The AI Readiness Checklist Your Board Actually Needs

Most AI readiness assessments are useless

Search “AI readiness assessment” and you’ll find consultancy frameworks that run 30 to 80 pages. They evaluate data maturity, technology infrastructure, organisational culture, talent pipelines, governance frameworks, and change management capacity. By the time you’ve completed one, three months have passed, nothing has been built, and the executive team has moved on to the next priority.

These frameworks aren’t wrong exactly. They’re just designed for Fortune 500 companies with dedicated strategy teams. For a mid-sized European company (50 to 500 people, real operational pressure, limited time for self-assessment) they’re the wrong tool entirely. The OECD AI Policy Observatory tracks SME AI adoption across member states and the pattern is consistent: mid-market readiness is gated by operational specificity, not by maturity-model scoring.

The real question isn’t “are we ready for AI?” in some abstract sense. It’s “do we have the conditions to run a successful first project?” That’s a much simpler question to answer. And the answer usually depends on seven things.

Seven questions that actually matter

1. Can you name a specific process that costs you identifiable money or time? Not “we want to be more efficient.” A specific process. “Our team spends 35 hours per week on invoice data entry.” “Client onboarding takes 14 days and involves 6 handoffs.” If you can’t point to a concrete process with a measurable cost, you’re not ready to build - you need an audit first. This is the single most important question on the list. Everything else follows from having a real target.

2. Is the data for that process already digital? It doesn’t need to be clean. It doesn’t need to be in a data warehouse. It just needs to exist in electronic form somewhere: spreadsheets, email, PDFs, a CRM, an ERP. If the process runs entirely on paper and phone calls, there’s a digitisation step before AI becomes relevant. That’s fine, but it changes your timeline and budget.

3. Is there someone internal who owns the problem? Not the technology - the problem. A person who lives with this process daily, understands its pain points, and will be accountable for whether the solution works. MIT Sloan’s state-of-AI tracking consistently flags absence of a clear business owner as one of the strongest predictors of a project stalling in pilot. This person doesn’t need to be technical. They need to care about the outcome.

4. Can you define what “success” looks like in a number? “Faster processing” is not a success metric. “Processing time drops from 14 minutes to 3 minutes per item” is. “Better customer experience” is not a metric. “First response time drops from 4 hours to 15 minutes” is. If you can’t put a number on it before you start, you won’t know if you’ve succeeded after you finish.

5. Does leadership support a pilot? You don’t need the board to approve a multi-year AI transformation. You need someone with authority to say: “Yes, spend four to six weeks and a defined budget testing this on one process.” That’s the minimum viable buy-in. If even that level of support doesn’t exist, the organisational readiness isn’t there, regardless of how good the data is.

6. Is the team willing to change how they work? This one gets underestimated. A technically perfect AI system that nobody uses is a failed project. The people who currently do the work need to be involved early, understand what’s changing and why, and see the tool as something that helps them rather than threatens them. McKinsey’s 2024 State of AI report found that change resistance accounts for 38% of AI implementation delays. Ask the team, not just the leadership.

7. Can you commit to a timeline measured in weeks, not quarters? The best AI projects start small and deliver fast. If your organisation’s default timeline for any new initiative is “six months of planning followed by twelve months of execution,” AI will die in committee. The companies that succeed commit to shipping something (even something small) within 4 to 8 weeks. Speed creates momentum. Momentum creates internal trust. Trust creates budget for the next project.

Using the checklist in practice

This checklist works in two contexts. First, as a self-assessment before you engage with any AI vendor, consultant, or internal project proposal. Run through the seven questions. If you can’t answer the first three, the right next step is an operational audit, not an AI project. The audit gives you the answers you’re missing.

Second, as a communication tool with your board or leadership team. Most boards are being asked “should we invest in AI?” when the real question is “do we have the conditions for a successful first project?” The seven questions reframe the conversation from abstract strategy to concrete operational assessment. Boards respond well to this. It’s specific, it’s measurable, and it gives them a clear basis for a yes-or-no decision.

One pattern we see repeatedly: companies that score well on questions 1 through 4 but poorly on 5 through 7 don’t have a technology problem or a data problem. They have an organisational problem. The fix isn’t more technology assessment. It’s stakeholder alignment. Sometimes a single meeting where leadership sees the cost of the status quo is enough to unlock the support you need.

• Run through all seven questions before engaging any vendor or starting any AI project
• If you can’t answer questions 1, 3, and 4, start with an operational audit, not an AI build
• Use the checklist to reframe board conversations from “should we do AI?” to “are we ready for a first project?”
• Weak scores on questions 5-7 point to organisational issues, not technical ones. Solve those first
• Revisit the checklist after each project; your readiness improves with every successful implementation

The final point is worth emphasising. Readiness is not static. Every company that ships a successful first AI project becomes dramatically more ready for the second one. The data gets better. The team gets more comfortable. Leadership trusts the process. The checklist becomes easier to pass each time. The hardest part is always the first project - which is exactly why the bar for starting should be lower than most frameworks suggest.

The audit-first engagement shape that converts a checklist pass into a shipped first project sits on the process page. The deeper question of whether you actually need an AI audit is the right read for companies sitting at three or four out of seven and unsure what to do next. A structured audit is the operational version of this checklist, applied to your specific process rather than to the org in the abstract.

When seven questions aren’t enough

The seven-question checklist is calibrated for a 50-500 person company running its first or second AI project on an internal process. Outside that profile, the bar for “ready” needs to be set differently and the light-touch checklist hides the real risk.

• Regulated workloads where the system itself is the compliance artefact. EU AI Act high-risk use cases, medical-device-adjacent decisions, credit decisioning, and HR screening don’t pass on three out of seven. They need formal risk classification, documented human-oversight protocols, and a conformity assessment plan before any build starts. The longer framework is the right tool here.
• The process touches customers in real time and a wrong answer is visible. If the AI output goes straight to a paying customer (pricing, eligibility, automated advice) the seven-question bar is too low. You need a failure-mode catalogue, a rollback plan, and an escalation path before shipping, none of which the checklist surfaces.
• Enterprise companies running a portfolio decision, not a project decision. When the board is allocating budget across 30 candidate use cases at a multinational, the question isn’t “are we ready for a first project.” It’s “which projects clear the bar and in what sequence.” That’s a portfolio scoring model, and the seven questions are the per-project filter inside it, not the whole tool.
• You’ve already failed a first AI project. Companies on attempt two carry organisational baggage the checklist doesn’t see: stakeholder fatigue, sceptical finance, a sunk vendor relationship. Passing the seven questions on paper is genuinely possible while the real readiness blocker is internal trust. The right diagnostic at that point is a post-mortem of attempt one, not a fresh readiness pass.