Process · The Bill-or-Bust Loop

We never start
with code.

We start with the one person your product is for. Then we measure, build, launch, and let your own data decide whether that part of the fee gets invoiced.

We build the person
before the product.

Before the first line of code, we study the one person the product is for. Not a persona deck. One real, specific person.

What we study

  1. P-01Where she eats.
  2. P-02What her Tuesday looks like.
  3. P-03Which tabs are open when she finds you.
  4. P-04What she was doing thirty seconds before she opened the app.
  5. P-05What she just gave up on.

What gets built for her

Copy
Written in the words she already uses.
User flow
Shaped around the thirty seconds before she opens the app.
Distribution channel
Chosen from the tabs she already has open.

Nothing is a template. By the first line of code, we already know who the first ten users are and how they'll arrive.

02 · The Bill-or-Bust Loop

Four stages. The decision is written first.

Every build runs the same loop. What counts as success is in the contract before we write any code.
  1. 01Week 1

    Baseline

    We measure before we build.

    • We measure the number we're going to move, as it stands today.
    • We study the one person the product is for.
    • The exit criteria go into the contract in week one.
  2. 02First code by day 4

    Build

    One dedicated team, scope locked.

    • One team that stays on your build from start to finish.
    • The first line of production code goes in by day 4.
    • Scope is locked, so the date holds.
  3. 03The promised date

    Blastoff

    Live on the date we promised.

    • Live in front of real users, not a demo.
    • Monitoring is on from hour 1.
  4. 04The check date

    Bill-or-Bust

    The scorecard opens on your data.

    • The decision was written down before we started.
    • Bill: the loop restarts, and the last result becomes the new baseline.
    • Bust: we fix it for free, or you walk away with the system.

If it's Bill

The loop restarts. Your last result becomes the new baseline, and we go after the next number.

If it's Bust

We fix it for free, or you walk away with the system. That part of the fee is never invoiced.

Five numbers.
Your data.

AI work is graded on a five-metric scorecard, every quarter, on a dashboard you can open any day.

Swipe to explore the table →

RefMetricTargetHow it's measured
M-01Automation rateRisesThe share of the work the system completes without a person stepping in.
M-02AccuracyNever fallsGraded monthly on 100 outputs against a rubric we write together.
M-03Time per taskFallsMeasured from your tools, not self-reported.
M-04AdoptionRisesRead from logs.
M-05Cost per taskFalls every quarterTracked from the same dashboard.
ΣThe triggerThree of five hit target, and accuracy never drops.

The accuracy rubric is written with you. Time per task comes from your tools and adoption from your logs, never from self-reporting.

04 · Ready when you are

Let’s build what’s
needed.

One call. Bring the number you want to move. If the outcome looks likely, we put part of our fee behind it. If it doesn’t, we’ll quote normally or tell you honestly.

An engineer replies within a day · no pitch deck required

Not ready for a call? Ask our GPT →
DF-A · First conversation
Booked this week
DF-B · Baseline & exit
Written in week one
DF-C · First commit
By day four