What is synthetic research?

If you're in law, consulting, accounting or financial services, you're probably hearing daily talk about AI-generated insights from large language models simulating buyer behaviour.

Using synthetic data to understand professional services clients

Synthetic research uses AI to create synthetic data that simulates how real clients think, decide and behave. When done properly, synthetic research is faster, cheaper and more flexible than traditional research. When done badly, it's expensive fiction.

Here's how to tell the difference between actionable synthetic data and a Copilot hallucination.

How synthetic research works

Synthetic research uses large language models (LLMs like ChatGPT and Claude) to create behavioural models that represent your clients. Instead of recruiting human respondents, you build Client Proxies that answer strategic questions about positioning, messaging, competitive perception and buyer behaviour.

The benefits of synthetic research

• No recruitment costs
• No scheduling delays
• No survey fatigue
• No clients annoyed about being bothered

You get the insights you need without the barriers that make traditional research prohibitively expensive in professional services.

What are the risks with synthetic data?

Synthetic research can generate plausible-sounding answers that have no connection to how real clients think. The difference between good synthetic data and garbage comes down to:

• What you ask
• How you structure the intelligence
• Whether you know what synthetic research can't answer

Why professional services firms need synthetic research

If you need to know "What percentage of our clients would renew?" or "How satisfied are clients with our service?" you should ask your clients.

For strategic questions about positioning, messaging, buyer behaviour and competitive perception, traditional research has frustrating limitations.

Access to clients is difficult

Partners don't want their clients "bothered". Senior decision-makers don't respond to surveys. When they do respond, they're either at the extremes (very happy or very dissatisfied) or they're not the actual decision-makers.

Low-effort responses waste your budget

When a supplier sends you a survey, how much time do you allocate? Multiply what you know is the reality over 500 respondents half-answering your questions.

Client Proxies give 100% effort, 100% of the time. Every answer is as carefully considered as the one before.

Social politics make people dishonest

You say the survey is anonymous. Your respondents aren't sure. Even if they trust the mechanics, they think their "voice" is distinctive. They believe you'll recognise them, so they censor their responses.

Client Proxies don't have careers to protect or relationships to manage. They give you the truth.

You can survey only your existing clients

You can generally survey only people where you have contact details or a relationship. That's your existing clients.

What about the rest of the market — the majority who haven't chosen to work with you yet?

If you're trying to understand why prospects don't choose you, you won't get the answer from people who did choose you.

Synthetic research lets you model the entire addressable market, not just the subset that's already bought from you.

People can't accurately report why they chose you

Memory is reconstructive. We rationalise decisions after the fact. Your clients genuinely believe their own explanations, but that doesn't make those explanations accurate predictors of future behaviour.

This isn't a criticism. It's cognitive science.

Client Proxies can be built to be 100% self-aware and 100% honest.

Asking the question changes the answer

The moment you ask "Would you pay more for sustainability?" you've made sustainability salient and created social desirability bias. You haven't measured a preference. You've manufactured it.

Client Proxies can be modelled to give answers untainted by biases that mask real decision drivers.

Traditional research is too slow for iteration

Traditional research can't test 12 positioning variants. Human respondents can't hold more than a few options in their minds separately.

Also, you can't resurvey a human panel when interesting questions come up. You get one shot, with one question set, with one timeline (usually measured in months).

Strategic decisions need iteration. You find something interesting—you want to explore it. You test a positioning variant—you want to understand why it resonated with one segment but not another.

With synthetic research, you can go back to your Client Proxies as many times as you need. No additional recruitment costs. No time delays. No survey fatigue.

Why professional services are the perfect environment for synthetic research

Synthetic research works best when:

• Buyers are sophisticated and follow predictable decision heuristics
• Buying decisions are high-stakes and considered
• The buyer pool is difficult to access via traditional methods
• Strategic questions require iteration and exploration

That's professional services buying.

Law firms, consulting firms, accounting firms and financial services are markets where traditional research is prohibitively expensive, access is difficult, and the strategic questions that matter most can't be answered with a single survey.

Why you can't just "ask ChatGPT" (or Copilot or Claude or...)

Synthetic research isn't prompting ChatGPT:

You are 200 CFOs in financial services. How would you evaluate cybersecurity consultants?

If you give that prompt, ChatGPT will answer in seconds. The speed alone tells you it hasn't modelled 200 individual CFOs. ChatGPT's off-the-cuff answer comes from averaging patterns in its training data; it's not modelling decision-making.

Real synthetic research requires three things ChatGPT doesn't give you.

1. Knowing what dimensions matter

Not all CFOs, general counsel or board members think alike. Some prioritise vendor track record. Others prioritise cost or integration. Some are risk-averse. Others are early adopters.

Building useful synthetic research requires knowing:

• What dimensions are relevant to the decision at hand
• How those dimensions interact
• Where individual variation matters and where it doesn't

You can't prompt your way to this in a chat window. Working with hundreds of variables across dozens of variations to ask multiple questions requires sophisticated code and API access to the right LLM model for the job.

It also requires expertise in the market, the decision and buyer psychology.

2. Knowing where LLMs are reliable and where they're not

Large language models are good at:

• Replicating language patterns
• Identifying decision criteria from existing data
• Generating reasoning chains

Large language models are bad at:

• Knowing how your firm is perceived (unless you're a Deloitte, your reputation isn't in training data)
• Admitting when they can't give useful answers
• Understanding context-specific reputation dynamics

If you don't know these limitations, you'll get answers that sound authoritative without any connection to reality.

3. Expertise that connects data to action

Asymmetric Strategic Intelligence (ASI) is built on our 30 years of direct response copywriting and decades in professional services.

The lineage matters because our clients want decision intelligence because their goal is to convince someone of something — a client to instruct them, a procurement team to shortlist them, partners to use the new CRM.

That's why we won't slide a deck-of-a-thousand-pie-charts across the table and walk away. We follow up the data with the strategic consulting you need to turn your client insights into positioning and marketing that works.

What good synthetic research delivers

Imagine you want to know what decision criteria matter most to general counsel evaluating environmental lawyers.

An LLM will tell you: "Track record, sector expertise, cost, responsiveness".

Likely true. Definitely not actionable. Every environmental lawyer claims those things.

What you need to know:

• Which criteria are hygiene factors (necessary but don't differentiate)
• Which criteria are differentiators (swing decisions between shortlisted firms)
• How buyers weight competing claims when they can't verify everything
• What language patterns signal credibility versus triggering scepticism

Synthetic research can tell you. Combine that with expertise in persuasion, and you have decision intelligence that turns your firm into the obvious choice.

That's Asymmetric Strategic Intelligence (ASI).

Does synthetic research actually work?

In mid-2024, Mark Ritson, marketing professor, consultant and columnist, wrote in Marketing Week that synthetic data is "as good as real".

The academic validation

Stanford Computer Science Department's peer-reviewed research shows that synthetic personas predict what their human counterparts would do with 85% accuracy.

That's not "good enough" — that's the ceiling of what's possible to know about decision-making. Ask your human clients the same question two weeks later, and probably less than 85% will give you the same answer.

In 2025, the Journal of Marketing published a study by the Wisconsin School of Business showing 95% alignment between synthetic insights and real consumer data.

Studies by Brand et al. (2023) (Microsoft and Harvard Business School), Li et al. (2024) (Stanford and Boston universities) and Dillion et al. (2023) (University of North Carolina) find the same alignment across diverse research applications — from conjoint analyses to brand perceptual mapping to moral judgement scenarios.

The studies show synthetic research replicates real human responses with high accuracy when the methodology is rigorous and the research question is well-suited to the approach.

The enterprise validation

EY, J.P. Morgan Payments, BP, Major League Soccer, Unilever and Coca-Cola all make high-stakes marketing decisions. They're all working with synthetic personas for strategic intelligence.

The market adoption data

Qualtrics Research (2025) surveyed 700+ senior marketing and insights executives about synthetic research and AI-powered intelligence.

Adoption velocity:

• 95% plan to use decision intelligence within 12 months
• 96% report positive impact on intelligence capabilities
• 91% more confident using AI for strategic intelligence

Accuracy perception:

• 92% say intelligence-based methods are MORE accurate than traditional approaches
• That's practitioners reporting actual experience, not vendor claims

The scepticism pattern:

• 51% initially cite stakeholder scepticism as barrier
• 96% report positive impact after use
• Pattern repeats: scepticism before use, validation after results

Where to use synthetic research

Synthetic research is excellent for:

Understanding decision heuristics: How do clients actually make decisions? What gets weighted heavily? What's ignored?

Testing positioning and messaging: Which value propositions resonate? Which trigger scepticism? How do buyers interpret competing claims?

Mapping buyer journeys: What questions do clients ask at each stage? When do they involve procurement? When does legal get involved?

Exploring strategic scenarios: How would buyers respond to a new service line? A price increase? A repositioning against larger competitors?

Identifying language patterns: What words signal credibility? What phrases trigger scepticism?

What synthetic research can't tell you

Synthetic research can't tell you:

• How your specific firm is perceived (unless you're a household name, market-specific reputation data isn't in training data)
• Quantitative market sizing (requires real survey data)
• Brand-specific sentiment (requires asking actual clients)
• Highly contextual decisions (if the answer depends on personal relationships or unique firm circumstances, you need qualitative interviews)

What synthetic research can do instead

Even when we can't tell you how your firm is perceived, we can tell you what clients use to make decisions. When you understand the actual decision heuristics — not the rationalised survey responses — you can position yourself strategically.

We can't tell you if clients perceive your firm as "innovative" (that requires asking them). We can tell you whether "innovation" is a differentiator or a hygiene factor in your category, and how buyers evaluate innovation claims.

That's often more useful than knowing your current perception, because it tells you whether "innovation" is worth investing in as a positioning pillar.

What ASI does differently

We built Asymmetric Strategic Intelligence (ASI) specifically for professional services markets. Our methodology is grounded in 30 years of direct response copywriting, decades of experience in professional services, and a clear understanding of what LLMs can and can't do.

We don't generate plausible opinions. We model real decision-making. And we only answer questions where synthetic research can give you actionable insight.

Our process:

• Map the decision architecture: What dimensions matter? Who's involved? What gets weighted heavily?
• Build behavioural models: Create Client Proxies that represent variation in your buyer population (not averaged opinions)
• Validate against real-world data: Ensure synthetic insights match actual human behaviour
• Translate intelligence to action: Connect findings to positioning, messaging and strategic decisions that drive conversions

See synthetic research in action: Client work

Explore what we can answer: How ASI works

Discover how our consulting operationalises your discoveries: ASI

See how we protect your investment: AI-powered marketing advisors

Find out what synthetic research can answer for your business

Book a 15-minute call with Steven Lewis to discuss a specific strategic question you're facing — positioning, competitive perception, messaging, buyer behaviour — and determine whether ASI Research can give you actionable intelligence.

In that conversation, we'll examine your question, establish whether synthetic research is the right methodology to answer it (we'll tell you honestly if it's not), and show you what the intelligence would look like — as well as how you'd use it.

You'll leave knowing whether ASI Research fits your situation or whether traditional research would serve you better.