Agentic API
Tell GWI's agent your goal. It works out what to ask, picks the right data and audiences, and returns a finished narrative with the evidence behind it.
Connect it to the tools you already use — a pitch tool, an assistant you've built, your CRM, or a product you sell. No agent infrastructure to build at your end. The reasoning happens on ours.
Here's what the agent does in between, and what comes back.
A goal or a question. For example: understand what matters to premium streaming subscribers in the UK, US and Germany, and where the three markets differ.
Works out what to ask, builds the relevant audience and returns a complete answer.
A written narrative, with the figures, the audience and the sample behind every claim.
Nothing in the middle is yours to build. Nothing in the answer is unsourced.
The agent plans the whole analysis, and GWI builds, tests and maintains it as models change. No orchestration layer or reasoning engine for you to build.
Built on 2M+ interviews a year across 50+ markets. Every figure, audience and sample comes back traced to people who were asked.
Ready to use, not ready to interpret. That's the difference between a capability people try once and one they use every week.
.svg)
Your research team owns the audiences and what can be surfaced. The hundredth request is reasoned like the first, so the numbers don't drift apart.
The analysis is an API response, so it renders wherever you want it.
A seller names the brand they're pitching and gets the audience case back: who the customers are, what they care about, and where your inventory meets them. Nobody opens GWI. This is the first guided workflow GWI has built, and it's live.
Your existing tool gains the ability to answer questions about people properly, rather than confidently and wrongly.
The analysis is the product, differentiated by real consumer data instead of a generic model.
A CRM action, an internal portal, a planning tool. The reasoning doesn't have to live at your end, so the system doesn't need to be intelligent to give an intelligent answer.
Some teams want the building blocks. Some want the finished answer. The difference is who does the reasoning.
Get the parts. Do the reasoning.
Build the audience, run the query, pull the detail, and assemble the answer in your own code. You decide what to ask and in what order.
Send the goal. Get the analysis.
GWI plans the steps, calls its own tools in order, and hands back the finished narrative with the evidence behind it.
What is the Agentic API?
It's an AI agent API where you send a goal and GWI's agent returns a finished analysis: a narrative plus the supporting insights. The multi-step reasoning runs on GWI's side, so you don't build or maintain an agent to get research-grade consumer answers.
How is the Agentic API different from the Spark API?
Both reach the same GWI data, audiences, and numbers. The Spark API hands you the parts and you do the reasoning, which means more effort and more control. The Agentic API does the reasoning and returns the result, which means less effort and less control. The narratives differ because a different agent does the thinking.
Do I need to build my own AI agent?
No, and that's the point. The orchestration, tool sequencing, and testing all happen on GWI's side. One call returns a finished answer instead of a data set your team still has to interpret.
Can our insights team control which audiences the agent uses?
Yes. The insights team defines and owns the saved audiences everyone works from, sets the boundaries on what can be surfaced, and decides when the wider team gets access. Everyone queries from the same definitions, so answers don't drift as you scale.
Does every answer come with evidence?
Yes. Each response returns with the figures, the audience, and the sample it rests on, so it holds up when someone challenges it.
How do we get started?
Read the API documentation, or talk to us about which route fits what you're building.