INTERFACE TYPE
AI workspace
INDUSTRY
Research
AGENCY
GoInvo
PROJECT STAGE
Not independently confirmed
Report wizard to convert insights into a client-ready deliverable
Ipsos Facto: interface example from the official agency account. Credit: GoInvo. Official source ↗. Open full image

The problem and context

Ipsos Facto is an internal generative-AI platform that helps market researchers turn Ipsos’s own historical research into cited insights. GoInvo reports that it led the design work, with a team of about 1.5 designers over six months, in collaboration with Ipsos’s product leadership. GoInvo’s original account.

GoInvo describes what it found when it arrived: Facto had launched as a basic enterprise chatbot. Chat, files, and tools were separate, which forced frequent switching. The experience was generic rather than shaped for research, and there was no clear path from an insight to a deliverable. The agency links this to low adoption and dissatisfaction with the AI’s results.

This examination focuses on one idea that runs through the project: a capable model inside a blank chat box is not yet a research tool.

The constraints

  • Researchers were expected to “prompt their way to insight” with no scaffolding.
  • Trust depends on citations. Research insights are only usable if a researcher can trace them back to the source material.
  • The product was evolving in place. Changes had to ship alongside ongoing use, not after a pause for a full redesign.

Design decisions worth examining

1. Replace the blank screen with guided workflows

GoInvo reports adding guided workflows and research-specific tools for core tasks such as content curation, meta-analysis, and transcript coding, plus templates that automate parts of a meta-analysis. GoInvo’s original account.

Our reading: this reframes the AI from an oracle to a set of familiar research procedures with AI inside them. A researcher who knows how a meta-analysis works can recognise the steps, which lowers the skill needed to get a good result.

2. Make the chat part of the system, not a sidekick

The agency describes letting users pull a template or tool into an existing chat, and making insights expandable to reveal details and citation metadata from different source types. GoInvo’s original account.

Our reading: this answers the switching problem directly. The conversation becomes a workspace that tools join, rather than one more panel to move between. Expandable citations put the evidence one click from the claim, which is where it is most useful.

3. Explain limits inside the interface

GoInvo reports adding suggestions, limitations, and disclaimers to the UI, and making visible how the AI reaches answers and where it fails. These sit within five stated principles: structured paths over blank screens, customisation by research type, transparency, visual cues to reduce complexity, and clear privacy and permissions. GoInvo’s original account.

Our reading: telling users where the system fails is unusual and valuable. It sets realistic expectations and turns the disclaimer from legal boilerplate into usable guidance.

Contribution, outcomes, and evidence

Documented contribution: UX strategy, guided workflows, research tools, interface improvements, and a clickable prototype for future direction. GoInvo credits Ipsos’s product leaders for the product vision and describes moving from vendor to core contributor in roadmap meetings. Ipsos built and runs the platform.

Stage: the account mixes shipped work (“dozens of UX band-aids”) with prototype concepts. It is not always clear which illustrated features shipped. The agency itself says the tool is “only 25% of the way” to a mature enterprise product.

Outcomes: the page reports high internal adoption, prompt and API volumes, and an AI-certification programme across many countries. These are Ipsos-wide figures. They show the platform is used heavily, but the page does not isolate how much of that came from the design work, so we do not treat them as design results.

What we did not verify: we have no access to the internal product. Feature descriptions come from GoInvo; “Our reading” passages are our interpretation.

A practical lens

  • Wrap AI in the procedures your users already know. Templates for familiar tasks beat a powerful blank prompt.
  • Bring tools to the conversation. If users keep leaving the chat to do something, that “something” belongs inside it.
  • Document failure modes in the product. “This tool is weak at X” is more useful than a generic warning.

Where this may not transfer: guided workflows suit repeatable professional tasks. For exploratory or creative use, rigid templates can get in the way, and an open prompt may serve better.

Original source and editorial responsibility

Source: GoInvo’s original account, checked 16 November 2026. Written by Best UX Portfolios from the agency’s published case study and images. We have not used the product or interviewed Ipsos. Spotted an error? Request a correction.