Case study
Fresh KDS is a kitchen display system that keeps orders organized, a team on track and every shift running smoothly. About 35 people.
They had already built a brain. Keeping it current was the problem
Fresh built a company brain on top of GitHub and asked people to update it. David, their CEO, is direct about what happened next. People didn’t do it consistently, and it drifted out of date. Some information lived in their CRM. The rest, the operational detail about a customer, lived only in people’s heads, in meeting transcripts, and in email and Slack.
Growth made compliance harder, not easier
Asking a team to keep a knowledge store current is a rule, and a growing team with more customers makes any rule harder to enforce. Nobody writes down the information most worth having twice.
The MSA with a certified-mail clause
Fresh finished a negotiation with a large customer, and the MSA carried requirements that would have been easy to miss. One of them meant sending something by certified mail. The AI pulled the tasks and deadlines out of the agreement and they met all of them. David is clear about the counterfactual: without the brain, they would have missed some.
Why they built the maintenance layer with us
Two things decided it, in David’s words. A direct connection to communications, which is where much of the operational knowledge lives, and a workflow engine that could update the brain. He was skeptical it would work. The brain feature existed before we started, but we built the maintenance process on Fresh’s input, so nobody knew in advance whether it would be good enough to adopt.
What the brain looks like
There’s a page per entity type: customers, partners, competitors, products. Each lists its entities, and each row links to two pages. A detail page holds a small amount of critical information about that entity. A log page holds every update about it. The log exists so anyone can rebuild the detail page if something goes wrong, and so it can be refreshed to keep it genuinely current.
How it stays current
When meeting notes land in someone’s inbox, or an exchange with a customer happens over email or Slack, this+that creates a task in that person’s DoBox with a proposed update. They accept it, reject it, or edit it to add color or fix a mistake. That’s semi-automatic by design. Fresh intends to automate more of it once the system has run long enough and performed well enough to earn that, and not before.
What changed
Operational information now lives in the brain, and it stays current as the work happens rather than in a cleanup pass nobody does. As David puts it, there’s no longer a chance that someone who talks to a customer won’t know the latest about them, unless they haven’t done their homework. Workflows that read the brain produce better output, because what they read is current. Automated work like building a pitch deck needs less oversight, because people can trust the information instead of confirming it with a colleague first.
What it saves, in David’s estimate
Each person on the team saves an hour or so a day by approving updates rather than making them by hand. The information about each customer is better and fresher. People spend more of the day on the job they were hired for, like selling, and the habit of reviewing updates keeps the quality up rather than letting it drift.
Who actually reads it
Agents are the primary consumers of what the brain holds. People read it too, mostly to check what an agent put into a document or an email before it goes out, and David describes that as far easier than searching mail and Slack threads to confirm the same thing.
It’s surprising how much information is known by only one person, even at a relatively small company like ours. That’s why we built a company brain ourselves earlier this year. But we quickly learned that it’s hard to get people to update it and we had to nag. this+that keeps our brain up to date semi-automatically and we’re eager to move to an automatic approach.
David Corts, CEO, Fresh Technology
What breaks if you turn it off
We asked. “We’d go back to the same problems we had before. People having different views about customers based on their previous conversations rather than having the complete picture. Decision-making would be worse, and automation, such as building pitch decks, would require more oversight since people would have to confirm information with colleagues rather than trust that it was accurate.”
Where it goes next
Fresh wants more of the maintenance automated, and that’s the direction we’re building: fewer proposed updates a person has to approve, and more that can be trusted to land on their own. The order matters. It earns that by running well first.