0-to-1 AI Product · Built with Lovable

BEACON — AI Customer Success Copilot

An AI-powered Customer Success command center for CSMs. BEACON reads customer signals — risk score, usage drops, support tickets, champion activity — and guides you through a clear workflow: understand the situation, investigate the reasons, decide on a strategy, prepare outreach and meeting briefs, then act.

Role
Solo builder — problem to working prototype
Method
CS workflow design, prompt engineering
Scope
Customer Success, AI copilot for CSMs
Deliverable
Working prototype ↗
The problem

CSMs know an account is at risk long before they know what to do about it

A dropping usage graph or a cluster of tickets tells a CSM something is wrong, but turning that signal into a coherent understanding — then a strategy, then an actual outreach message — is still manual work done differently by every CSM on the team.

Solution

One workflow: understand → investigate → decide → prepare → act

UNDERSTAND

Every signal in one view

Risk score, usage drops, support tickets and champion activity surface together, so a CSM starts from the full picture instead of one dashboard at a time.

INVESTIGATE & DECIDE

From "what's wrong" to "what to do"

BEACON walks the CSM from investigating why an account looks risky to a recommended strategy — not just a red flag with no next step.

PREPARE & ACT

Outreach and meeting briefs, drafted

Once a strategy is set, BEACON prepares the outreach message and meeting brief itself, so the CSM's time goes to the conversation, not the write-up.

Prototype

A working copilot, not a static risk dashboard

BEACON is built and live — the full understand-investigate-decide-prepare-act workflow works end to end in the prototype below.

Live prototype — BEACON Open full screen ↗
Open interactive prototype ↗
Process

Designed from having run this workflow manually

Discovery
Problem discovery — started from how differently CSMs each turn a risk signal into action, with no shared playbook.
Flow
Workflow design — mapped the understand → investigate → decide → prepare → act sequence as a guided path, not a dashboard.
Prompts
Prompt engineering — wrote the prompts that turn raw account signals into a recommended strategy and drafted outreach.
Output
Build — shipped a working, clickable copilot on Lovable rather than leaving the concept as a deck.