Pal is the ultimate all-in-one health companion

Pal is a mobile health coach app with features like a conversational voice agent, nutrition tracking, and SmartShop, a tool that rates supermarket products and suggests healthier alternatives.

Each of these features runs on Rightbrain Tasks, which are small, production-ready AI workflows that power the intelligence behind Pal’s app. The app focuses on collecting data and delivering a great experience, while Rightbrain handles the processing, reasoning, and outputs.

Building tools for agents in Rightbrain 

One of Pal’s features,  SmartShop, is a health product scanner that helps users make better choices in the supermarket. When a user takes a photo of a product, SmartShop:

• Analyses the product and gives it a health rating
• Flags hidden ingredients or misleading marketing claims
• Suggests healthier alternatives with links to buy

Behind the scenes, SmartShop is powered by a Rightbrain Task. Pal have built this task to:

• Accept the product photo and ingredients label as inputs
• Run it through a model they selected after testing multiple options in Rightbrain
• Use a carefully crafted prompt that captures edge cases and defines exactly how the analysis should be performed
• Return a structured JSON output with health rating, flags, and recommended alternatives that developers can easily render in the app

Because SmartShop lives on  Rightbrain, Pal can quickly adjust prompts, update output formats, and ship improvements instantly, all without redeploying their app.

Monitoring AI tasks in one place

Pal could have connected to a model API directly, but they needed a way to operate, monitor, and improve AI in production without waiting for engineering cycles or app redeploys, and to future-proof their AI approach through access to all leading models.

With Rightbrain, Chris can:

• Test and swap models side by side to find the most reliable one
• Refine prompts on the fly and deploy updates instantly to all users
• Handle edge cases quickly by iterating on prompts and output format

And when things fail, they’re not black boxes. Chris can inspect the inputs and outputs, diagnose the issue, adjust the prompt, and redeploy, all without downtime.

Pal builds and ships multiple AI agents in Rightbrain

With Rightbrain, Pal are able to create and embed multiple AI agents to give users a complete suite of features.

SmartShop: scan a product, get healthier alternatives

In a supermarket aisle, a user snaps a yoghurt pot. Pal analyses the label, flags misleading claims like “low-fat” that hide high sugar content, and suggests healthier swaps with purchase links. Built as a Rightbrain task with structured JSON outputs, it’s easy to refine; whenever accuracy drifts, Chris tweaks the prompt and updates go live instantly.

Daily nutrition targets with guardrails

When one user was assigned a dangerously low calorie target, Chris added a simple rule, “never recommend below safe minimums”, and rolled it out in seconds. The fix went live for all users without requiring a developer.

Content moderation in under an hour

Pal’s profanity filter, which blocks offensive usernames, was conceived, built, and deployed in a single afternoon.

Impact (so far)

Pal went from prototype to production with Rightbrain at the core. The results are measurable:

• 20+ tasks live today, with 30+ in the pipeline.
• Founder-led iteration: Chris pushes updates daily without adding to engineering backlog
• Model flexibility: test new models side by side, switch them in minutes, and optimise for speed, cost, and quality without re-engineering.
• Operational confidence: failures aren’t black boxes, Chris can monitor failed runs, detect anomalies in performance, and patch fixes fast.

“If Rightbrain shut down tomorrow, the impact across the business would be huge. There’s nothing else that makes it this easy to run AI in production.”

— Chris Davison, Founder, Pal