Path 1 · Hair on Fire
"Help me now."
You are entering a fight. Aggression is the strategy, and speed is the moat while the moat gets built.
Product-Market Fit · AI Scale-ups · 2026
An AI product can book a year of pilot revenue on curiosity budgets alone and still have no business underneath it. This page is how I tell motion from fit at early-stage AI, AI tooling and AI security companies: three paths, one diagnostic, and the signals that survive contact with a renewal.
01 · The Three Paths
Sequoia's Arc framework sorts product-market fit into three paths, defined by how the customer already feels about the problem. It is the most useful PMF model I have used with AI companies, because AI founders routinely misdiagnose their path and then execute the wrong playbook at full spend.
Adapted from Sequoia Capital's Arc product-market fit frameworkPath 1 · Hair on Fire
"Help me now."
You are entering a fight. Aggression is the strategy, and speed is the moat while the moat gets built.
Path 2 · Hard Fact
"It is what it is."
Two jobs, both mandatory: a problem that matters enough to change for, a solution compelling enough to believe.
Path 3 · Future Vision
"Yeah, right."
Expect pit stops. The viable path rarely runs through the customer you first imagined.
Path determines playbook. Hair-on-Fire aggression in a Hard-Fact market torches cash on buyers who feel no urgency. Hard-Fact patience in a Hair-on-Fire market hands the land grab to someone faster. Diagnose before you spend.
02 · The Diagnostic
Eighteen statements, adapted from Arc's diagnostic. Check what is true today, not what the roadmap promises. The tallest column is your path. A scattered result is a finding too.
Check what is true today. The tallest column is your path, and the playbook changes with it.
03 · Early Signal
Every AI company I work with can show traction. The job is sorting which of it predicts a renewal. Two lists, built from deals I have watched close, and watched unwind.
Signals that hold
Weight these. They compound.
Paid at real prices
Design partners who negotiated. Discounts trade for references and roadmap input, never for silence. Free pilots teach nothing about willingness to pay.
Usage before signature
The team built on you mid-evaluation without being asked. Procurement becomes paperwork instead of persuasion.
POV wins in the 75-85% band
With success criteria signed before kickoff. A low win rate against signed criteria is a message problem, not a sales problem.
Wins outside the network
Net-new logos with no tie to the founder: bought, deployed, renewed. Everything before that is anecdote.
The same story, eight times
Same buyer, same pain, same words in the win notes. Repetition is what fit sounds like.
Expansion nobody engineered
Workloads, environments and teams growing without an account manager pushing. NRR of 105-110% in the landing year means the product is doing the selling.
Inbound that names an assistant
A buyer who says ChatGPT or Claude recommended you arrived pre-sold by the answer box, which is where buying now starts.
Signals that lie
Discount these. They flatter.
Innovation-budget ARR
The experimentation fund bought a curiosity. That line item sunsets, and the logo goes with it.
Pilot purgatory
POVs with no signed criteria and no exit date. An evaluation that cannot end is a hobby you are funding.
The novelty curve
Week-one usage spikes, week-six silence. Judge cohort curves, not launch-day dashboards.
Free design partners
They will tell you the product is interesting. They cannot tell you what it is worth, because to them it is worth zero.
Security-team meetings
Taking briefings is the job; it costs them nothing. The signal is a named owner and a budget line, not a calendar of polite thirty-minutes.
Stars and logo walls
GitHub stars, waitlists and borrowed logos measure curiosity. Retention measures value.
Services dressed as software
If every deployment needs your engineers indefinitely, you built a consultancy with a login page.
Two questions sort every signal above. Did they pay real prices? Did they change how they work to keep using it? Money and changed behavior. The rest is commentary.
04 · The Operating Play
PMF work fails as a vibe and works as a weekly discipline. Five moves, in order, running on the same stage gates as the rest of this playbook.
Graduate when two net-new logos outside the network have bought, deployed and renewed. Then prove transfer: two non-founder reps at 70%+ attainment, two consecutive quarters. Spend scales after evidence. Never before.
Andrew Wesbecher
If you are running an AI or AI security company between $0 and $20M, and the signals on this page look familiar, or worryingly do not, my inbox is open.
This page is one module of a larger argument: the $0 → $20M Sales & GTM playbook for AI scale-ups.