Human Judgment in the Age of AI
How AI is changing the role of the analyst
I. The First Draft
I was sitting at my desk, looking at a paragraph that had appeared in seconds. It looked finished, or close enough to seem that way. But it didn’t sound like me, and that was what stopped me. I’ve spent thirty years writing in different registers – stories, poems, academic pieces, the operational scraps of everyday analysis – yet here I was, hesitating over a line I hadn’t written, trying to decide whether to claim it. The model had done its part. The accountability stayed with me.
II. The Pace of Work
Lately I’ve noticed how the pace of the work has shifted. A draft returns before I’ve finished deciding what I want to say. Market notes arrive faster than I can decide what matters. A client’s question folds together claims, and each sounds certain on its own. None of it is wrong. It’s just fast. Faster than the patterns I learned to trust.
There was a time when writing moved at the speed of reading, back when the pace still matched the patterns I relied on. You gathered what you needed, compared it, let the argument settle. Now the work comes in fragments. Information is abundant. What’s scarce is the time to sit with it long enough to see what holds.
When I pause over these drafts, the hesitation isn’t about trust. It’s a quiet recalibration, a way of gauging what the work asks for now. The older sequence – gather, think, write – has given way to something closer to sorting: inputs that outrun attention, signals that collide before you’ve had a chance to weigh them.
None of this feels dramatic. It’s more like realizing the angle of the floor has changed just enough to shift your balance. The foundations that held for years no longer match the speed of what reaches us now. They weren’t built for this.
The work isn’t harder. It’s different.
Some days I’m struck by how little of that work is writing, the part I once thought defined the job. I delete more than I keep. I spend longer deciding what doesn’t belong than shaping what does. Most of the real work happens before anything reaches the page.
Clients and colleagues come with questions that don’t match the old rhythms. They aren’t looking for more material. They want to know where to begin, what to ignore, what carries weight. They’re trying to find the shape of the problem, and that work happens in conversation long before a draft exists.
Often the clearest thing I can offer is a narrower frame, placing one fact beside another, setting aside a claim that held until you followed it further in. Small shifts like that change the direction of the work. None of it looks like expertise as we used to describe it, but it decides what comes next.
I used to think analysis meant assembling the right information. Now it feels closer to sitting with tensions that won’t resolve. Several claims can make sense on their own and still fall short of what’s happening. Clarity comes from staying with the problem long enough, sometimes with someone across the table, to see where the line runs.
Somewhere along the way, the role changed. We used to write about markets. Now we help people understand them. Not through certainty or prediction, but through the steady work of interpreting what doesn’t line up. Most of this happens quietly: in margin notes, in brief exchanges after a call, in the kinds of conversations that never appear in a report but shape how people see their choices.
If the work has a center now, it’s this: finding what matters in the noise, and helping others see it too.
III. What Tools Can’t Do
I’ve learned not to confuse a quick draft with understanding. A model can give me a clean paragraph before I’ve finished settling on the point. It can arrange the pieces, smooth the edges, even sound confident about what belongs. But it doesn’t see the stakes. That part stays with me.
The tool handles the mechanical work. It can gather what I asked for, stack it neatly, offer a version that looks finished at a distance. Sometimes it produces a line close to something I might have written. But it treats everything as equal. It can’t tell what should fall away.
Its limits show up where the work won’t resolve. It leans toward coherence and fills gaps quickly when the ground is shifting. I can feel the boundary when it offers a conclusion too soon or misses an absence I’ve been circling.
That boundary isn’t a flaw. It’s the place where the work turns back toward judgment. Context, constraint, memory, responsibility: none of that can be retrieved on command. It comes from knowing what a client means but hasn’t said, or what a pattern suggests when you follow it past the obvious.
I noticed the shift during our analyst training this year. The final case studies came in cleaner and faster than anything I had seen from earlier cohorts. They had worked alongside AI, and the acceleration was visible on the page. Around the same time, I heard about a small consulting shop producing work at the pace of a full team. A small signal, but clear [1], [2].
The change runs through the field. Professional services now work at a quicker tempo, and the signs are easy to see. Reports land sooner. Proposals sharpen. Conversations move past steps that used to slow us down.
The acceleration has tightened expectations. Timelines that once felt ambitious now read as standard. Clients look for quicker cycles. Competitors release work on schedules that would have been unlikely not long ago.
The tools PSOs are building to support operations aren’t speculative. They answer to the way the profession already works.
The changes show up in small moments. A client will come with a question shaped by pressure, and I can tell the real problem sits somewhere else, a little further in. The work begins when I slow the conversation enough to find it. Sometimes it’s a single phrase they repeat without noticing. Sometimes it’s the pause after they describe the options they don’t want. Naming that center point gives the rest of the work a focus.
Later, back at my desk, I sort through what belongs in the model. A few details will anchor the draft; the wrong ones will turn it sideways. Choosing the right pieces is its own judgment call, and I feel the weight of it more than I used to.
The first drafts are plausible until I press on them. I follow each claim to where it rests, checking which ones collapse and which ones hold. The contradictions sit there waiting for me to decide what they mean.
Meanwhile, some of the work happens outside the draft. A colleague will ask how to structure a problem, and I’ll sketch a workflow or a prompt that helps them see the next step. These small tools don’t replace the analyst. They help others reach the point where the work can begin.
Ambiguity threads through the job. Clients bring partial information, competing pressures, and the things they’re afraid might be true. The model can summarize the situation, but it can’t stay with the tension long enough to see what follows. That part is still mine.
And then there’s judgment. Someone will ask whether a risk deserves urgency, and I try to answer without reaching for fear. Fear can force a decision, but it distorts it. Clarity gives people room to act.
If the role has a center now, it’s here, in the choices that shape the work before any of it reaches the page. The tools help. The responsibility stays with the analyst.
IV. The Work That Remains
When I look beyond my own desk, the pattern is plain. The old advantage of having access to information others couldn’t reach has thinned. Anyone can open a tool and get a quick answer. Knowledge no longer distinguishes one firm from another. The work has shifted toward interpretation.
Clients arrive with claims they’ve gathered on their own. They don’t need us to repeat any of it. They want to know how the pieces fit, or why sources that sound certain point in different directions. The value is in helping them sort through the noise, not in restating what they already have.
Inside firms, the same change holds. Teams use the tools to move quickly through the early steps — gathering, summarizing, drafting — and the bottleneck moves to interpretation. Producing material is easy. Making sense of contradictions isn’t. The tools can lay out the tradeoffs, but they can’t decide which ones matter.
Advisory work becomes something done with people rather than delivered to them. The analyst sits at the center of that process, connecting what the tools surface with what the moment requires. Sometimes that means clarifying the question. Sometimes it means slowing the room so the real concern can be said aloud. It’s less about authority now, and more about staying in the work with the people who need to choose a direction.
If the profession is heading anywhere, it’s toward this kind of shared sensemaking — work that unfolds between people, with the tools in the background. The pace has changed, but the choosing hasn’t.
The other day a colleague reached out with a question that had gathered more weight than it needed. The facts were straightforward; the situation around them wasn’t. We talked for a few minutes, setting aside what didn’t matter, holding on to what did. When the next step came into view, it wasn’t dramatic. It was a small turn, enough to steady things.
Everything around us moves quickly now, yet the work that lasts still happens at this scale: a single judgment, made quietly in the middle of an ordinary afternoon.
[1] https://journals.sagepub.com/doi/10.1177/20438869241280977