AI vendor pitches often open with some variation of “the technology is ready.” That statement says little about whether a particular firm, workflow or group of users is ready to adopt it.
The slide does not show the expression on a senior associate's face when you announce an AI-powered document-review tool. It misses the practice lead who nods in the steering committee and never logs in, and the IT director shaped by several transformation projects that failed to deliver.
Technical capability and organisational readiness are separate evidence questions. Find out where people stand before the firm commits substantial money and reputation to a rollout.
Culture is too soft to measure. We just need to get on with it and people will adapt.
I understand the instinct. You're running a firm that bills by the hour and do not want an abstract HR exercise. Buying the tool and providing training may still leave incentives, trust, leadership behaviour and workflow untouched. Measure those conditions so the adoption plan addresses this firm instead of relying on unsupported industry failure percentages.
Measuring cultural attitudes means a structured, repeatable assessment using behaviour and evidence as well as survey responses. It should show where concerns and support sit, why they exist and which change interventions deserve attention before a broad licence commitment.
When good technology meets a bad room
I was working with a 300-person professional services firm last year - the kind of firm that does everything thoughtfully. They'd done their research. Selected a genuinely good AI tool for proposal generation. Built a solid business case. Got board sign-off. Ran training sessions. The whole playbook.
Six months in, recorded adoption was 14%. The tool worked well for the people using it. Others had found workarounds or carried on with the old process. The CTO said: “We gave them everything they needed.” The implementation had supplied technology and training; it had not examined the social conditions around use.
The project had not examined attitudes before launch. It assumed that capable technology and thorough training would be sufficient. The adoption result challenged that assumption.
Interviews suggested three broad groups. Enthusiasts wanted to use the tool quickly. A smaller group of senior people opposed it because they saw AI as a threat to the expertise model that had built their careers. The largest group was undecided and watched what respected colleagues did. When senior sceptics signalled that the tool was not part of serious practice, many in that middle group followed their lead.
The firm had made a six-figure commitment while most people remained outside the active user group. Earlier measurement could have exposed the concerns, identified credible champions and given the undecided group evidence from peers.
I had a shorter, messier version of this conversation with a legal firm. Six weeks into its pilot, a senior partner sent a firm-wide email questioning whether the AI tool was “appropriate for client-facing work.” The message exposed a concern that the project had not addressed before launch and disrupted the programme. Asking that partner earlier would have created an opportunity to examine the risk, evidence and permitted use before a firm-wide challenge.
The dimensions that actually tell you something
Right. So how do you measure something as apparently woolly as "cultural attitude toward AI"? You break it down into dimensions that are specific enough to assess and act on.
We use five when we work with firms on AI readiness. None of them require a psychology degree to evaluate. All of them produce data you can do something with.
The first is general experience of change. How did the firm absorb a recent CRM, team restructure or billing-model change? Look at the last three significant initiatives: how long did adoption take, where did use stall and what was the gap between announced go-live and real work? System logs, project reviews and direct conversations with team leads provide a stronger basis than memory. Score the dimension from 1 to 5 and record the evidence behind the score.
Trust in technology is subtler. Do partners trust CRM pipeline data enough to use it in resourcing decisions? Do people maintain shadow spreadsheets and feed central systems without relying on them? Those behaviours reveal the history an AI rollout inherits. Survey a cross-section with specific scenarios: “If the system flagged a client as at risk of churn, what would you check before acting?” The answers show both trust and the conditions people need for appropriate challenge.
Then there's data literacy. Can people read a dashboard, explain a rate and distinguish correlation from causation? AI outputs also require interpretation and challenge. Give a sample group a real dashboard from an existing system and ask what it shows, what it cannot establish and which decision it could support. That task produces better evidence than a confidence rating alone.
Willingness to experiment is the fourth. Some firms have a culture where people try new approaches, share what worked, and aren't punished for things that didn't. Others - and this is common in professional services, where mistakes can have regulatory consequences - have a culture of caution. Neither is inherently wrong, but a firm with low experimentation tolerance needs a very different AI rollout strategy. Ask people when they last tried a new tool without being told to. Ask team leads whether their people suggest process improvements. Look at whether pilot programmes in the past attracted volunteers or had to be staffed by assignment.
Leadership signals are the fifth dimension. What do senior leaders demonstrate in practice? Does the managing partner use the tools they endorse? Do practice leads discuss appropriate AI use, limitations and results? A gap between the official narrative and observable behaviour will influence the rest of the firm.
I've sat in firms where the partner championing AI adoption still dictated everything to an assistant and had someone else print emails. Anonymous upward feedback can help surface the gap safely. Ask whether a line manager visibly uses approved tools and whether the team leader has discussed AI use in a recent meeting. Combine responses with observation rather than treating one survey as a verdict on an individual.
What to do with the numbers
Once you've scored across all five dimensions, you'll have a cultural readiness map. The results may reveal patterns that are useful to group as champions, sceptics and an undecided middle, without treating people as fixed categories.
Champions show confidence or useful experience across several dimensions. Make them visible, involve them in the pilot and let them present evidence to the wider firm. Do not confuse enthusiasm with representative user insight or give champions permission to bypass controls.
Sceptics may raise legitimate concerns about data quality, client confidentiality or output reliability. Engage those concerns and test them. Where someone remains opposed, understand their authority, incentives and effect on the workflow before deciding how to proceed. A survey score alone does not establish that a person should be managed around.
The undecided middle often watches credible peers before changing practice. Give this group bounded tasks, time to learn, visible support and evidence from work they recognise. Make appropriate use easier and safer without turning social pressure into a substitute for informed consent or professional judgement.
One or two vocal people should not set the pace by default. Equally, seniority or enthusiasm should not silence a risk that the project has failed to answer. Use agreed decision rights and evidence to resolve the issue.
This isn't a one-time exercise
Culture changes as people accumulate experience. Trust may improve after a well-controlled pilot, or fall after a poor rollout. Leadership behaviour can change when commercial evidence becomes visible.
Repeat the measurement at points tied to decisions: before the pilot, after people have enough experience to judge it, and before expanding the use case. The right interval depends on the programme's pace and workforce. Compare like with like, record material events between rounds and look for changes by team or role rather than relying on the firm-wide average.
Repeated measurement can also signal that the firm takes people's experience seriously. That benefit disappears if feedback produces no visible response. Publish what the firm heard, what will change, what will remain and why.
What leadership may not want to hear
Measuring cultural attitudes toward AI will, in some firms, surface things leadership would rather not know. You might discover that your senior team is the primary source of resistance. You might find that trust in technology is so low that you need to fix your existing systems before anyone will take AI seriously. You might learn that a significant portion of your firm views the whole initiative as a management fad that will pass if they wait long enough.
None of that is a reason not to measure. All of it is a reason you need to.
The firms I see making operational progress treat cultural readiness as seriously as technical readiness. They examine attitudes and behaviour before broad rollout, adapt the plan to the evidence, and keep investing in the human side after the technology is live.
An AI initiative can underperform because the technology, workflow, data, controls or adoption conditions are weak. Cultural measurement helps diagnose one part of that system. It should lead to a specific intervention rather than becoming another readiness score filed away after the launch meeting.
If you want a broader view of where your firm sits on AI readiness - covering data, infrastructure, and governance alongside culture - we've put together an AI readiness guide that covers all four dimensions. And if you'd like a downloadable version of the cultural readiness assessment framework described here, something you can adapt and run internally, get in touch and we'll share the template.
