistant living in the owner’s own Telegram.
Plugged straight into the sales system and the CMS we built.

The routines
  • Checks sales daily, replies to leads, sends the emails
  • Spots employee input errors and corrects them as it finds them
The content flow
  • Ideate, create, post, straight to Instagram and Meta
  • One loop, no marketing meeting
The boundary
  • Employees still enter data by hand. The owner chose that
  • A working system respects chosen boundaries
Why the name matters The team treats a named assistant as a colleague. Adoption is the hard part, not the model.
Picky’s loop Picky named colleague in the owner’s Telegram Wake daily, and on events Read sales, leads, CMS entries Act reply, email, fix, post Report owner digest, Telegram

Module 03 / Live

The QA gates

AI output is a draft, and nobody outside the team sees a draft.
These are the gates we actually run, on real client work, every time.

01 Screenshot or it did not happen Every UI change is rendered and screenshot verified before it counts as done. The builder reads its own screenshot and fixes what looks wrong.
02 Fixed-viewport overflow check Layouts are checked at fixed desktop and phone viewports. Overflowing or clipped content fails the gate, whatever the code claims.
03 Banned-pattern scan Output is scanned for banned patterns before it ships. Our house example: em dashes are forbidden, a style gate that runs on every file.
04 Second-model scoring Work is scored against an approved reference by a second model before shipping. The builder never grades its own work.
05 Cheap hands, strong judge Cheap models do the mechanical work, a strong model judges it. Spend on the judging, not on the typing.
When a gate fails

Fix, rerun the gate, only then ship. Never argue with the gate, never ship around it.

The lesson

The gate list is the product. A model without gates is a demo.

More modules are coming

New modules land
in your inbox first.

Modules 04 and up are being written from installs that are live right now.
Leave your email and each new module arrives the day it ships.
Nothing else, no drip sequence, no upsell.

Your email

Free, like the first three. We won’t spam you, ever. Your address is stored securely. Privacy, plainly.

Want it installed instead?

See the offer Book a call