AI Readiness Assessment: A Practical Guide (and Checklist)

An AI readiness assessment is a structured review of whether your organization can actually put AI into production and keep it there. It scores four dimensions: data readiness (can the model reach clean, permissioned data), workflow readiness (is the process mapped well enough to redesign), talent readiness (do people have the skills and the time), and adoption readiness (will the team actually use it after launch). The last one is where most projects quietly die. Run the checklist below, score each dimension honestly, and you will know before you spend a dollar whether a use case is ready or whether you are about to fund another pilot that demos well and dies in month six.

Why do I need an AI readiness assessment at all?

Because the failure rate is brutal, and it is not a technology problem. MIT found that 95% of AI pilots deliver no measurable business impact. RAND puts the share of AI projects that fail to meet their goals at roughly 80%, far higher than typical IT projects. Those numbers are not caused by weak models. Models are commoditized and improving weekly. Projects fail because the organization around the model was not ready: the data was locked in a system nobody could query, the workflow was never mapped, the team had no time to learn the tool, or the people meant to use it were never brought along.

A readiness assessment forces those questions to the surface before you commit budget. It is cheaper to learn you are not ready in an afternoon than in month six. And readiness is not binary. You come out with a per-dimension score that tells you exactly what to fix first, rather than a vague "we should do more AI."

What does "data readiness" actually mean?

Not a data lake or a fashionable warehouse. Data readiness is whether the specific model you want to build can reach the specific data it needs, in a usable state, with permission to use it.

That means four things. The data exists and is captured, not trapped in an inbox or a PDF. It is accessible: there is an API, an export, or a query path, not a screen a human has to read. It is clean enough that you would trust a decision made from it today, without a three-week cleanup first. And you hold the rights to use it, including customer data under whatever privacy terms you operate on.

A common trap: teams assume that because data is "in the CRM," it is ready. Then they find half the fields are free text, the other half are blank, and the one field that matters is filled in differently by every rep. That is a data readiness failure, and it is fixable, but only if you catch it before you scope the build.

How do I assess workflow readiness?

McKinsey's finding is blunt: end-to-end workflow redesign is the number one driver of AI value. Not the model, the workflow. So workflow readiness measures whether you understand the process well enough to redesign it.

Ask yourself: can you draw the current process end to end, including the handoffs and the exceptions? Do you know how long each step takes and who owns it? Do you know where the real bottleneck is, as opposed to where you assume it is? If you cannot map the current state, you are not ready to redesign it, and dropping an AI tool into a process you do not understand just automates the confusion faster. The strongest signal of readiness here is a single named owner who can walk the whole flow without hedging. The weakest is "it depends who is doing it." This is the dimension teams most often overrate, because the process feels obvious to the people living in it until you ask them to write it down.

What is talent readiness, and how much does it matter?

Talent readiness is whether the people involved have the skills, and just as important, the time, to build and run the thing.

Two failure modes. The first is a genuine skills gap: nobody can evaluate a model output, write a decent prompt, or judge whether a result is good. That is trainable and often faster to close than people expect. The second is subtler and more dangerous: the skills exist but the capacity does not. The one person who understands both the workflow and the tooling is already at 110% on their day job. A project staffed on someone's nonexistent spare time is not ready, no matter how skilled that person is.

Assess this honestly by asking who owns it after launch, whether that person has real hours protected for it, and whether a second person could keep it running if the first one leaves. If any answer is shaky, note it. It is fixable with a fractional team or protected time, but only if you name it up front.

What is adoption readiness, and why does everyone skip it?

Adoption readiness, or stickiness, is whether the team and the process will actually absorb the change. It is the dimension everyone ignores, and it is the reason pilots demo beautifully and then flatline.

Here is the pattern. A tool gets built. It works in the demo. Then it lands in a real team that was not consulted, does not trust it, has no incentive to change how they work, and quietly routes around it. Six months later usage is near zero and someone declares AI "did not work here." The model was never the problem. Adoption was.

Assessing it means asking uncomfortable questions early. Do the people who will use this know it is coming, and did they help design it? Does the new way cost them more effort in the first weeks, and is anything protecting them through that dip? Is there a clear owner accountable for usage, not just for shipping? Does anyone's incentive actually change? If the honest answers are no, the use case is not ready, and no amount of model quality will save it. Spinning up the tool is the easy 20%. Adoption is a large part of the durable 80%, and it has to be designed in from the start.

The AI readiness checklist you can self-run

Score each item yes or no. A "no" is not a disqualifier, it is a work item. Count the no's per dimension. Any dimension with two or more no's is not ready yet, and that is your first fix.

Data readiness

• The data the model needs exists and is captured today.
• There is a real access path (API, export, query), not a human reading a screen.
• The data is clean enough to trust a decision from right now.
• You have the rights to use it, including customer data.

Workflow readiness

• You can draw the current process end to end, including handoffs and exceptions.
• You know how long each step takes and who owns it.
• You know where the real bottleneck is, with evidence, not assumption.
• One named person can walk the whole flow without hedging.

Talent readiness

• Someone can evaluate model output and tell good from bad.
• A named owner will run this after launch, with protected hours.
• A second person could keep it running if the first one leaves.
• The team has time for this, not just interest in it.

Adoption readiness

• The people who will use it know it is coming and helped shape it.
• Someone is accountable for usage, not just for shipping.
• The first-weeks effort dip is acknowledged and protected against.
• At least one person's incentive or workflow actually changes to favor the new way.

The lowest-scoring dimension is where your risk lives, and it is almost never the one your team expected.

How this maps to a real engagement

This checklist is the self-serve version of the first phase of our work, the Audit and Teardown. There we run the same four dimensions with evidence rather than self-report: we pull the actual data and test whether it is reachable, we map the real workflow with the people who run it, and we pressure-test adoption by talking to the humans who would use the tool, not just the sponsor who wants it.

The output is the Durable AI Index, which scores each candidate use case on Impact, Feasibility, and Stickiness. Readiness feeds directly into Feasibility and Stickiness. A use case can be high impact and still score low because the data is not reachable or the team will not adopt it. That is exactly the expensive mistake a readiness assessment catches before you build.

Frequently asked

How long does an AI readiness assessment take? The self-run checklist above takes an afternoon per use case. A formal Audit and Teardown across a portfolio of candidate use cases typically runs a few weeks, because the work is in gathering evidence and talking to the people who run the process, not in filling out a form.

Do we need to be "AI ready" everywhere before we start? No. Readiness is per use case, not company-wide. You almost certainly have one or two use cases that are ready now and several that are not. The point is to find the ready ones and sequence the rest, not to wait until everything is perfect.

What is the difference between readiness and feasibility? Feasibility is mostly technical: can it be built. Readiness is broader: can it be built, adopted, and sustained in your organization. A feasible project can still be unready if the team will not use it.

We have great data. Doesn't that mean we are ready? Data is one of four dimensions. Great data with an unmapped workflow and a team that was never consulted still fails. The dimensions are not additive: a hard failure in adoption sinks the project regardless of how strong the data is.

Where to go next

If your checklist turned up more no's than you expected, that is the assessment doing its job. Start by scoring your top two or three candidate use cases and see which dimension is dragging. If you want the evidence-based version with a scored use-case portfolio, that is our Audit and Teardown, and you can see how it applies to your model on the use-case hub. When you are ready to talk specifics, get in touch.

To go deeper, read why AI pilots fail for the adoption failure pattern in detail, and how to prioritize AI use cases for turning a set of ready use cases into a sequenced plan.