How to Run an AI Readiness Assessment (In-House or With a Partner)

    Somebody on the board has said the words "AI readiness", and now there's a line in the quarterly plan with no clear next step underneath it. Maybe it came from a competitor announcement, maybe from McKinsey's 2025 finding that 88% of businesses have AI live in at least one function and only 39% report a measurable EBIT effect. Either way, the question landed and the answer is missing. A lot of businesses skip straight to vendor demos at this point, which is how MIT NANDA ends up with a 95% pilot-failure number. The honest move is an assessment first.

    Two routes are available, and the right one depends on how far along you are. The free AI Necessity Test on this site walks one candidate use case through five questions in eight minutes, and it produces a verdict you can defend in a leadership meeting on Monday. The paid AI Readiness Audit is a 14-day engagement, with two digital Gemba sessions, three process maps, and a board-ready roadmap at the end. The DIY route is the right place to start. If, halfway through, the answers feel softer than the questions deserve, that's the signal to escalate. Both routes use the same four-step structure underneath, and the rest of this article walks through it.

    The DIY assessment, 4 steps, 30 minutes

    The DIY route gets a single use case to a clear go, no, or fix-the-foundations-first verdict in under an hour of focused work across two people. It won't replace a full audit. It can't see the three other use cases competing for the same budget, and it relies on the team being honest with themselves about gaps. What it does do is filter out the obviously broken briefs before procurement starts, which is where most of the wasted spend in mid-market AI actually happens. Run the four steps below in a single sitting, ideally with the operating sponsor in the room.

    Step 1: Define the candidate process in one sentence

    Write down the workflow you're considering for AI in a single sentence with the format: 'We want to [verb] [thing] for [team] so that [outcome], measured by [metric].' If the sentence doesn't fit, the use case isn't ready.

    Worked example, from a B2B services firm we walked through this with last quarter: "We want to qualify inbound demo requests for the sales team so that fresh conversations start within 24 hours, measured by qualified-leads-per-week versus the current 4.2-day cycle." That sentence survives contact with procurement. The common pitfall is loading two workflows into one sentence ("qualify leads and write the follow-up emails"), which guarantees a fuzzy pilot. You're done when the sentence names one workflow, one team, one metric, and a number the team would recognise on the current dashboard.

    Step 2: Score the four readiness pillars

    Score Process, Data, People, and Strategy on a 1-5 scale. A pillar at 1-2 is a blocker that must be fixed before the pilot. A 3 is workable. A 4-5 is ready.

    Score each pillar with the person closest to it in the room. Process: is the workflow documented end-to-end on a single page. Data: can the team get to the records, are they clean, does someone own them. People: is there a Champion with the authority to unblock decisions and the credibility to land the change. Strategy: is the metric baselined, is the budget signed off, is the kill criterion written down. The common pitfall is scoring on aspiration ("we're a 4 on data because we're investing in it"). Score on what's true today, not on what the roadmap promises. You're done when each pillar has a number and one line of evidence behind it.

    Step 3: Identify the weakest pillar and its specific gap

    Name the single biggest gap in the weakest pillar. 'Data is messy' is too vague. 'Customer records live in three CRMs with inconsistent IDs' is specific enough to fix.

    Take the lowest-scoring pillar and name the specific thing that's broken. "People is a 2 because no one currently uses AI tooling in the team and the marketing lead is openly sceptical" is actionable. "People aren't ready" isn't. The common pitfall is naming a symptom instead of a cause. "The team doesn't trust the output" is a symptom. The cause is usually further upstream, in the brief, the data, or the measurement framework. You're done when the gap is specific enough that a 30-day plan to close it would write itself in the next paragraph.

    Step 4: Decide on the next 30 days

    Pick one of three paths: fix the weakest pillar (if a pillar scored 1-2), run a 4-week pilot on the strongest use case (if all pillars scored 3+), or invest in deeper assessment (if you can't honestly score yourself).

    Three paths, picked on the scores. Path one, fix the weakest pillar: if any pillar landed at 1-2, that's the next 30 days, full stop, no procurement conversation until it moves to a 3. Path two, run a 4-week pilot: if every pillar scored 3 or higher, scope a time-boxed test on the strongest candidate workflow with a baselined metric and a kill date. Path three, deeper assessment: if the team couldn't honestly score itself, or if more than two use cases are competing for the same quarter, get an outside eye on it. You're done when the next 30 days has one named owner and one written outcome.

    When to bring in an external partner

    The DIY route covers a lot of ground for the cost of an afternoon. There are five situations, though, where doing it in-house costs more than it saves. The common thread across all five is that the gap isn't the assessment itself, it's the honesty or the breadth the assessment needs to land properly. If any of these triggers apply, the fee for an external audit is the cheapest insurance against the £50k-plus that gets wasted on a misscoped pilot.

    Trigger 1: You can't honestly score yourself

    The first trigger is the one most businesses won't admit to. A lot of teams score themselves a 4 on data because the alternative is uncomfortable, then watch the pilot break on dirty records in week six. We see this constantly. The score is aspirational, not observational, and there's no neutral party to call it. An outside audit produces a written score backed by evidence the board can read, which removes the politics from the conversation. If the team is hesitating on the score, that's the signal.

    Trigger 2: Multiple pillars score 1-2

    Two or more pillars at the bottom of the scale isn't a project plan, it's a transformation. The sequencing matters more than the work, and getting the order wrong wastes a quarter. The audit produces a sequenced fix plan with named owners and dated milestones, in writing, as part of the deliverable. That document is the difference between fixing three pillars in 90 days and fixing one and a half in 180. If the in-house scoring keeps returning multiple low pillars, the assessment itself isn't the bottleneck. The fix plan is.

    Trigger 3: You've tried before and it didn't work

    A failed first attempt raises the bar on the next one. The board has weaker patience, the team has the scars, and the budget owner has internalised "AI doesn't work here". The honest read on a dead project is that one of the four readiness pillars was the actual cause, and the in-house team is rarely the right people to run that post-mortem on themselves. An outside audit produces a named root cause from the previous attempt and a sequenced plan for the next one. Without that document, the second pilot is the first one with a different vendor name on the invoice.

    Trigger 4: Budget over £50k is on the table

    At spend levels above £50k, the cost of a misscoped pilot exceeds the cost of the audit by an order of magnitude. A 14-day engagement that produces a written ROI projection per opportunity and a board-ready roadmap is the cheapest way to convert a large budget into a defensible decision. A lot of finance directors will sign off the audit faster than the pilot for exactly this reason. If the budget conversation is real, the audit conversation should be ahead of it.

    Trigger 5: Multiple use cases competing for the same budget

    The DIY tool scores one use case at a time. When the conversation is "we've got six candidates and one quarter of budget", that's a prioritisation problem the in-house team can't easily solve without a shared framework. The audit produces an AI Opportunity Scorecard with ROI projections across the top three workflows, which lets the leadership team pick on numbers rather than opinions. Internal politics rarely survive a written scorecard. They almost always survive a Slack thread.

    What a professional AI Readiness Audit covers

    The LeverageAI AI Readiness Audit is a 14-day remote engagement. Week one is observation. Two digital Gemba sessions, live screen-share walkthroughs of the actual tools and workflows the team uses today, not the documented version of them. Team interviews with the operators, not the planners. A current-state value stream map of the top three highest-waste workflows, with handoffs, cycle times, and decision points written down on a single page each.

    Week two is analysis. A prioritised use case shortlist, ranked against an AI readiness scorecard, with ROI projections per opportunity. A capability gap report covering Process, Data, People, and Strategy, with named fixes and a sequenced 30-60-90 plan. A 90-day pilot roadmap on the highest-scoring workflow, with a baselined metric, a target, and a written kill criterion. A competitive AI landscape brief for the sector, so the board sees where the business sits relative to peers. A board-ready briefing document and a recorded executive summary video for the directors who couldn't make the presentation. The audit closes with a 30-minute follow-up call two weeks after delivery, to pressure-test the first decisions made on the back of the work.

    Two pieces of structural honesty are built into the engagement. The first is the guarantee: at least three actionable AI opportunities with positive ROI projections, or the fee is refunded in full. The second is the credit: the audit fee credits in full toward the 90-Day Implementation Sprint, so the audit isn't a cost added on top of the build, it's the first 14 days of it. That structure exists because the right shape of partner contract is outcome-tied, not effort-billed. Anything looser produces the kind of engagement the 95% of failed pilots started with.

    Where to start

    References

    • MIT NANDA. "The GenAI Divide: State of AI in Business 2025." MIT Media Lab Project NANDA, 2025. Source of the 95% no-measurable-return figure that frames why an assessment matters.
    • McKinsey & Company (QuantumBlack). "The State of AI." 2025. Source of the 88% adoption versus 39% measurable EBIT impact gap cited in the opening section.
    • Gartner. "Data Quality and AI Project Failure." Ongoing research, 2024-2025. Source of the data-quality-as-leading-cause finding referenced in the data pillar scoring guidance.
    • Deloitte AI Institute. "State of Generative AI in the Enterprise." 2026. Source of the worker-skills barrier finding referenced in the People pillar guidance.