Planr Pal
About
Planr Pal is the AI assistant for Planr Flow and Planr Ledger, sitting on top of a company’s order data and turning it into answers, actions, and a bit of foresight. The idea was simple: people already keep their information in Planr, so why make them dig for it? Instead of hunting through tables and filters, you should be able to just ask.
The idea
Most AI features bolted onto software feel like a chatbot in the corner that does not really know anything about you. I wanted Planr Pal to be the opposite, grounded in your actual data, aware of the orders, stages, and context that already live in the system, so its answers are specific rather than generic. Ask what is overdue, what is waiting on a customer, or what changed this week, and it answers from the real state of your business.
The bar I set was that Pal should feel like asking a colleague who has the whole system open in front of them. It greets you, asks what you want to know about your production data, and then goes and looks. You can point it at a specific order or customer with a mention instead of describing it in words, and the requests that come up every day are packaged as skills, so a question a planner asks every morning is one click rather than a paragraph of typing.
Building it
Planr Pal is built into the existing Next.js and TypeScript stack, so it shares Planr’s data model rather than living beside it. The hard part was never calling a model. It was giving the model the right context without overwhelming it, and making sure every answer could be traced back to something real in the data. Guardrails, clear sources, and a tone that stays helpful without pretending to be more certain than it is.
Pal lives inside the product rather than in a separate tab. The Ask Pal button sits in the header of every screen in Flow, with a keyboard shortcut next to it, so you can open it in the middle of the work orders list, ask about the row you are looking at, and get back to the list without losing your place. It runs under the same session and permissions as the rest of Planr, so it only ever sees what the person asking is allowed to see.
Under the hood it runs on the AI SDK. A question is turned into a set of typed lookups against the same data the rest of Planr uses, and the results are handed to the model as context, so the answer is built from the current state of the business rather than from whatever the model happens to remember. When Pal names an order or a date, that comes straight from a record it fetched, and when the data is not there it says so instead of guessing.
What’s next
Planr Pal is still early and evolving alongside the rest of the product. The direction is less about flashy chat and more about quietly removing the small friction of finding things, so the system does more of the looking, and you do more of the deciding.