Thanks to my friends at Laurel for sharing this article by Diana Buccella.
The firms that will look back on their early AI investments and wonder what went wrong won't have chosen the wrong tools. They'll have deployed them without knowing which work deserved to be automated and which deserved to be protected.
Here’s an uncomfortable truth about the AI investments being made across professional services right now: most of them are solving the wrong problem.
Not because the tools aren’t capable. Not because the strategy is misguided. But because the work being accelerated is only a fraction of the work actually being done. And in many cases, it’s the less valuable fraction.
Speed is only useful if you’re going the right direction
When a firm deploys AI to draft documents, scan contracts, or summarize research, something real happens: that work gets done faster. That’s a genuine win. But it’s a win that only touches the work that was already visible, the activities that made it into time entries, project logs, and system records.
The work that didn’t get captured, the research that happened before the formal engagement started, the judgment call a partner made in a hallway conversation, the hours absorbed managing a client crisis, that work still doesn’t exist, as far as your systems are concerned. AI can’t accelerate what it can’t see. And the activities it can’t see are often the ones your clients are actually paying for.
The result is a firm that’s getting faster at the visible work while the invisible work stays unmeasured, unprotected, and unpriced.
The cost of optimizing in the dark
Consider what happens when a firm decides to automate a category of work without knowing how that work connects to profitability.
A practice group that looks efficient on paper, high utilization, fast turnaround, may be quietly absorbing hours of invisible work that never surface in any report. When AI speeds up the visible portion, the firm celebrates a productivity gain while the invisible hours continue to erode margins in the background. The problem doesn’t get solved. It gets obscured.
This plays out in pricing too. Fixed-fee arrangements are only defensible if you understand your true cost of delivery. But if a meaningful share of the work that went into an engagement was never recorded, your cost model is based on a partial picture. You’re not setting a price. You’re making a guess and hoping it holds.
The firms that will look back on their early AI investments and wonder what went wrong won’t have failed because they chose the wrong tools. They’ll have failed because they deployed those tools without knowing which work deserved to be automated and which deserved to be protected.
What it means to actually know your work
Distinguishing between work that should be automated and work that should be protected requires a complete picture of what your professionals are actually doing, not what they report, not what their time entries suggest, but the discrete activities that make up the working day.
That means capturing work as it happens across every surface: calls, laptops, virtual environments, collaboration tools. It means classifying those activities against a definition of work that reflects how your firm operates. And it means connecting that data to business outcomes so you can see, with precision, where human judgment drives the most value and where automation is genuinely the right call.
Without that foundation, every AI investment is a bet. The technology may deliver exactly what it promises and still point the firm in the wrong direction.
The question driving better decisions
The firms getting this right aren’t asking “how do we deploy AI faster?” They’re asking a harder question first: which of our work should exist at all, which should be automated, and which is too valuable to hand off?
That question can’t be answered with submissions and reported time. It requires intelligence about work itself, the kind that most firms are only beginning to build.
The good news is that building it creates compounding returns. Every activity captured is a data point. Every data point classified is a clearer picture of where your firm’s value actually lives. And every decision made with that picture is more defensible than the one made without it.
Faster is only better when you know what you’re accelerating.
