Fifty-one percent of companies already have AI agents in production. I keep coming back to that number because of what it implies about the other forty-nine percent. They're not all sitting around ignoring AI - most of them are stuck somewhere between "we should probably do something" and "we tried something and it didn't really land." And when I dig into why, the answer is almost never that the technology wasn't available. It's that the firm wasn't ready in ways they hadn't thought to check.
I was working with a 200-person consulting firm earlier this year - sharp leadership team, decent tech stack, genuine enthusiasm for AI. They'd invested in a document summarisation tool, rolled it out to their consulting teams, and within six weeks it was gathering dust. Nobody was using it. The tool worked fine. The data it was pulling from was a mess. Half the documents it was summarising were outdated versions sitting in a SharePoint folder that hadn't been cleaned since 2019. The consultants tried it twice, got unreliable outputs, and went back to doing things manually.
That firm didn't have a technology problem. They had a data readiness problem masquerading as a technology failure. And because nobody had diagnosed which dimension of readiness was actually the constraint, they burned budget, lost internal credibility for AI initiatives, and the next time someone suggested an AI project, the room went quiet.
We don't need a checklist. We either have the capability or we don't, and we'll know when we try.
I hear this a lot. And I understand the instinct - it sounds pragmatic, action-oriented. But it's a bit like saying "we don't need a survey, we'll know if the building is structurally sound when we start the renovation." You might. Or you might get three weeks in and discover the foundations can't support what you're trying to build. By then you've spent the money, lost the momentum, and - critically - damaged the internal appetite for trying again.
AI readiness isn't binary. It's a profile. And the reason that matters is because different AI use cases make demands on different parts of your organisation. Automated proposal generation needs strong data and systems readiness but can work even if your cultural readiness is middling. A client-facing AI chatbot needs governance nailed down tight but can tolerate weaker technical skills internally. Knowing your profile doesn't just tell you whether you're ready - it tells you what you're ready for.
That's the thinking behind the framework I'm going to walk through here. Five dimensions, each scored independently, each mapping to a different set of use cases. It's designed to take twenty to thirty minutes, not an afternoon. And the output isn't a pass or fail - it's a map.
If you want the conceptual grounding for what AI adoption actually involves in a mid-market firm, I've written about that separately in a piece on what using AI in your business actually means. But if you're past the "should we?" stage and into the "where do we start?" stage, this is where to focus.
Before I walk through each one, a quick note on how to use this. Score each dimension on a 1–5 scale. Don't average them into a single number - that's where the useful information gets lost. A firm scoring 4 on data readiness and 2 on cultural readiness has a fundamentally different starting point than a firm scoring 3 on both. The profile is the point.
And be honest. Score where you are today, not where you plan to be after the project you haven't started yet.
This is the one that trips up most firms. Not because they don't have data - mid-market service firms are usually swimming in it - but because the data isn't in a state that AI can actually use.
There are four things to assess here.
Quality. Are the data sets you'd want AI to work with clean, consistently formatted, and free of significant duplicates or errors? A score of 1 looks like: contact records with three different date formats, duplicate entries everywhere, fields that say "TBC" or "ask Sarah." A score of 3: mostly clean, some known issues that get fixed periodically, the odd rogue spreadsheet floating around. A score of 5: actively maintained, validated at the point of entry, with a defined process for catching and correcting errors.
Accessibility. Can the relevant data be accessed programmatically - meaning an AI system can pull from it via an API or direct connection - or does someone need to export a CSV and email it across? A 1 means everything lives in individual spreadsheets, email inboxes, or people's heads. A 3 means most key data is in a central system but extraction still requires manual steps. A 5 means systems expose data through APIs or structured queries, and an AI tool could connect to them without someone manually bridging the gap.
Integration. Is your data consolidated, or scattered across disconnected applications? I worked with a law firm last year that had client information in their practice management system, billing data in a separate finance platform, marketing engagement in HubSpot, and matter-specific documents in a mix of iManage and shared drives. Each system was fine on its own. But asking an AI tool to do anything useful across that landscape was like asking someone to write a report using four different libraries in four different buildings. A score of 1 means heavily siloed with no connections between systems. A 3 means some integration exists - maybe CRM talks to the website, or finance connects to the practice management system - but significant gaps remain. A 5 means a well-integrated stack where data flows between core systems with minimal manual intervention.
Governance. Is there clear ownership of data quality? Does someone actually care whether the CRM is clean, and is there a process for keeping it that way? A 1 means nobody owns it and data hygiene happens sporadically, usually right before a board meeting. A 3 means there's informal ownership and periodic clean-ups. A 5 means there's a named person or team responsible, with documented standards and regular review cycles.
Poor data quality costs organisations an average of $15m annually, according to Gartner. For a mid-market firm, the absolute number is smaller, but the proportional impact is often worse because you've got fewer resources to absorb the inefficiency. I'll be honest - we've seen this catch out firms who thought they were in decent shape. They weren't wrong exactly, just optimistic.
This one is frequently stronger than firms expect, which is a nice surprise for once.
Integration capability. Do your core systems - CRM, practice management, finance, document management - have APIs or other integration capabilities that would allow AI tools to connect to them? Most modern platforms do. If you're running Salesforce, HubSpot, Microsoft 365, or Dynamics, you've already got integration points available. Score a 1 if your systems are largely closed, with no APIs and limited export capabilities. A 3 if your main platforms have APIs but you haven't used them extensively. A 5 if you're already running integrations between systems and have experience connecting third-party tools.
Platform modernity. Are your primary systems modern enough to support what AI applications need? This doesn't mean everything has to be brand new - it means the platforms can handle the connections, data flows, and processing that AI tools require. A 1 means you're running legacy systems that struggle with basic integration, let alone AI. A 3 means a mix - some modern platforms, some older ones that would need workarounds. A 5 means your core stack is current, cloud-hosted, and extensible.
Existing AI features. This is the one that surprises people most. What AI-adjacent capabilities do your current tools already include? Microsoft Copilot, Salesforce Einstein, HubSpot's AI features, even basic things like smart email categorisation or predictive text. If you're in the Microsoft ecosystem - and most mid-market professional services firms are - you're sitting on more AI capability than you probably realise. You might not need to buy anything new to start. A score of 1 means your tools have no AI features or you're unaware of what's available. A 3 means some AI features exist but haven't been activated or explored. A 5 means you're already using embedded AI features and have a sense of what they can and can't do.
Right. This is the awkward one. Because you can have perfect data, modern systems, and a clear governance framework, and still fail at AI adoption because your people aren't having it.
Leadership buy-in. Is there genuine senior sponsorship for AI, or has it been delegated to the IT team as a "technology project"? There's a world of difference. When the managing partner or CEO is actively championing AI - asking about it in meetings, allocating budget, making it part of the strategic conversation - it signals something different to the organisation than when it's sitting on the CTO's to-do list. A 1 means leadership hasn't engaged meaningfully with AI beyond reading a few articles. A 3 means there's interest and some budget allocation, but it's not a strategic priority. A 5 means AI is part of the firm's strategic plan with named senior sponsorship and protected budget.
Staff willingness. This is where professional services firms in particular run into trouble. Partnership structures create a specific dynamic: if the partners aren't using it, nobody else feels they have permission to. And partners, by nature, tend to be sceptical of anything that changes how they work until they've seen proof it works for someone exactly like them. A 1 means active resistance or widespread anxiety about AI. A 3 means a mix - some enthusiasts, some sceptics, most people waiting to see what happens. A 5 means a general expectation that AI tools will become part of the workflow, with early adopters already experimenting.
Change management capability. Has your firm successfully managed technology-driven behaviour change before? Rolling out a new CRM counts. Moving to Teams from email-based collaboration counts. If those went well, you've got muscle memory. If they went badly - and honestly, in a lot of firms they did - that's useful information too, because it tells you what not to repeat. A 1 means previous technology rollouts have gone poorly, with low adoption and lingering resentment. A 3 means mixed results - some changes stuck, others didn't. A 5 means the firm has a track record of managing technology transitions effectively, with internal champions and a tested approach.
This dimension is often the binding constraint for professional services firms specifically. The partnership governance model creates conditions that resist rapid adoption in ways that, say, a SaaS company or an MSP simply doesn't face. That's not a criticism - it's a structural reality that needs to be factored into any honest assessment. And if I'm being straight with you, it's the dimension where I've seen the most wishful scoring.
If cultural readiness is the awkward dimension, governance is the boring-but-essential one. And in regulated industries - financial services, legal - it can be the difference between a successful pilot and a compliance incident.
Policies. Does your firm have a written AI use policy? Not a 40-page document gathering dust, but something practical that tells people what's acceptable, what's not, and where to go with questions. A 1 means no policy exists and AI use is ad hoc - some people are using ChatGPT, nobody's sure if they should be. A 3 means there's informal guidance or a draft policy that hasn't been formally adopted. A 5 means a clear, communicated policy that's been reviewed by compliance or legal and is actively referenced.
Oversight structures. Is there a named person responsible for AI governance? In larger firms, this might be a committee. In a 150-person consultancy, it might just be the COO or CTO with an explicit remit. The key is whether someone is actually accountable. A 1 means nobody owns it. A 3 means someone's been informally tasked with keeping an eye on it. A 5 means there's clear, documented accountability with regular review.
Ethical framework. Has the firm thought through the client confidentiality, data handling, and professional obligation dimensions of AI use? For a law firm, this means: can we put client data into an AI tool? Under what conditions? What do our professional obligations say? For a financial services firm, it's: what are the regulatory implications of using AI in advice, reporting, or client communications? A 1 means these questions haven't been seriously discussed. A 3 means they've been raised but not resolved. A 5 means the firm has a considered position, documented and reviewed by relevant stakeholders.
If governance is your lowest-scoring dimension, I'd strongly recommend reading the companion piece on AI governance for professional services, which goes into this in much more depth. And if you're in a regulated industry and you score a 1 or 2 here, sort this before deploying anything client-facing. The reputational and regulatory risk isn't worth the speed.
Last one. And the temptation here is to score based on aspiration rather than reality. Don't.
Technical capability. Does the firm have the internal or externally accessible capability to configure, integrate, and maintain AI tools? This isn't about having a team of machine learning engineers - most mid-market firms don't and don't need to. It's about whether someone can set up an integration, configure a workflow, troubleshoot when something breaks, and understand what's happening under the bonnet enough to make good decisions. A 1 means no internal technical capability relevant to AI and no external partner. A 3 means some technical resource that could be redirected, or an existing technology partner who could extend into AI. A 5 means dedicated technical capability with AI experience, either in-house or through an established partner relationship.
AI literacy. Do the people who will actually use AI tools understand enough about how they work - and where they fail - to use them safely? This isn't about everyone becoming a prompt engineer. It's about understanding that large language models can produce confident-sounding nonsense, that outputs need checking, that the quality of what goes in determines the quality of what comes out. A 1 means most staff have no real understanding of AI beyond what they've read in the news. A 3 means some awareness, probably from personal use of tools like ChatGPT, but no formal understanding of limitations. A 5 means structured literacy across the firm, with people who understand both the capabilities and the failure modes.
Training infrastructure. Is there a mechanism for building AI capability across the firm, or would it require creating something from scratch? If you already run regular professional development, CPD programmes, or technology training, you've got a vehicle. If the last firm-wide training initiative was a compliance refresher three years ago, that's a different starting point. A 1 means no training infrastructure. A 3 means existing L&D capability that could be adapted. A 5 means an active training programme that's already incorporating or could readily incorporate AI content.
Most frameworks fall apart here - they tell you to "improve your weakest dimension" and leave it at that. That's about as useful as a doctor saying "get healthier." The whole point of scoring five dimensions independently is that the pattern tells you something specific.
If your strongest dimensions are data and systems - say, 4s and 5s there, but 2s on cultural or skills readiness - your best starting point is a back-office automation use case. Something like automated data extraction from documents, or report generation from structured data. These use cases don't require widespread behaviour change across the firm because they affect a small number of people doing a specific task. You get a win, build evidence, and use that evidence to shift the cultural needle.
If your strongest dimensions are cultural and skills - leadership is bought in, people are willing, but the data is messy and the systems are creaky - start with a communication assistance use case. Things like drafting support, email triage, or meeting summarisation. These tools work with unstructured data that people generate in real time, so they're less dependent on having a pristine data estate. They're also highly visible, which helps maintain the cultural momentum you've already got.
If governance is your lowest dimension - and especially if you're in a regulated industry - address that before deploying anything client-facing. Full stop. I've written about how to build the internal case for that preparatory work if you're finding it hard to get budget and attention for governance when everyone wants to jump straight to the shiny stuff.
If everything is sitting at 2 or 3 - you're in the messy middle, and honestly, that's where most firms are. Don't try to fix everything at once. Pick the dimension that unlocks the most valuable first use case, focus there, and let the first project create the momentum and learning for the second.
If you're scoring 4s and 5s across the board - you probably don't need this article. But you might want to read the piece on how to build confidence in AI without overpromising, which covers pilot design for firms that are ready to move.
I've laid out the thinking here so you understand what these dimensions are and why they matter. But self-assessment has its limits, and I say that as someone who's sat in enough rooms to know how it plays out. The CTO thinks the data is cleaner than it is. The managing partner thinks the culture is more receptive than it is. The compliance lead thinks governance is further along than it is. It's almost universal - and we've been guilty of it ourselves when assessing our own operations, so I'm not throwing stones.
That's why we built the AI Readiness Assessment as an online tool. It scores your firm across these five dimensions and returns a personalised readiness profile - takes about ten minutes, and you walk away with something concrete rather than a vague feeling. More usefully, it gives you something you can share with colleagues, which turns "I think we should look at AI" into "here's specifically where we're strong and where we're not, and here's what that means for what we should do first."
And if the results surface complexity that needs a proper conversation - or if you want to run it as a facilitated session with your leadership team - we do that too. We've run guided AI readiness workshops with firms from 80 people to 800. The conversation that comes out of a room where everyone has scored independently and then compared notes tends to be more useful than most strategy sessions I've been in. People say things they wouldn't put in a deck. Disagreements surface that were quietly blocking progress. It's worth doing for that alone.
The question isn't really "are we ready for AI?" It's "what are we ready for, and what do we need to sort out before we tackle the rest?" That's a much better question. And now you've got a way to answer it.