Ninety days. That's thirteen weeks. One quarter. The time it takes most professional services firms to agree the agenda for their AI steering committee - if they've even formed one yet.
I want to walk you through what happened when a mid-market professional services firm decided to stop talking about AI and actually do something with it. Not in eighteen months. Not after hiring a Head of AI. Not after a data transformation programme that would take until 2026 to complete. In ninety days, starting from a position that probably looks a lot like yours.
But I need to be honest about something first. This isn't a story about a firm that had everything figured out and just needed a push. It's a story about a firm that was genuinely uncertain, had real gaps in its foundations, hit a specific problem halfway through that required us to change course, and still ended up with a working AI application in production by week thirteen. The honest version is more useful to you than the polished one.
[INTERVIEW REQUIRED: James to confirm the firm's sector, agreed anonymisation approach, and level of detail that can be published about the use case and results. All specifics below are placeholder structures that must be replaced with real case material.]
[INTERVIEW REQUIRED: James to describe the firm's actual starting point in detail.]
The firm was a [SECTOR] practice, roughly [SIZE] people, with a reputation built over [YEARS] in their market. AI was on the agenda - a few partners were using ChatGPT for personal productivity, drafting emails, summarising documents, that sort of thing. One associate had built a GPT wrapper for [SPECIFIC TASK] that a handful of people used informally.
But there was no formal programme. No data assessment. No governance framework. No one had looked at whether the firm's data was actually in a state where AI could do anything meaningful with it. The managing partner was genuinely curious - not a reluctant convert being dragged into it by a technology team - but had no clear direction on where to start.
We've been meaning to do something with AI, but we're not sure where to begin. And honestly, every AI project I've seen has taken longer than expected and delivered less than promised.
That was almost exactly what [ROLE] said in our first conversation. And they weren't wrong to be sceptical. Most AI projects in professional services do take longer and deliver less. But that's usually because they start too broadly, skip the foundations work, or try to boil the ocean with a firm-wide AI strategy before anyone's proved that AI can solve a single specific problem.
We took a different approach.
I've written about how we approach AI readiness in more detail elsewhere, but here's the specific timeline for this engagement.
Weeks one and two: finding out what was actually true
We ran a structured readiness assessment across five dimensions - data, systems, culture, governance, and skills. This isn't a box-ticking exercise. It's designed to produce a specific profile of where the firm actually is, not where it thinks it is or where it wants to be.
[INTERVIEW REQUIRED: James to describe what the assessment found - specifically what was ready and what wasn't. The honest gaps are the most valuable part of this section.]
The assessment surfaced [SPECIFIC FINDING ABOUT DATA READINESS] and [SPECIFIC FINDING ABOUT GOVERNANCE]. It also identified [NUMBER] potential use cases aligned with the firm's current state. Some were more ambitious than the foundations could support. That's normal. And it's better to discover that in week two than in week eight.
The three weeks we spent on assessment and prioritisation were what made the remaining ten weeks viable. I know that sounds counterintuitive - spending nearly a quarter of your timeline on preparation rather than building. But every AI engagement where I've seen things go sideways, the root cause was skipping this step. You can't build on foundations you haven't examined.
Weeks three and four: choosing one thing, not everything
This is where most firms get it wrong. The temptation - and I completely understand it - is to look at a dozen potential AI use cases and try to launch a programme that addresses several at once. More bang for your buck, right? Except it's not. It's more risk for your buck, more complexity for your buck, and more likelihood that nothing gets finished.
We worked with the firm's leadership to identify one use case that met four criteria: it addressed a specific workflow problem, it had measurable success criteria, the data foundation was adequate (not perfect - adequate), and a senior partner was willing to sponsor it personally, not just approve it and check back in three months.
[INTERVIEW REQUIRED: James to describe the specific use case selected and why it was chosen over other candidates.]
The use case they chose was [SPECIFIC USE CASE]. The decision to focus on this single use case rather than launching a broader programme was the most important decision in the entire ninety days. I'd go further - it's the single most important decision any firm makes in its first AI engagement.
Weeks five to ten: building the pilot
The pilot was designed with specific scope, pre-agreed success criteria, realistic data, and human oversight built in from day one. We used existing tools rather than procuring new infrastructure - [INTERVIEW REQUIRED: James to specify which tools and platforms were used].
Now, here's where I need to be honest about something that didn't go smoothly.
[INTERVIEW REQUIRED: James to describe the specific problem encountered during the build phase - what went wrong, why it happened, and how it was resolved. This is the most commercially important paragraph in the piece for credibility. A managing partner reading this needs to see that problems arose and were handled, not that everything went perfectly.]
[PLACEHOLDER: During week [X], we discovered that [SPECIFIC PROBLEM]. This meant [SPECIFIC CONSEQUENCE]. We adapted by [SPECIFIC RESOLUTION], which added [TIME IMPACT] but ultimately [OUTCOME]. If I'm being candid, we should have [WHAT COULD HAVE BEEN DONE DIFFERENTLY]. That's a lesson we've carried into subsequent engagements.]
The build phase involved the firm's existing staff working alongside our team from day one. This wasn't a case of us disappearing into a back room and emerging twelve weeks later with a finished product. The firm's people were in the work throughout - partly because that's how we operate at Distinction, and partly because if the firm's own people don't understand how the thing works, it won't survive past the pilot.
Weeks eleven to thirteen: measuring what actually happened
We measured the pilot against the success criteria agreed back in week three. Not against a broader set of aspirations that had crept in during the build. Not against what we hoped it might do. Against what we said it would do, twelve weeks earlier.
[INTERVIEW REQUIRED: James to provide specific, measurable results expressed in terms a managing partner will recognise - time saved, adoption rate, quality improvement, cost reduction. State the methodology alongside the figure.]
The results: [SPECIFIC RESULT]. [SPECIFIC RESULT]. [SPECIFIC RESULT].
What didn't improve, or was harder than expected: [INTERVIEW REQUIRED: James to describe what fell short of expectations or what remained difficult. This is essential for credibility.]
I should be clear - these results came from a contained pilot with a specific team and a specific use case. Extrapolating from a pilot to a firm-wide projection is exactly the kind of AI hype that gives managing partners justified scepticism. What the results did demonstrate was that the approach worked for this use case, with this team, in this timeframe.
I want to be careful here. Ninety days is not a guarantee. It was the result of four specific conditions. If your firm has these, you can replicate something close to this timeline. If you don't, you need to address the gaps first.
The first is clear scope. One use case. One workflow. One team. The managing partner who tries to solve six problems simultaneously in their first AI engagement will solve none of them. I've seen it happen enough times that I'm fairly dogmatic about this.
The second is willing leadership. [ROLE] wasn't a sponsor who signed off and reappeared at the final presentation. They were present at every fortnightly review. They made decisions when decisions were needed. They shielded the pilot from the inevitable "can we also add..." requests that accumulate in any project involving new technology. This isn't about a leader who's technically literate. It's about a leader who's committed enough to protect the scope.
The third is honesty about foundations. The readiness assessment in weeks one and two identified [SPECIFIC DATA QUALITY ISSUE] that was addressed before the pilot build began. If we'd skipped the assessment and started building in week three, we'd have discovered that problem in week seven or eight - and the project would have stalled while we went back to fix it. The firms that move fastest are the ones willing to hear difficult things about their data and governance early.
The fourth is focused measurement. The success criteria defined in week three were the reference point for every decision in weeks five through thirteen. The pilot was never asked to do more than it was designed to do. When someone suggested adding [SPECIFIC FEATURE CREEP EXAMPLE], the answer was straightforward: that's not what we're measuring in this pilot. Put it on the list for phase two.
None of these conditions is exceptional. But all four need to be present. If you're reading this thinking "we have those" - you're probably closer to ninety days than you think. If you're reading it thinking "we definitely don't have condition three" - that's actually useful information, because the readiness assessment is exactly where you'd start.
[INTERVIEW REQUIRED: James to describe the firm's post-engagement trajectory - what was deployed next, how the internal capability changed, whether the assessment and prioritisation practices persisted.]
The ninety-day engagement didn't just produce [SPECIFIC AI APPLICATION]. It produced the organisational habits - the governance framework, the prioritisation discipline, the measurement practice - that make subsequent AI initiatives faster and more reliable. The second and third use cases [INTERVIEW REQUIRED: confirm whether these were commissioned and what they were] were scoped and initiated within [TIMEFRAME] of the pilot completing, and they moved faster because the foundations were already in place.
[INTERVIEW REQUIRED: specific quote or observation about how the firm's relationship with AI changed as a result of the engagement.]
That, honestly, is the bit I'm most proud of. Not the pilot results - though they were good. The fact that the firm now has the internal confidence and the practical framework to keep going without us standing next to them. That's what "done with you" actually means in practice.
I'm not going to pretend that every firm can replicate this exact timeline. Your starting point is different. Your data is different. Your culture is different. But the pattern - assess honestly, choose one thing, build with discipline, measure against what you said you'd measure - that pattern is replicable.
If you want to understand whether the conditions that made ninety days possible are present in your firm, book an AI discovery conversation. It's a conversation, not a pitch.
And if you want to start where this firm started - with a structured assessment of where you actually are, not where you think you are - book a readiness assessment. The assessment that opened this engagement took two weeks and cost a fraction of what the firm had been quoted for a full AI strategy. It also stopped them spending money on the wrong things, which - if you've read our piece on how to build confidence in AI without overpromising - you'll know is half the battle.
We've also put together a one-page overview of the 90-day implementation roadmap - the four phases, the typical timeline, the client involvement required, and the deliverables at each stage. If you want a clear picture of what the engagement actually looks like in practice, download the 90-day AI implementation roadmap. It's written for managing partners, not technologists.