Short answer: they’re a sequence, not a choice.
Synthetic research is for breadth, speed, and the questions you can’t put to a real client. Real conversations are for depth, emotional truth, and the decisions that carry real money. The firms that get value from synthetic research don’t use it to replace client research — they use it to make their real research sharper and cheaper. Run synthetic first to find the right questions and narrow the field, then spend your limited real-client time settling what actually matters.
This is also, for once, the settled view. The responsible providers across the whole category say the same thing: synthetic is a supplement, strong for directional work, wrong for high-stakes final decisions. A vendor who tells you it replaces talking to your clients is not describing the technology honestly.
What each is actually good at
Synthetic research is strong on breadth and candour. It can explore hundreds of client types in the time a single recruited study would take, at a fraction of the cost. It reaches clients you could never assemble — a rival firm’s general counsel, the prospects who quietly chose someone else, the senior clients who’d never sit for an interview. And it lets you ask confrontational questions — “why would you drop this firm?” — that would be relationally reckless to put to a real client, because there’s no relationship to damage.
It also finds things interviews miss. In a peer-reviewed study in the Journal of Marketing (Arora, Chakraborty and Nishimura, 2025), researchers reran a Fortune 500 company’s study with AI working alongside human analysts. Depending on the set-up, the combined approach recovered 77% to 96% of the themes the human-only study found, and surfaced new themes it had missed (14% to 23%).
Real research is strong on depth and truth. A real client surprises you. They hesitate, contradict themselves, say the thing they didn’t plan to say, and react in the moment in ways no model reproduces. For anything emotional, relational, or genuinely high-stakes, the real conversation is irreplaceable — and it is the only one that also maintains the relationship while it gathers the intelligence.
The mistake is asking which is better. They’re good at different things, and a firm needs both.
Where synthetic research earns its place
Reach for synthetic first when you’re:
- exploring a market before you commit — mapping how a category of client thinks, and which considerations drive them, before you’ve spent a penny on recruitment;
- testing messages and positioning — running a claim past dozens of client types to see what persuades and what bounces, so the version you take to real clients is already the strong one;
- reaching the unreachable — the clients you can’t recruit, the lost pitches you can’t call back, the competitor’s clients;
- pressure-testing a pitch before it’s in front of a real client (done properly — the naive version flatters you);
- asking the questions you can’t ask — the confrontational, decomposing lines of enquiry that would make a real client feel cross-examined.
In each case the point is the same: synthetic research front-loads the thinking and widens the field, cheaply, before you commit scarce real-client access.
Where you still need real people
Use real clients — not synthetic — when the question is:
- a genuine go/no-go decision where being wrong is expensive;
- about price — what clients will actually pay is where synthetic data is least reliable;
- emotionally or relationally sensitive, where the truth lives in tone and hesitation, not words;
- the kind of thing that, asked well of a real client, also deepens the relationship — which synthetic research, by definition, never can.
Synthetic research reads how a market thinks. It doesn’t predict what a specific client will do, and it can’t strengthen a real relationship. For those, there is no substitute for the conversation.
The sequence that works
The firms getting the most from both run them in order, not in competition.
For a firm, synthetic first usually means a study across a few hundred simulated clients, run by a research consultancy such as Taleist, or on a self-serve platform if you have research skills in-house.
Synthetic first: generate the hypotheses, map the client types, find the messages worth testing and the questions worth asking. Real second: a smaller, sharper set of client conversations that confirm what the synthetic work surfaced — and only that. The synthetic stage means your real research is no longer a broad, expensive fishing trip; it’s a focused, confident test of a few things that matter. You spend less on real research and trust the result more.
Synthetic first, real second
Run it the other way — or skip the real stage entirely — and you lose the benefit of both.
The mistake to avoid
The real danger isn’t using synthetic research. It’s using it to avoid the discomfort of real client feedback. Synthetic personas are seductive precisely because they never contradict you, never deliver the inconvenient finding, never make you justify a hard truth to a partner. That’s not a feature — it’s the trap. A firm that leans entirely on synthetic research ends up with a confident, comfortable echo of what it already believed, and none of the friction that changes a mind.
For a professional services firm the risk has a particular edge: your own records and your partners’ instincts already over-index on why you win and go quiet on why you lose. Synthetic research built to flatter that story just launders the blind spot. Used honestly — to widen the field and sharpen the real questions — it does the opposite.
For the questions that separate a synthetic research provider who understands this from one who oversells, see the companion guide, how professional services firms should evaluate synthetic research.
For synthetic research on a question you can’t ask a client outright, see why corporate clients stay with their incumbent law firm.
Taleist builds decision intelligence for professional services firms — designed to widen the field and sharpen the questions your real client conversations then settle.