In-App AI Copilot. @in-app-copilot

The In-App AI Copilot embeds an assistant inside your existing product — not a chat widget bolted on, but a copilot wired into your data model and permissions that does a defined set of things well: drafting within the product's domain, explaining data on screen, executing multi-step actions users would otherwise click through. The finish state is the copilot live for real users on a scoped capability set, with permissions enforced, quality measured against an evaluation set, and usage telemetry showing where it earns its place.

Scrums.com builds LLM-powered product features for production, with senior AI engineers embedded in your existing SDLC and latency, quality, and token cost tracked as delivery metrics from sprint one. The discipline that makes copilots succeed is scope: a copilot that does five things well beats one that does fifty things plausibly.

Capability Scoping

The copilot's capability set chosen deliberately from your product's jobs-to-be-done — each capability with a defined input, output, and quality bar, and an explicit list of what the copilot will refuse.

Context & Permission Layer

The copilot grounded in the user's actual context — current screen, account data, history — with your product's permission model enforced at retrieval and action level, so it never sees or does what the user could not.

Copilot UX Integration

Native integration into your product's UI: inline suggestions, command surfaces, or side-panel chat as fits the workflow, with streaming responses, undo for actions, and graceful failure states.

Evaluation & Telemetry

A per-capability evaluation set run before launch, plus production telemetry on usage, acceptance rates, latency, and token cost — the numbers that tell you what to build next.

1. Scope — Pick the capability set, fix the context and permission design, and build the evaluation sets.
2. Build — Implement the copilot against your product APIs, iterate on prompts and retrieval until evaluations pass.
3. Handover — Staged rollout to users, telemetry dashboards live, and a playbook for adding capabilities safely.

The In-App AI Copilot is a menu item on the Scrums.com delivery catalog, available at every plan tier. Add it to your plan backlog and your delivery team schedules it like any other item — scoped, tracked, and reported through the SEOP. See Delivery Plan Tiers.

Which model does the copilot run on?

Whichever fits your quality, latency, cost, and data-residency constraints — commercial APIs or self-hosted open models. The choice is made at scoping against your actual constraints, and the architecture keeps it swappable.

Can the copilot take actions, not just answer?

Yes, within the scoped capability set: actions go through your existing product APIs with the user's permissions, confirmation steps where consequences are real, and undo where possible.

How do we grow it after launch?

Telemetry shows which capabilities earn usage; new ones ship behind the same evaluation gate. When the copilot becomes core product, the AI Feature Productionization item hardens it for scale.