Skip to content 
Insight teams have more AI tools available to them than they know what to do with. The tools are getting cheaper and more capable all the time. The harder problem is whether the team has the judgement to evaluate them, integrate them well, and advise the rest of the business on where they help and where they don’t. That’s the capability gap, and it’s now the bigger problem to solve.
Walk into most insight functions today and you will hear two things at once. The first is excitement: AI is letting the team do work that wasn’t possible six months ago. The second is unease: nobody is quite sure which AI tools are worth using, where they break, or how to handle a CMO who has just been pitched a synthetic-data platform and wants to start tomorrow.
Both reactions are reasonable. AI is genuinely opening up new kinds of research work. It is also being oversold by vendors with strong commercial incentives. Telling the difference is now where the real value sits, and that is a human-capability question, not a tool question.
The tools are now the easy part
Three things have shifted in the last twelve months: AI research tools have proliferated, pricing has come down sharply, and the basic capability gap between the best and the rest has narrowed. Most insight teams that want access to a synthetic-respondent platform, an LLM-powered probing tool, an AI-driven trends scanner, or a qual themeing engine can have one inside a quarter.
That changes the bottleneck. The bottleneck used to be “do we have the technology?” The bottleneck now is “do we know how to use it well?”
What 'AI capability' actually means inside an insight team
There are three dimensions worth distinguishing.
Critical evaluation. Can the team tell a credible AI tool from a polished but unaudited one? Can they ask the right questions about training data, accuracy claims, model transparency, bias testing? When a vendor pitches 95% accuracy on synthetic respondents, can the team work out what the inverse means for a global segmentation? This is partly technical literacy and partly applied scepticism. It is the muscle Blog 1 in this series argued for.
Integration. Can the team thread AI into the research workflow without losing rigour? Where does AI sit in the project plan, where doesn’t it, and how does the team make sure outputs are still curated by a named human before they shape a recommendation? Knowing what to automate and what not to is harder than it looks once a tool is in the room.
Advisory voice. Can an insight leader walk into a conversation with the CMO, the CFO or the board and explain, calmly and credibly, where AI is the right answer for a specific brief and where it isn’t? This matters because most senior leaders inside client organisations are getting AI advice from vendors. Insight teams that can be the trusted internal voice on AI tend to keep their seat at the table. Insight teams that can’t tend to lose it.
All three are learnable. None is a one-day workshop.
Why the gap shows up where it shows up
Three patterns we see regularly inside client organisations.
The first is the team that adopts a tool without auditing it. A vendor lands on a friendly stakeholder’s desk, a procurement decision happens, and the insight team finds itself integrating a black-box product into existing programmes without ever interrogating what is inside it. By the time the methodology questions come up, the contract is signed.
The second is the team that lets AI write the deck. Outputs that come out of an LLM look polished and confident, which is the problem. Without a strong critical-thinking discipline, recommendations get accepted because they read well rather than because they are right. Errors get past the review because the prose is clean.
The risk is brain rot. A team that gets used to AI giving polished answers can start to outsource its thinking to the tool, and lose the muscle that made the team valuable in the first place.
The third is the team that gets sidelined. The CMO has already decided AI is the answer. The insight team, sensing the conversation has moved on, doesn’t push back. Six months later the campaign underperforms, the post-mortem is uncomfortable, and the question of why nobody flagged the methodological risk goes unanswered.
All three patterns are capability problems, not tool problems. None of them is solved by buying a different platform.
How to start closing the gap
Three places that are practical to start, in order.
Get clear on principles. Before evaluating any specific tool, the team needs a shared view on what they are evaluating against. Six principles we use, and have written about elsewhere in this series, are humans in control, purpose-led adoption, governance and trust, bias awareness, continuous improvement, and transparency over salesmanship. They give a team something to actually test a vendor’s claims against.
Build critical thinking specifically about AI. The general critical-thinking skills insight teams already have do not automatically transfer to AI inputs and outputs. Synthetic data needs different scrutiny than survey data. Black-box models need different governance than published research. Teams that explicitly train this dimension catch problems earlier.
Develop the trusted-advisor capability around AI. The other side of the same coin: not just being right about AI internally, but being able to walk a CMO through it credibly. That is a different skill from delivering insight. It is the consultancy skill, applied to AI specifically. Many insight leaders find this is the bit that gets the most resistance and pays back the most.
Build new craft skills, not just sharper old ones. The skills the next five years will reward are things like insight engineering (working with LLMs the way data managers used to work with databases), facilitation, scenario planning and scrum-style ways of working with internal stakeholders. They are not all in the current insight skills curriculum, which is partly why the teams that have them stand out.
What we’re doing at STRAT7
STRAT7 invests in this on both sides. On the tools side, Nucleus is our AI hub: design, discovery and impact agents working alongside our consultants under the six principles above. On the human side, we are running an internal training programme to develop AI capability across our team. Critical thinking, evaluation, and the trusted-advisor skill set, applied specifically to AI. We will share more on that work when there is something concrete to share, including what we are learning that other insight teams might find useful.
In the meantime, the practical move for any insight leader reading this is simpler than buying another tool. Start a conversation inside your own team about which AI tools you have, which ones you trust, and how you actually know. Most of the value sits in that conversation. The tools are the easy part.
Webinar on-demand
Beyond speed: How AI is broadening insight's commercial value
To learn more about Nucleus, our AI hub, watch our webinar recording where we explore:
- Where the real value of AI lies, beyond speed and cost.
- Why briefs that were impossible a year ago are now on the table.
- Case studies and real-world applications with leading brands.
- How this changes the conversations insight teams have with the board.
- What changes when teams move from “what can we afford?” to “what do we need to know?”
About the author
- Hasdeep Sethi
- Group AI Lead & Data Science Director, STRAT7
Hasdeep is a data science and AI leader with 10+ years’ experience building, delivering and setting strategic direction on data science, machine learning and AI projects. He helps lead STRAT7’s AI offer through Nucleus and the AI Innovation Lab, and speaks regularly at industry events including MRS, GRBN, IIEX and ASC. His particular interest is how AI and synthetic data are reshaping insight and research.