Home/Blog/AI Readiness Assessment in 2027
Updated November 2026

AI Readiness Assessment in 2027

By Arsh Singh/November 2026/9 min read

An AI readiness assessment is a structured check of whether your data, workflows, people, governance and leadership can support AI in daily work. In 2027, the people and workflow scores will decide your result more than the technology score.

Imagine a 300-person distribution company. The COO approves an AI assistant licence for every office employee in March. By June the dashboard says 80% of seats are "activated." By September, two teams use the tool weekly and everyone else opened it once. The purchase was fine. What nobody did was train the 25 team leads who decide how work actually gets done on a Tuesday afternoon.

That story is hypothetical, but I have spent 20+ years in growth work across 300+ brands, and the shape is familiar from every channel I have run. Buying is easy. Changing behaviour is the work. This post shows you how to find out where you stand and what usually breaks next.

Key takeaways

  • Almost every company invests in AI, yet just 1% of leaders call their company mature at deployment (McKinsey, 2025).
  • Only 21% of organizations using gen AI have fundamentally redesigned at least some workflows, and workflow redesign had the biggest effect on profit impact of 25 attributes tested (McKinsey, 2025).
  • Gartner predicted that at least 30% of generative AI projects would be abandoned after proof of concept by the end of 2025 (Gartner, 2024).
  • 63% of organizations lack, or are unsure they have, the data management practices AI needs (Gartner, 2025).

What does an AI readiness assessment actually measure?

A good assessment measures five things: data, workflows, people (managers above all), governance, and leadership alignment. It scores them for specific workflows, such as invoice processing or candidate screening, rather than for "the company" in general. A company-wide score hides the one weak spot that will stall the first rollout.

Start with the gap that makes the exercise necessary. 78% of organizations say they use AI in at least one business function (McKinsey, 2025). Yet only 1% of leaders describe their company as mature in deployment (McKinsey, 2025). Use is common. Readiness is rare. Those are different things.

The same McKinsey study found that executives misread their own workforce. Leaders estimated that just 4% of employees use gen AI for at least 30% of their daily work. Employees reported 13%, more than three times as many (McKinsey, 2025). I read that as a visibility problem. If leadership cannot see current use, it cannot see current risk either. People paste customer data into personal accounts because nobody gave them an approved option, and the HR or operations lead finds out after something goes wrong.

So the first job of an assessment is to learn what is already happening, including the unofficial use. Only then does it ask whether the official plan has a foundation.

For illustration: a 120-person services firm runs a short, anonymous survey before buying anything. Suppose 40 people say they already use a public chatbot for client emails. That one number changes the plan. The firm needs an approved tool list and a data rule before it needs a new platform.

Readiness is not a measure of how much AI you own. It measures how much of your daily work you can change safely, on purpose, with your current managers.

How do you run an AI readiness assessment step by step?

Pick three workflows, score each one on five dimensions from 0 to 3, find the lowest score, and fix that before you buy or build anything. This takes a few weeks of interviews, not a six-month audit. The scoring below is my own working scale, not an industry standard.

DimensionScore 0 looks likeScore 3 looks likeWho answers
DataKey data sits in personal spreadsheets and inboxesNamed data owner, documented source, access rules for this workflowOperations and IT
WorkflowProcess is undocumented and differs by personDocumented steps, clear handoffs, a human checkpoint at the AI stepTeam leads
Managers and peopleTeam leads cannot say what AI may and may not do on their teamLeads are trained, review AI use weekly, and can coach staffHR and team leads
GovernanceNo approved tool list, no ownerApproved tools, named owner, incident path, review dateExecutive and legal
Leadership alignmentGoal is "use more AI"One business metric per workflow and a named sponsorExecutive team

Five dimensions at a maximum of 3 gives a total out of 15. My rule of thumb: 0 to 5 means fix foundations first; 6 to 10 means pilot one workflow with tight scope; 11 to 15 means you can scale. One override matters more than the total. A zero in any single dimension caps the workflow at "not ready," because the zero is where the project will fail.

  1. Choose three workflows with real volume and a clear owner. Avoid anything that touches regulated decisions on your first pass.
  2. Interview the people who run them, not only the sponsor. Team leads know where the process really bends.
  3. Score each dimension using the table, with a short written reason for every number.
  4. Find the lowest score across all three workflows. That is your first fix.
  5. Set an adoption metric before launch.

On that last step, count active use, not licences or installs. We track the same thing in app work. In the Enjoy Movies Your Way case study, 76% of downloads were active monthly (38K+ monthly active users on 50K+ downloads by December 2023). Installs said little; monthly activity told the real story. Seat activation in an AI rollout works the same way.

The scoring is the easy half. Turning the lowest score into a sequenced plan with owners, budget, training and review dates is the harder half, and it is where a paid engagement earns its fee.

The data shows AI projects stall at the data, workflow and manager layers

The published evidence points the same way: AI projects rarely fail because the model is weak. They stall because the data is not ready, the workflow was never redesigned, or the managers were never equipped to lead the change. Here is the evidence side by side.

FindingFigureSourceWhat it means for your assessment
Companies calling themselves mature in AI deployment1%McKinsey, 2025Assume you are early, even if you have spent heavily
Companies planning to raise AI investment over three years92%McKinsey, 2025Spending pressure will rise before readiness does
Gen AI users who have redesigned some workflows21%McKinsey, 2025Score the workflow dimension honestly
Organizations unsure they have AI-ready data practices63%Gartner, 2025Check data ownership per workflow
Gen AI projects predicted abandoned after proof of conceptAt least 30%Gartner, 2024Define the business metric before the pilot
Agentic AI projects predicted canceled by end of 2027Over 40%Gartner, 2025Do not skip governance when tools start acting on their own

Gartner names the reasons behind the abandonment forecast: poor data quality, inadequate risk controls, escalating costs and unclear business value. Three of the four are readiness items you can score before spending a dollar. The fourth, cost, usually grows out of the other three.

My synthesis: none of these causes is a model problem. McKinsey's finding that workflow redesign had the biggest effect on profit impact of the 25 attributes it tested matters here (McKinsey, 2025). Add the 4% versus 13% visibility gap and the pattern is plain. The cheap part (the tool) gets the budget. The expensive part (redesigning work and training the people who run it) gets skipped.

If you want help turning scores like these into delivery, our AI automation services start from the workflow, not the tool catalogue.

Which mistakes make an AI readiness assessment useless?

The most common mistake is scoring technology while ignoring the managers who must lead the change. Tools get bought, licences get assigned, and adoption stalls because team leads were never shown how to change how their team works. The scenarios below are hypothetical, but each follows a mechanism you can test against your own plan.

Assessing the company instead of the workflow

For illustration: a 500-person firm scores itself 11 out of 15 overall and feels ready. Break it down and finance scores 14 while customer support scores 4. The average hid the truth. Always score per workflow.

Treating training as one webinar

Imagine an HR leader books a single one-hour session for all 400 staff. Staff leave with general awareness. Their managers leave with no answer to "can my team use this on client files?" Training that sticks is role-specific and repeated, and the managers get it first. Europe has made part of this a legal duty: the EU AI Act's AI literacy obligation has applied since February 2025 for organizations that provide or deploy AI systems.

Leaving governance as a document

AI governance is the set of owners, rules and review routines that decide which AI tools are allowed, what data they may touch, and who answers when something goes wrong. A PDF policy nobody owns scores a zero. The NIST AI Risk Management Framework is a useful free reference for structuring it, but you still need a named person and a review date.

Letting the vendor grade its own work

An assessment run by the company selling you the platform will find that you need the platform. Keep the assessor independent of the product.

Running it once

An assessment dated January is stale by June, because tools, staff behaviour and regulation all move. Treat the result as a baseline and re-score the lowest dimension every quarter.

AI readiness in 2027 will be a quarterly habit, not a one-time audit

These are my predictions, based on the published trends above, not reported facts. In 2027 the readiness question shifts from "can we run a pilot?" to "can our managers run AI-assisted work every week without us watching?" Four changes follow.

First, agent-style tools raise the governance bar. Gartner predicts over 40% of agentic AI projects will be canceled by the end of 2027, citing cost, unclear value and weak risk controls (Gartner, 2025). A tool that drafts text needs a reviewer. A tool that takes actions needs an owner, a limit and a log. Companies that score zero on governance today will find that gap expensive.

Second, manager capability becomes the scarce resource. Gartner's 63% data figure shows that data gaps are widespread, but data can be bought, cleaned or outsourced. A team lead who coaches people through a changed process cannot be bought in a week. Expect the best-run mid-size companies to put team leads through training before the next tool purchase.

Third, regulation arrives in stages. The EU AI Act has been phasing in obligations since 2025, and timelines for some provisions have been under discussion, so check the European Commission's current guidance rather than relying on a dated summary. Even a company with no EU exposure benefits from the discipline of documenting its tools and data use.

Fourth, scorecards become dashboards. The firms that win will re-score every quarter and tie each workflow to one business metric. Active use per team, not seats sold, will be the number the board asks for.

Frequently Asked Questions

How long does an AI readiness assessment take?

For a company of 50 to 1,000 people, scoring three workflows takes me roughly three to four weeks, mostly interview scheduling. That is a rule of thumb, not a benchmark. If a scope stretches past six weeks, it has usually grown into an AI strategy project. Keep the first pass narrow and re-score later.

Who should own the assessment inside the company?

Give it to an operations or HR leader with an executive sponsor, not to IT alone. IT can answer the data and security questions, but the biggest gaps sit with managers and workflow design. In my view, the owner should be whoever is accountable for how work changes, because that person holds the lever that matters most.

Do we need clean data before we start?

No, but you need to know where each workflow's data lives and who owns it. Gartner reports 63% of organizations are not sure they have AI-ready data practices (Gartner, 2025). Score data per workflow. A single well-owned dataset is enough to start a narrow pilot, and it exposes the cleanup you actually need.

What is the difference between AI readiness and AI governance?

Readiness asks whether your whole organization can adopt AI: data, workflows, people, governance and leadership. Governance is one of those five dimensions, covering who owns tools, what data they may use, and how problems are handled. I treat governance as a gate: a zero there stops a workflow from moving forward, whatever the other scores say.

How often should we repeat the assessment?

Re-score the lowest dimension every quarter and repeat the full scorecard once a year. Tools, staff habits and regulation all shift quickly, so a single audit goes stale. I would also re-run it whenever you add a new workflow or change a major tool, since the manager training needs change with them.

Conclusion

Three principles hold up across everything above. Score workflows, not the whole company. Treat managers as the main adoption lever, because tools do not change habits on their own. And measure active use, since licences and installs flatter you. A zero in data, governance or manager capability stops a project no matter how good the tool is.

You can run the first pass of the scorecard in this article yourself. The harder part is sequencing the fixes, training the right managers first, and setting up governance that people follow. That is where an outside view saves months. If you want to see where your company stands and what is likely to go wrong next, book an AI readiness consult. You can also read how we approach delivery on our AI automation page.

Want a second pair of eyes on your growth?

Book a free 30-minute strategy call. Bring your numbers, leave with two or three moves worth making. No pitch, no deck.

Book a Free Strategy Call