How Day AI’s product team hears every customer, in Linear

Getting customer feedback has always been touted as a “good” thing, with the idea being that more feedback equals more opportunities to make your product better.

Of course, this ignored the rather inconvenient reality that turning feedback into something useful was expensive, not to mention exhausting.

Every support ticket, sales call, and Slack thread had to be tracked by someone before being painstakingly translated into something Product could prioritize and, at long last, picked up by an engineer. Faced with all this, PMs got selective about the feedback they collected, cultivating relationships with a handful of customers and relying on instinct for the rest.

AI turns this on its head, with LLMs making it cheap and easy to ingest huge amounts of information and turn it into something structured. As a result, the alpha has shifted from deciding which feedback is worth collecting to getting context in front of teams (and their agents) where they make product decisions.

Day AI is already running this in production, turning customer conversations into actionable feedback in Linear. In just three months, their automated intake system has monitored their team’s conversations with more than 200 customers and captured 1,200 pieces of real feedback that now sit alongside and inform their development work.

Building an always-on intake system⁠

Day AI made customer feedback a continuous part of product development by bringing it into Linear, where it can be viewed in context of existing work and used to decide what to build next.

Day AI helps companies understand their customers by capturing every conversation and touchpoint and structuring the information into a “context graph,” preserving what was actually said and connecting it to the relevant people, companies, and deals in the sales pipeline.

To connect that feedback to real product development, the team enabled Customer Requests in Linear, connected Day AI through MCP, and described in natural language how work is organized in their workspace. Day AI also syncs companies into Linear as Customers, keeping attributes such as tier, revenue earned, and sales owner current. With Linear’s views and filters, the team can see which requests are coming from their most valuable customers and prioritize accordingly.

This intake system is always on. Day AI monitors the channels where its team talks to customers, using an in-house LLM classifier to identify feedback and distinguish between bugs and feature requests. The agent checks that feedback against work already tracked in Linear. When there’s a match, it adds a Customer Request to the existing issue; when there isn’t, it creates a new issue.

The foundations of agent orchestrated product development⁠

While automating intake has solved an immediate operational problem, it’s had bigger implications on how Day AI builds product. A recent example is a feature that allows customers to change who can see their meeting recordings in bulk, instead of one by one. On each call, the request sounded like a nice-to-have, the sort of thing a salesperson might not even think to pass along to Product. But when the team saw the issue in Linear with seven of their most engaged customers attached, they shipped it.

Engineers don’t have to change how they work to make use of all this context. It shows up directly on the issue in Linear, where Day AI’s own cloud coding agents can also access it through MCP. The agents run in a sandbox (rather than on any one person’s laptop) and engineers can tag them in Slack against a Linear issue. “By the time something is fully specced in Linear,” Day AI founder and CEO Christopher O’Donnell says, “it’s effectively in production.”

Each Customer Request links back to the moment a customer asked for the feature, captured through Gong, Zoom, or whatever meeting recorder the customer uses; so when the team ships it, the Day AI PM can pull every clip of the customer asking for it and play the highlight reel while demoing the product.

There’s now a much shorter distance between hearing what a customer needs and doing something about it at Day AI. By acting on feedback before it has a chance to balloon into a problem, support tickets have slowed to a trickle, and Day AI has no one working in customer support.