Assessment
Will your pilot survive day 91?
A twelve-question assessment of the operating-model decisions that determine whether your AI pilot graduates to production or quietly disappears.
Ninety-five percent of GenAI pilots never reach production. Most die because four operating-model questions went unanswered before the demo: who owns this on day 91, what graduation looks like, which P&L absorbs the cost, and what integration debt the org is willing to take on. This diagnostic puts those questions, and a few others senior practitioners ask, in front of you. If you are sponsoring a pilot, or about to greenlight one, spend five minutes here before the next steering committee.
Begin the diagnostic
Twelve questions across five operating-model dimensions. About five minutes. No account required. Your answers stay in your browser.
The diagnostic runs in your browser and your answers stay there. Our analytics counts that a result was shown, with the score and band, and nothing else.
The five dimensions
What the twelve questions cover
Executive Mandate
Whether a named executive, rather than a steering committee, owns AI outcomes with real authority, real budget, and compensation tied to delivery.
- Who is accountable when the board asks why AI investment did not show up in last quarter's results?
- When an AI initiative needs an unplanned eight-figure investment to make it to production, what happens?
Operating Model
Whether the four operating questions are answered before a pilot starts: day-91 ownership, graduation gate, P&L absorption, integration debt.
- Your AI champion just got promoted to a new role. Who runs the system Monday morning?
- What separates a pilot that graduates into production from one that doesn't, at your firm today?
- How does your most promising current AI pilot integrate with the systems of record it depends on?
Data Foundations
Whether the data, lineage, and access behind the use cases on the table have been mapped for those use cases specifically.
- For the top AI use case on your roadmap, do you know where the underlying data came from, who owns it, and what changed last quarter?
- How long would it take an engineer on a new AI project to get production-quality access to the data they need?
Engineering Capacity
Whether the organization has its own engineers accountable for systems after launch, or relies on vendor licenses and pyramid staffing.
- Who writes the production code that runs your AI systems, and where do they sit organizationally?
- Six months after an AI system goes live, who is responsible for retraining, monitoring drift, and shipping improvements?
Value Realization
Whether the ROI case was designed into the system before launch with a finance partner who will book it.
- When did you and finance agree on the dollar value the top AI system on your roadmap will produce, and how it will be measured?
- How do you separate the value AI produced from the value the business would have produced anyway?
- If asked at the next board meeting what AI contributed to last year's earnings, what would you say?
Methodology
How the diagnostic works
Twelve questions across five operating-model dimensions: ownership and accountability, graduation criteria, P&L and budget absorption, integration and technical readiness, and measurement. Each answer carries a weight, and the dimensions are not equal. Ownership and graduation criteria are weighted most heavily because they are the two failure modes we see most often inside client AI programs. The diagnostic takes about five minutes to complete and returns a score between zero and one hundred, along with the dimension where your pilot is most exposed. The scoring logic is the same one we use in our first working session with new clients.
About the diagnostic.
How is the score calculated?
Each question maps to one of five operating-model dimensions and carries a weight based on how often we see it predict pilot failure across client programs. Ownership and graduation criteria are weighted most heavily. The total is normalized to a zero-to-one-hundred score, with the weakest dimension surfaced alongside it so you know where to look first.
Can I take it again?
Yes. Many users run it once on their own, then a second time with their program lead or sponsor in the room. The questions tend to produce a more useful conversation than they do a useful score. Retaking it with a different person answering is one of the more productive uses of the tool.
Will I get spam after I take this?
No. The diagnostic does not ask for your email, your company, or your name, and there is nothing to opt into. If you want to talk through your result with us, you can book a call at the end. Otherwise the interaction ends when you close the tab.
What if none of the answer options fit my situation?
Pick the option closest to your reality and keep going. The diagnostic is calibrated for the patterns we see most often rather than for every edge case. If your situation does not fit any of the four options on a given question, that is itself a useful signal. Bring it up on the call.
Can I share my result?
The result lives in your browser session and disappears when you close the tab, so there is no shareable link. If you want to share the output with a colleague, screenshot the result page or take the diagnostic together. There is no share-this-score button, on purpose, because the conversation is worth more than the number.