A new tool held to a high standard.

AI is the most useful thing to happen to research analysis in a generation, and the most over-promised. We've taken our time with it. The point was never to have AI for its own sake. It was to get clients clearer answers, faster, without giving up an ounce of the rigor this firm was built on.

This is new equipment, not new territory. For decades we've worked at the frontier of how survey data gets analyzed — segmentation, modeling, the methods that turn a pile of responses into a decision someone can act on. AI is the next method in that line, and we adopt it the way we've adopted every method before it: deliberately, and only once we trust it. We didn't bolt an AI feature onto someone else's software. We built our own client dashboard around it, so we control exactly how it touches your data.

In practice, AI does the heavy, slow work that used to sit between fielding and findings. It sorts thousands of open-ended comments into themes in minutes instead of days. It flags patterns across segments that are worth a closer look. It drafts a first read of the data for a human to sharpen. We can use it to build bespoke tools as we progress as a firm. That speed goes straight to you as a faster turnaround, and as more time spent on what the numbers mean instead of wrestling them into shape.

Here's the part we care about most: AI is confident even when it's wrong, and in our line of work a confident wrong number is dangerous. So we build everything around one rule: every claim has to trace back to the data behind it. The dashboard is designed so that a figure or a finding carries a footnote leading straight to its source, where you can check it yourself. We're constantly improving how well that works, but the principle isn't up for negotiation. You should never have to take the AI's word for it.

Your data stays yours. Anything you or your respondents enter into the dashboard is never used to train AI models. It's handled through Anthropic's commercial API, which states it does not learn from prompts, and it's kept for your project and no one else's.

Speaking of Anthropic, after months of testing the available tools against real projects, we chose to build on Anthropic's Claude. The deciding factor was a match in priorities: Anthropic builds for safety and for grounded, checkable output, which is the same standard we hold our own work to. Tools will keep changing, and we'll keep choosing whichever one serves the work best. But we'll always tell you what's under the hood.

AI hasn't changed what we do. It's made us faster, and it's given clients a clearer line of sight into their own data.