We use artificial intelligence openly and carefully across our research. It makes the work faster and sharper, but it never replaces human judgment, and no AI output reaches a client before a researcher has checked it. This page explains exactly how that works.
AI works as a fast assistant inside every stage of our method. It handles volume and speed, so our researchers spend their time on judgment, context and the field itself.
Drafts questionnaires and first pass translations, always for researcher review.
Structures raw field notes into clean, comparable records.
Recalculates weighted scores as an independent second check.
Summarizes transcripts and drafts report sections for human editing.
Used well, AI is a powerful assistant across the research cycle. It handles volume and speed so our researchers can spend their time on judgment.
Left unchecked, AI produces work that looks confident and reads well but falls apart under scrutiny. That is why humans stay in charge.
AI never replaces our researchers. It cannot stand in a showroom, feel a welcome, or judge local nuance. People do the research. AI helps them do it faster, and check it twice.
A named researcher reviews every AI output before it is used. Nothing generated by a model goes into a deliverable, a score or a client conversation unchecked.
Critical outputs are cross checked across more than one AI system. When the models disagree, a human investigates the difference and makes the final call.
Client identities and confidential data never enter an AI tool. Models see anonymized, need to know material only.
Field evidence wins. If a model says one thing and the shopper's photos, receipts and notes say another, the evidence decides, every time.
Give everything to AI without checking and the research turns weak. Use AI as a fast assistant under human judgment and the research becomes stronger, faster and easier to trust.
We treat AI the way good research treats any instrument, with documented methods, traceable sources and independent checks.
Claims of fact carry a source a reader can check. If we cannot trace a claim to evidence, it does not get published.
Critical outputs are cross checked across independent AI systems, and disagreements are investigated by a person, never averaged away.
How AI was used in a study is recorded and available to you, the same way we document sampling, fieldwork and quality control.
Confidentiality is a condition of this work. Our AI use is built so that client identities and sensitive material stay out of external tools entirely.
Every AI output passes a named human checkpoint before it is used
Two systems must agree before a critical number stands
Field evidence outranks any model output, every time
Our team is trained and tested on these rules in the Learning Lab
For any finding, you can ask exactly how it was produced
Expert human analysis makes the final call on every insight