The 2-Week Build Log: Zero to Paying Customers

The timeline (free)

This is the compressed version of what I actually did: empty repo on a Monday, first paying customers by the end of the following week. Not a growth-hack fairy tale — a constrained build with one outcome, one AI surface, and an outbound motion that started before the product felt "ready."

Days 1–2 — Scope and walking skeleton

Day 1 decision: One outcome only. "Help operators install AI workflows" was too wide. I cut to: a buyer can describe their GTM motion, get a scored readiness readout, and book a conversation. Everything else waited.

Day 1 dead end: I almost built a full "AI OS" dashboard on day one. That would have burned the week on UI. I deleted the sketch and wrote the data model for one scorecard flow instead.

Day 2: Wired the thinnest possible path — form → score → result page → calendar link. No auth. No admin. No analytics beyond a single event log. Shipped a walking skeleton to a private URL so I could feel the loop end to end.

Days 3–6 — Core AI against real inputs

Day 3: Built the scoring prompt against answers I already knew were good and bad from client work. If the model couldn't separate a serious team from a "we're exploring AI" team, the product was fake.

Day 4: Fed it messy real answers — incomplete forms, vague goals, contradictory tools lists. Fixed the prompt until weak inputs produced "unclear / not ready" instead of polite fluff.

Day 5: Instrumented one metric: completed scorecards that reached the booking CTA. Not pageviews. Not "engagement."

Day 6: Hardened failure modes — timeout copy, empty-state copy, and a manual override path if the model returned garbage. Still no marketing site polish.

Days 7–9 — Surface, edges, production

Day 7: Wrote the result page like a field note, not a SaaS dashboard. One headline, the score breakdown, three recommended next moves, one CTA.

Day 8: Deployed to production with a custom domain. Fixed the embarrassing mobile bugs. Added a plain privacy note because people paste real process details.

Day 9: Soft-launched to a handful of people who'd already talked to me about AI enablement. Watched them complete the flow. Changed two questions that caused drop-off. Did not add features they "might want someday."

Days 10–14 — Outbound while the product is ugly-honest

Day 10: Started outbound to operators and revenue leaders who already had the problem — not "AI curious" lists. Message was specific: readiness score in two minutes, no email gate on the score itself, conversation if the gap was clear.

Days 11–12: Replied to every response the same day. Qualified hard. Booked calls only when the score and the conversation both showed a real install need.

Days 13–14: Closed the first three paying customers from that sequence — people who had already felt the gap in the scorecard or a prior conversation. The product didn't close them alone. The outbound + the clear offer did. The scorecard made the problem visible fast enough that the sales call wasn't a cold education session.

What the free timeline leaves out

The prompts behind the score, the exact outbound sequence, how I scored inbound replies, and the daily standup notes that kept a one-person build from drifting into "just one more feature." That's the playbook below.

The full playbook

The constraint I wrote on day zero

Before I opened the repo I wrote three lines and taped them above the desk:

One outcome: score readiness → recommend next move → book.
Ship something a stranger can complete without a Zoom explanation.
Start outbound on day 10 whether I feel ready or not.

If a task didn't serve those three, it waited. Most "nice" ideas failed that test.

Day-by-day decisions (the ones that mattered)

Day 1 — Cut the ICP. I was tempted to serve founders, ops, and sales leaders equally. I picked revenue / GTM operators inside B2B teams who already felt adoption pain. Broader ICP would have made the questions mushy.

Day 2 — No accounts. Login would have doubled the build. I accepted that people might redo the scorecard. Speed beat retention theater.

Day 3 — Rubric before model cleverness. I wrote a manual scoring rubric on paper first: five dimensions, what "1" vs "5" looked like in plain English. Only then did I ask the model to apply it. Without the rubric, the model flattered everyone.

Day 4 — Rejected "helpful" tone. Early outputs sounded like a consultant deck. I added instructions: name the gap bluntly, no "exciting opportunity" language, recommend one next move not five.

Day 5 — Metric discipline. I almost added a heatmap. Instead I logged scorecard_completed and booking_clicked. Two events. That was enough to see whether the AI surface created motion.

Day 6 — Human escape hatch. If scoring failed, show a short manual path: "Tell me your biggest GTM AI bottleneck in one sentence" and a book link. Never leave someone on an error page during a launch week.

Day 7 — Copy as product. I spent more time on the result page words than on visual design. The product is the diagnosis. Pretty charts would have lied about readiness.

Day 8 — Production before pride. Shipped with known rough edges. Fixed only what blocked completion.

Day 9 — Change questions, not architecture. Drop-off taught me two questions were academic. I rewrote them to force a concrete answer ("What happens in week two after an AI rollout?") instead of adding a progress bar.

Day 10 — Outbound with a scar. Message led with the problem and the artifact, not "I built a thing." People who didn't have the problem ignored it. That was correct.

Day 11–12 — Qualify in the reply thread. I did not book everyone who said "interesting." I asked one qualifying question before calendar links.

Day 13–14 — Sell the install, not the scorecard. Paying customers bought help installing workflows / a sprint / ongoing partnership. The scorecard was proof I understood the failure mode — not the whole offer.

Outbound sequence (exact shape)

I used a short sequence. No seven-touch novel. Channel was email + selective LinkedIn notes to people I'd already crossed paths with or who publicly talked about AI enablement pain.

Touch 1 — Problem + artifact

Subject: week-two AI adoption

[Name] — noticing a lot of GTM teams "roll out AI" and quietly lose it by
week two.

I put together a two-minute readiness scorecard that scores the boring
stuff that actually predicts whether workflows stick (install depth,
manager loop, boundaries, etc.). No email gate on the score.

If useful: [link]

Either way, curious whether week-two decay is showing up on your team.

Touch 2 — Three days later, only if no reply

Subject: re: week-two AI adoption

Quick bump with context: the scorecard is deliberately blunt. If it says
you're not ready, it's usually because the rollout was announced as a tool
instead of installed as a workflow.

Happy to send the three questions I use in the first enablement session if
easier than running the form.

Touch 3 — LinkedIn note (selective)

Saw your note on [specific AI / enablement topic]. Built a short readiness
score for GTM teams fighting shelfware — figured it might be in your world.
No pitch in the link: [link]

Stop rules: Two emails + optional note. If no engagement, remove. I did not "add value" with a fourth touch that was actually a pitch in costume.

Qualification scoring

I scored replies mentally on a 0–2 scale across four questions before I sent a calendar link:

Do they own or influence GTM / sales enablement? (0–2)
Is there a live pain (failed rollout, unused licenses, manager push)? (0–2)
Can they fund a sprint or kit purchase in the next 30–60 days? (0–2)
Are they looking for install help — or a magic tool? (0–2; magic tool = low)

8–8: Book same week. 5–7: Send a sharper question or the kit page; book only if answers tighten. Below 5: Be kind, send the free scorecard / a relevant Signal post, don't fake a fit.

Qualifying question I used most:

When AI was rolled out on your team, what broke first — the workflow design,
the manager coaching loop, or trust after a bad output?

People with a real answer were buyers. People who said "we just need better prompts" usually weren't ready for paid install work yet — I pointed them to free Signal posts and stayed in touch.

Prompt skeleton behind the scorecard

Not the full production prompt — the shape that mattered:

You are scoring GTM AI readiness for a B2B team. Be blunt. No hype.

Dimensions (1–5 each): problem clarity, workflow install depth, manager
inspection loop, written boundaries, measurement honesty.

For each dimension: score, one-sentence evidence from their answers, one
concrete gap.

Then:
## Overall
One paragraph. Name the primary failure mode.

## Do this next
Exactly one recommended move (not a menu).

## Don't do yet
One thing they should stop or postpone.

Rules: if answers are vague, lower the score and say why. Never invent
maturity they didn't demonstrate.

I iterated this daily on days 3–6 with real answer sets until it stopped flattering thin inputs.

Daily standup notes (solo)

I ran a ten-minute standup with myself every morning. Same four lines in a doc:

Date:
Yesterday shipped:
Today's one outcome:
Blocked by:
Not doing today (parked ideas):

Sample entries, lightly cleaned:

Day 3

Yesterday shipped: rubric v1 + first scoring pass
Today's one outcome: fail loud on vague inputs
Blocked by: none
Not doing: dashboard, PDF export, team accounts

Day 7

Yesterday shipped: result page copy v2
Today's one outcome: production deploy + domain
Blocked by: DNS (solved midday)
Not doing: blog, brand illustrations, secondary CTAs

Day 10

Yesterday shipped: soft launch tweaks
Today's one outcome: 15 outbound touches
Blocked by: ego ("product isn't ready") — ignored per day-zero rule
Not doing: new features requested in soft launch

Day 14

Yesterday shipped: second and third close conversations
Today's one outcome: write onboarding notes for first customers
Blocked by: none
Not doing: rebuild scorecard UX

The "Not doing today" line saved the build. Without it I would have spent days 8–12 on cosmetics.

First-customer onboarding notes (what I sent)

After the first close I stopped improvising the handoff. This is the note I sent — short on purpose:

Welcome — here's how the first two weeks work.

Week 1
- 45-min install session on one workflow (usually call prep)
- Real accounts only; we fix instructions live
- Boundary doc signed off by you / your manager champion

Week 2
- Manager coaching loop starts ("show me the artifact")
- One named win posted in your team channel
- Day-14 check: keep, fix, or kill the workflow

You own: picking the workflow, naming the manager champion, bringing live deals.
I own: project instructions, session facilitation, prompt fixes, day-14 audit.

Async: one shared doc. No second tool unless security requires it.

That note prevented the classic post-sale drift where "paying customer" turns into endless Slack advice with no install date.

Failure modes inside the two weeks

Feature gravity. Soft-launch users asked for exports, team seats, and history. Valid ideas. Wrong week. I logged them under "Not doing today" and revisited after three paying customers, not before.

Polite scoring. When I softened the prompt to sound friendlier, completion went up and booking quality went down. Blunt scores filtered better. I put the bluntness back.

Outbound to the wrong list. My first ten touches included "AI curious" operators with no GTM ownership. Replies were nice. Pipeline was empty. I rewrote the list around revenue / enablement owners with a visible pain and resumed.

Selling the scorecard. One early conversation tried to buy "just the tool." I declined. The offer was install help; the scorecard was the wedge. Keeping that line saved me from supporting a product I wasn't ready to run as standalone SaaS.

What I'd repeat vs. change

Repeat: one outcome, rubric before prompts, outbound on a fixed day, qualify before calendar, sell the install not the artifact, morning standup with a parked-ideas line.

Change: I would have recorded five real scorecard sessions on day 9 and watched the recordings before outbound. A couple of question phrasings still confused people. Also: I under-invested in a simple "what happens after you book" page — first customers asked operational questions I should have answered in writing. Last: I would have drafted the outbound touches on day 7 so day 10 was send-only, not write-and-send under fatigue.

Soft CTA

If you're staring at an empty repo (or a bloated roadmap) and need the install-and-ship version of this thinking applied to your GTM motion, start with the free readiness scorecard or look at the GTM AI Operating System. If you want me in the room for two weeks of install work, that's the GTM AI Sprint.