AI Readiness Assessment: The Questions to Ask Before Your First Pilot

Readiness has four layers: ownership (who owns the outcome), process (is the workflow documented well enough to hand to AI), data (is your CRM data trustworthy for the fields this pilot needs), and infrastructure (can this actually integrate with what you have). Score yourself honestly against a specific pilot, not your organization in the abstract, and the gap usually shows up in one layer more than the others. Fix that layer first.

This is a self-serve framework. No gate, no signup, no proprietary score. Just the questions. Most readiness assessments in this space are gated behind a form, which is a strange choice for something meant to help a VP of Sales make a fast decision. The version below is the whole thing, on the page, because giving it away is more useful to you and a better test of whether the framework holds up than hiding it ever was.

Layer 1: Ownership

• Who owns this if the pilot underperforms, by name, not by department?
• Does that person have the authority to kill the pilot, not just champion it?
• Is there a defined kill criterion, stated before launch, not decided after the fact if it goes badly?
• Will this person actually use the output weekly, or is this someone else’s workflow being automated on their behalf?
• If the tool makes a wrong call, who’s accountable for catching it?

If nobody can answer the first question with a name, that’s the gap. A pilot with no named owner tends to drift, get partially adopted, and quietly die without anyone officially deciding to kill it. Ownership sounds like the easiest layer to score well on, since most teams can name a champion quickly. The harder version of the question is whether that person has the standing to actually kill the pilot if it’s not working, or whether they’re the same person who fought to get budget for it and has a reason to keep it alive past the point it should have ended.

Layer 2: Process

• Could someone new to the role follow this workflow today from written documentation alone, or does it live only in a few people’s heads?
• Is the workflow you’re piloting stable, or does it change meaningfully deal to deal, rep to rep?
• Have you actually mapped the steps end to end, or are you assuming the tool will figure out the steps for you?
• Is there a version of this workflow that already works well manually, that the tool is meant to speed up, versus one that’s broken and you’re hoping AI fixes it?
• Who currently does this work, and have they been part of scoping the pilot?

AI tools automate what you hand them. A workflow nobody’s documented gets automated inconsistently, because the tool inherits whatever ambiguity was already there. This layer is easy to overrate your own team on. “Everyone basically knows how we do this” usually means three senior reps know, informally, and a new hire would have to ask around for a month to piece it together. Try writing the workflow down in five steps before the pilot starts. If that’s hard to do, the tool will find it hard too.

Layer 3: Data

• Would two different people pulling the same report from your CRM agree on the numbers?
• Are the specific fields this pilot depends on filled in consistently, not just marked required?
• Is activity logged in the CRM itself, or in personal notes and spreadsheets the tool can’t see?
• Do you have enough historical data in the lookback window the pilot needs, and is it trustworthy?
• Has anyone actually looked at a sample of records for this pilot’s fields, or are you assuming the data’s fine?

This is usually the layer with the most silent gaps, because bad data doesn’t announce itself; it just produces confidently wrong output. For the deeper version of this check, see CRM data hygiene: the prerequisite nobody wants to do before AI.

Layer 4: Infrastructure

• Does this tool need to read from or write to systems it currently can’t reach?
• Who owns the integration work, and do they have time on the roadmap for it?
• What happens to your data and any build artifacts if you cancel the tool later?
• Does the tool need real-time access, or is batch sync enough for this workflow?
• Have you tested the integration against your actual data, not just the vendor’s demo environment?

Infrastructure gaps are usually the most visible ones and the easiest to underestimate. “We integrate with your CRM” covers a huge range of actual depth, from a one-way field sync to full bidirectional real-time access. Ask a vendor to show the integration against a sample of your own data, not their demo environment, before you assume it works the way the sales deck implies. This is the same “bring your own scenario” instinct that matters in any vendor evaluation, and it applies just as much when you’re checking your own readiness as when you’re checking a vendor’s claims.

How to score yourself

Don’t force a number. A fake precision score (out of 100, with decimal points) implies more validation than a self-assessment can actually provide. Instead, read across the four layers and find the one with the most “no” or “not sure” answers. That’s your biggest gap, and it’s what to fix before the pilot, not during it.

Most teams find their weakest layer is data or process, not ownership. Ownership is usually clear on paper. Whether the workflow is actually documented and the data actually trustworthy is where the honest answers get uncomfortable. It’s also common to score reasonably across all four layers individually and still not be ready, because readiness for one pilot doesn’t transfer automatically to the next one. A team that’s ready to pilot a research assistant, where the data dependency is light, might not be ready for a forecasting tool, where stage and close-date accuracy matter far more. Re-run the relevant questions for each new pilot, not just once for the organization as a whole.

What to fix first, by gap

If you’re carrying an AI mandate from leadership with no plan attached, this assessment is a reasonable first move before any vendor conversation. See Your AI mandate has no plan. Here is where to start.. And if you haven’t inventoried what you’re working with yet, the stack audit is the input that makes this assessment concrete instead of guessed. See How to audit your sales tech stack before you buy anything AI.

Where this self-serve version runs out is stress-testing your own honest answers. That’s what an outside diagnostic conversation is actually useful for, and it’s a reasonable next step once you’ve been through the four layers yourself.