AI Concierge for Retail
End-to-end conversational agent resolving 62% of post-purchase requests without human escalation.
- Client
- Lumen Retail
- Year
- 2025
- Duration
- 10 weeks
- Team
- 3 engineers, 1 designer
Lumen Retail's contact volume multiplied roughly five-fold during peak weeks, and their answer had been seasonal hiring — which put their least experienced agents on the highest-stakes conversations of the year. We built an agent that resolves routine post-purchase requests end-to-end in their order and fulfilment systems, rather than deflecting customers into a form.
The brief
Absorb the peak-season volume spike on routine post-purchase requests without diluting brand voice or hiding the path to a human.
- Retail
- Agents
- CX
The decisions that mattered.
Every engagement turns on a handful of calls. These are the ones that decided whether this system reached production.
Resolution, not deflection
The agent writes to the order system: it issues refunds, generates return labels, and reschedules deliveries. Deflection metrics look good and customers hate them, so we measured resolution instead.
Every write is reversible and idempotent
Refunds and cancellations run through tools with idempotency keys and a defined rollback path, so a retry can never double-refund an order.
Brand voice enforced, then sampled
Tone and policy guardrails run on every response, and a daily random sample is surfaced to Lumen's brand team for review.
Escalation is never hidden
A request for a human is honored immediately, with full conversation context handed to the agent. Trust in the automated path depends on that exit being obvious.
What shipped, and what changed.
Outcomes
- 62% of post-purchase requests resolved without escalation
- Median resolution time down from 9 hours to 4 minutes
- Peak-season seasonal hiring reduced by 40%
- CSAT on automated resolutions 4 points above the human baseline
“We went into peak with 40% fewer seasonal agents and better CSAT than the year before. That combination is not supposed to happen.”
What we built it with.
- Claude
- LangGraph
- TypeScript
- Node.js
- Redis
- Shopify
Other projects from the shelf.
Clinical Documentation Copilot
Ambient AI scribe that drafts structured notes, ICD-10 codes, and prior-auth letters in real time.
Read moreEquity Research Terminal
Multi-modal RAG over 10 years of filings, transcripts, and broker research with citation-grounded answers.
Read morePredictive Maintenance Platform
Time-series + vision models predicting equipment failure 72 hours in advance across 14 plants.
Read moreTell us about the problem.
A 30-minute call with the senior team. We will tell you what is realistic, what it would cost, and whether we are the right people for it.