Why 61% of Professional Services Firms Abandon AI — And What the Other 39% Did Differently

The Data Most Firms Don't Want to See

General Assembly surveyed professional services firms in March 2026 and found that 61% had abandoned at least one AI initiative in the past year. Thirty-five percent had abandoned multiple. These aren't firms that dabbled once and walked away — many of them tried more than once, failed more than once, and are now in the early stages of a third attempt.

The study asked firms why. The answer wasn't technical. It was organizational.

Stat	Source
61% abandoned at least one AI initiative	General Assembly, March 2026
35% abandoned multiple initiatives	General Assembly, March 2026
58% cite change management as #1 skills gap	General Assembly, March 2026
55% say partners/senior leaders are least prepared	General Assembly, March 2026
74% of legal firms say leadership is the bottleneck	General Assembly, March 2026
Only 26% feel "great" about selling AI capabilities to clients	General Assembly, March 2026
94% believe they're ahead of clients on AI	General Assembly, March 2026
42% have clients questioning their billing model due to AI	General Assembly, March 2026
37% are proactively renegotiating client pricing	General Assembly, March 2026

Fifty-eight percent of firms cited change management and stakeholder communication as their number one skills gap — not prompt engineering, not tool selection, not data security. The thing they were worst at was getting their own people to actually change how they worked.

The most common explanation for why firms struggle with AI isn't "we picked the wrong technology." It's "we didn't know how to lead our team through a process change."

That's a different problem. And it has a different solution.

---

The Bottleneck Is Upstairs

Here's the statistic that should make every firm owner pause: 55% of professional services firms say their partners and senior leaders are the least prepared people in the firm to use AI tools.

Not the associates. Not the junior staff. The senior people.

In legal, the number is sharper: 74% of legal firm respondents specifically identified firm leadership — partners, principals, managing attorneys — as the primary bottleneck in AI adoption. The people with the most authority over how the firm operates are the people least likely to personally use the tools.

This creates a predictable failure pattern. A firm buys an AI tool. The owner or senior partner sees the business case — time savings, cost reduction, competitive positioning. They direct staff to adopt it. Staff tries. Staff encounters friction. Without visible, credible leadership engagement, the change dies in the middle of the org chart. Two months later, the tool is barely being used. Six months later, someone cancels the subscription.

There's a second number that tells the same story from a different angle. Only 26% of firms feel confident selling their AI capabilities to clients. Yet 94% believe they're ahead of clients on AI adoption. Almost every firm believes they're ahead — but barely a quarter can demonstrate it when a client asks.

That gap between self-reported readiness and the ability to prove it isn't a marketing problem. It's a depth problem. The firms that can't demonstrate their AI capabilities aren't hiding a genuine capability — they simply don't have one yet. The capability never got built because it never got used.

The firms in the 39% — the ones who didn't abandon — have one thing in common that the data consistently points to: the owner or senior partner is a personal, hands-on user of the tools. Not a sponsor. A practitioner.

---

Adding Tools vs. Redesigning Workflows: The Critical Difference

There's a distinction that separates firms that get results from firms that don't, and it's not about which tools they chose.

Adding a tool looks like this: you subscribe to an AI writing assistant. When someone on your team needs to write a proposal, they can use it. The underlying process — how proposals get assigned, drafted, reviewed, and sent — hasn't changed. AI is available as an option within that process. Most people use it when they remember to. Some don't use it at all.

Redesigning a workflow looks like this: you pick one specific, high-volume process in your firm (say, client meeting follow-up) and rebuild it from the ground up with AI as a defined step. Every client meeting gets recorded. The transcript goes through an AI summary tool. The summary gets reviewed before billing is entered. A client-facing recap goes out within two hours. No one has the option to skip the step — it's part of how the work gets done now.

The performance difference between these two approaches is significant. In the staffing sector, firms that redesigned their candidate placement workflows — rather than just adding AI tools on top of existing processes — saw 142% more job orders fulfilled per recruiter. The firms that added tools without redesigning the underlying workflow saw minimal ROI and abandoned their AI investments at a higher rate.

This is the mechanism behind the 61% abandonment figure. Most of those firms tried to add AI. Very few tried to redesign. The ones that redesigned are the ones still using it.

The reason redesign is harder is obvious: it requires the people at the top of the firm to commit to a specific new way of working before they can ask anyone else to. You can't redesign a client meeting follow-up workflow if the senior partner still does their post-meeting debrief the old way. The workflow redesign only works if the person with authority over the workflow actually changes how they work.

Which brings us back to the bottleneck being upstairs.

---

The 3-Step Sequence That Avoids the Change Management Failure

The firms that successfully implemented AI and stayed with it didn't follow a big rollout plan. They followed a sequence. And the sequence matters.

Step 1: The owner learns the tool personally — before asking anyone else to.

Not a demo. Not a vendor walkthrough. Personal, hands-on use in actual work. Pick one task — meeting notes, client email drafts, research synthesis — and spend one week doing it with AI instead of without. The goal isn't mastery. It's fluency. You need to know what good output looks like, where the tool requires intervention, and what the workflow feels like from the inside. You can't credibly lead an AI change initiative if you haven't experienced the change yourself.

This step also does something important: it gives you stories. When you tell your team "I've been using this for three weeks to prep for client calls and it's cut my prep time in half," that's a different kind of organizational communication than "here's a tool I want everyone to try."

Step 2: One workflow gets redesigned from scratch.

Not the whole firm. One workflow. Pick the highest-volume, most repetitive process in the firm — the one that consumes the most cumulative hours across your team. Map how it currently works, step by step. Then map how it would work with AI handling a defined step. The redesign doesn't need to be complex. It needs to be specific. "AI does X, human reviews Y, output Z goes to the client" is a workflow. "We use AI to help with drafts sometimes" is not.

This is where change management actually happens — in the specificity of the new process, not in the vision for AI transformation. People can follow a specific process. They can't follow a general direction to "be more AI-enabled."

Step 3: Train staff on the new workflow — not on the tool.

This is where most firms make their second mistake. They buy a tool, they run a training session on how to use the tool, and then they expect people to figure out when and how to apply it. That's the wrong curriculum.

The training that works is: "Here is the new process. Here is your role in it. Here is exactly what you do at each step." The tool is incidental. The workflow is the thing they need to learn. When staff know the specific new steps they're responsible for, adoption happens because it's just part of how the work gets done now — not because people are motivated to use AI.

Sequence matters here too. The owner goes first. One workflow gets redesigned. Staff gets trained on the workflow. The firms that try to skip the first step — that ask staff to adopt a tool the owner hasn't personally used — face the change management failure the 61% survey is documenting.

---

What "Implementation Success" Actually Looks Like at a 5-Person Firm

Here's a concrete before-and-after. A five-person accounting firm handled client meeting notes the same way they had for years: the partner took handwritten notes during the meeting, typed them up later, used them to prepare billing entries, and sent a brief follow-up email to the client. Average time from meeting to follow-up: 3–4 hours of partner time.

After a workflow redesign:

The partner runs a Fathom recording link in every virtual client meeting. After the meeting, Fathom produces a timestamped transcript and an AI-generated summary with action items. The partner reviews the summary (5–10 minutes, not 30–45). Billing entries are populated from the action item list. The client gets a clean, professional recap email the same afternoon — adapted from the summary, not written from scratch.

Time from meeting to follow-up: under 30 minutes. Client experience: visibly better. Partner capacity freed: roughly 3 hours per client meeting.

That firm didn't implement a comprehensive AI strategy. They redesigned one workflow. The owner learned the tool first, redesigned the specific steps, and trained the one associate who supported meeting prep. Three weeks to implement. Results visible immediately.

This is what the 39% did. Not a big AI transformation. One workflow. Owned personally by the senior person. Then another workflow. Then another.

Firms that are still in the 61% are usually waiting for a more comprehensive plan, a better tool, a clearer ROI case, or a time when the team is ready. Those conditions don't arrive. The sequence above creates them.

---

Frequently Asked Questions

What percentage of professional services firms abandoned AI initiatives in 2026?

61% of professional services firms abandoned at least one AI initiative in the past year, according to a General Assembly survey conducted in March 2026. 35% abandoned multiple. The most common reason wasn't a bad tool choice or a technical failure — it was an internal skills gap. Specifically, 58% of firms cited change management and stakeholder communication as the #1 thing they lack. In other words, most firms failed at the human side of implementation, not the technical side.

Why do professional services firms fail at AI implementation?

The dominant failure mode is a leadership readiness gap, not a technology problem. General Assembly's March 2026 survey found that 55% of firms say their partners and senior leaders are the least prepared people in the firm to use AI tools. In legal specifically, 74% of respondents identified firm leadership as the primary bottleneck. Only 26% of firms feel confident demonstrating their AI capabilities to clients. The firms that succeed at AI implementation share one common trait: the owner or senior partner is a personal, hands-on user of the tools — not just a sponsor who delegates adoption to staff.

What is the difference between adding AI tools and redesigning AI workflows?

Adding a tool means using AI within an existing process — for example, asking ChatGPT to draft one email instead of writing it manually. Redesigning a workflow means restructuring how work gets done so AI handles a defined step systematically — for example, all client meetings are recorded, the transcript is reviewed before billing is entered, and a client recap goes out within two hours. The first requires no process change. The second requires intentional redesign. Firms that redesigned workflows saw 142% more output. Firms that added tools without redesigning the underlying process saw minimal ROI and a high abandonment rate.

How should a professional services firm owner start with AI if previous attempts have failed?

Three-step sequence. First, the owner learns the tool personally — not as a sponsor, but as a user. Pick one task (meeting notes, email drafting, research synthesis) and spend one week doing it with AI instead of without. Second, one workflow gets redesigned from scratch with AI embedded as a defined step. Pick the highest-volume, most repetitive workflow in the firm. Map how it currently works, then map how it works with AI. Third, train staff on the redesigned workflow — not on "how to use AI tools." People resist abstract tool adoption. They adopt specific new processes. The sequence works because it puts the owner's credibility behind the change before asking anyone else to make it.

Are professional services firms actually ahead of clients on AI adoption?

94% of professional services firms believe they're ahead of their clients on AI adoption, according to General Assembly's March 2026 survey. But only 26% feel confident demonstrating those capabilities to clients when asked. That gap — between self-reported readiness and the ability to prove it — is the most revealing number in the data. A firm that is genuinely ahead on AI should be able to show a client, clearly and specifically, how AI makes their work better, faster, or more accurate. Most firms can't do that yet. The 39% that succeed at AI implementation can.

---

What to Do This Week

If you recognize your firm in the 61% — if you've tried AI, seen limited results, and aren't sure what went wrong — the answer isn't to try harder or buy a different tool.

The answer is to go back to Step 1 and actually do it. Pick one task you do personally, every week. Spend the next five working days doing that task with an AI tool instead of without. Not delegating to an assistant. Not piloting with your team. You, personally, using the tool for a real work task.

After five days, you'll know whether you were the bottleneck. Most firm owners who do this find out that yes — they were. And that's useful information, because it's fixable.

The broader AI adoption gap in professional services is well-documented at this point. The question isn't whether to adapt — it's how. The workflow redesign approach is where the ROI actually shows up. And once you've redesigned a workflow, measuring the results is how you build the internal case for the next one.

---

The Crossing Report is a weekly intelligence briefing for professional services firm owners navigating the AI transition. Every Monday, we cover what's actually working — specific tools, specific workflows, specific results — for firms like yours. Subscribe here for the practical AI adoption framework, delivered to your inbox.