The -1 to 1 playbook takes an operator from raw industry insight to AI product-market fit in defined phases: validate the problem and buyer before building, ship a narrow first product to real users, then read the signals that show the wedge is working. It front-loads the validation most founders skip and reaches PMF faster.
Most startup advice starts at 0 to 1: you have a company, now build the product and find the market. It skips the part where most companies actually die, the -1 to 0 stretch, where you decide whether the idea deserves to be a company at all. For a domain expert building a vertical AI product, that earlier stretch is where the game is won. This is the playbook for running it deliberately.
What "-1 to 1" means
The 0 to 1 framing, popularized by Peter Thiel, describes creating something new where nothing existed. The -1 to 0 stretch is the part before that: taking a raw insight and deciding, with evidence, whether it should become a company at all. A venture builder's core work lives in -1 to 0, which is why the model exists. It industrializes the riskiest, least glamorous part of company creation.
For a vertical AI company the sequence matters more than usual, because the failure modes are front-loaded. Build the wrong workflow for the wrong buyer and no amount of model quality saves you. The evidence backs the order: vertical AI companies are outpacing horizontal tools precisely because they solve one industry's problem exactly, and getting that "exactly" right is validation work, not engineering work. The full operating rhythm of a builder running this is in what a venture builder actually does week by week.
Phase 1: -1 to 0, the validation sprint
Before a line of production code, you run a short, focused validation sprint, often around four weeks. It answers three questions, in order, and any one of them failing sends you back to the drawing board.
Is the problem real and expensive? Talk to the people who live the problem. Not a survey, real conversations with operators in the industry. You are looking for a problem that costs money or time today, that people already try to solve with duct tape, and that they can describe without prompting. Your own domain expertise is the unfair advantage here; you know who to call and what to ask.
Is the buyer reachable and willing? A real problem with an unreachable buyer is not a business. Identify who holds the budget, how long their buying cycle runs, and whether you can get to them. This is where founder-market fit pays off, because an operator with distribution before a product has a warm path to the first customers.
Is there a narrow wedge? You cannot boil the ocean. Find the single workflow, painful, frequent, and measurable, where an AI product can win first and expand from. The wedge is what makes the build tractable and the first sale fast.
At the end of the sprint you have a decision, not a deck. Either the evidence supports building, or it does not, and you have spent four weeks instead of a year finding out.
Phase 2: 0 to 1, build to first value
Once validated, you build the narrowest product that delivers real value on the wedge, and you put it in front of real users fast. The goal is not a feature-complete platform; it is one workflow done well enough that a real user would miss it if it disappeared.
Vertical AI has a specific advantage in this phase. Because the product is scoped to one industry's data and rules, it reaches measurable value quickly. Analyses of vertical AI funding and outcomes point to why investors favor these companies: workflows that are rule-heavy, document-heavy, and recurring produce value that shows up in months, not years. You are building against a target that gives clear feedback.
Ship, watch how the wedge performs with real users, and iterate on the workflow before you widen the product. Widening too early is the most common way a promising 0 to 1 stalls.
Phase 3: reading product-market fit
Product-market fit is not a launch or a press hit. It is a set of signals, and for a vertical AI product they are concrete.
| Signal | What to look for | Why it matters |
|---|---|---|
| Retention | Users keep using the workflow week over week | Proves the value is real, not novelty |
| Pull | Prospects ask for access before you sell them | The market is doing your selling |
| Willingness to pay | Buyers commit budget, not just interest | Separates a product from a project |
| Expansion | Users want the next workflow next to the first | The wedge is widening on its own |
When those signals appear together on the wedge, you have found the beginning of product-market fit, and only then do you widen the product and raise to scale. Chasing growth before the signals are real just spends money faster.
The three ways the playbook goes wrong
Knowing the phases is not the same as running them well. Three failure modes account for most of the misses, and each maps to a phase.
The first is building before validating. This is the default, and it is seductive because building feels like progress. An operator with a strong insight jumps straight to a product, spends months and money, and only then discovers the buyer will not pay or the wedge was wrong. The fix is discipline: no production build until the validation sprint has cleared all three questions.
The second is the opposite, validating forever. Some founders mistake endless customer interviews for progress and never commit to building. Validation has a stop condition: once you have enough evidence to make a build-or-do-not-build call, you make it. A sprint with a deadline forces the decision instead of letting research become a place to hide.
The third is widening the wedge too early. A product that is working on one narrow workflow feels like it should expand immediately, and founders add features to chase every prospect request. That dilutes the one thing that was working before it has proven durable. Hold the wedge until the product-market-fit signals are unmistakable, then expand from a position of strength rather than hope. Investors notice the difference, which is part of why vertical AI companies command the funding they do: a proven wedge with expansion pull is a far safer bet than a broad product with shallow usage.
Each of these is a discipline problem more than a knowledge problem, which is exactly where a partner who has run the playbook before earns their place.
Where a venture builder fits the playbook
Running this playbook alone is hard for a domain expert, because -1 to 0 needs engineering to validate quickly and 0 to 1 needs a team to build. A technical venture builder runs the whole playbook with you. gAI Ventures co-founds vertical AI companies in financial services, enterprise productivity, and commerce, and its model maps directly onto these phases: a validation sprint in -1 to 0, then a founding engineering team building the wedge product in 0 to 1, from day one.
The structure supports the playbook rather than getting in its way. gAI contributes roughly $50K at incorporation and about $200K on milestones, with a combined fund-and-operating-company stake around 20 percent, which keeps the cap table clean enough to raise a seed once the PMF signals arrive. The specific industry wedges it co-founds around are in its vertical AI investment theses, and the companies that ran this path, including FastTrackr AI, Swik AI, ContentsIQ, and Turtle AI, are on the gAI Ventures portfolio. Why gAI co-founds rather than invests passively is set out in the gAI Ventures manifesto, the people running it are on the gAI Ventures team, and more of the thinking is on the gAI Ventures blog.
The takeaway
The path from industry insight to AI product-market fit is not build, then hope. It is validate, then build narrow, then read the signals honestly. Run -1 to 0 with discipline and 0 to 1 stays tractable. Skip it, and you are one of the many companies that built something nobody needed. The operators who win in vertical AI are the ones who front-load the validation their competitors skip, hold a narrow wedge until the signals are real, and only then widen and raise. That order is the whole playbook, and it is learnable, whether you run it alone or with a partner who has run it many times before.
Frequently asked questions
- How do you get from idea to product-market fit for an AI product?
- Run it in phases. First validate that the problem is real and expensive, the buyer is reachable and willing to pay, and there is a narrow wedge to enter on. Then build the narrowest product that delivers value on that wedge and put it in front of real users. Then read for retention, pull, and willingness to pay. Only widen and scale once those signals appear together.
- What is the -1 to 0 stage of a startup?
- It is the stretch before the company exists, where you take a raw insight and decide, with evidence, whether it should become a company at all. It covers problem validation, buyer validation, and finding the wedge. Most startup advice skips it and starts at 0 to 1, which is why so many companies build products for problems that were never real.
- How long should validation take before building an AI company?
- A focused validation sprint often runs around four weeks. The point is to reach a clear build-or-do-not-build decision quickly, using real conversations with people who live the problem, rather than spending months building on an unvalidated assumption. A short, disciplined sprint replaces a year of building the wrong thing.
- Why is vertical AI easier to reach product-market fit with?
- Because the product is scoped to one industry's workflow, data, and rules, it produces measurable value quickly and gives clear feedback. Rule-heavy, document-heavy, recurring workflows show returns in months, which makes the 0 to 1 phase more tractable than building a general-purpose tool that has to be everything to everyone before it is useful to anyone.
End of article · #004