Work

AI operations

Mano Swartz Furs

I built software and AI agents for Mano Swartz Furs' nationwide fur-buying program: a texting agent that checks photos and drafts replies, and voice agents that answer calls.

Mano Swartz texting inbox with a fur photo, an AI reply draft and the steps from photo review to return shipping.
20,054customer photos deliveredOrganized and returned to the client’s Google Drive, totaling 4.66 GB.
Client
Heritage retail
Role
Designed and built
Timeline
2026 · Ongoing
Stack
Next.js 15, Bun, Neon Postgres, Drizzle, Vercel, Twilio SMS/MMS/Voice/SIP, Retell, Vapi, Claude vision, OpenAI embeddings, Google Sheets & Drive APIs

Technically strong, communication skills on point, and work completed well before the deadline. I highly recommend.

Richard Swartz, Mano Swartz Inc.Heritage retail

A shared SMS work queue connects the AI texting agent, AI voice agents, offer follow-up, returns, and operating alerts.

The same-day-offer promise depended on reviewing customer photos, drafting replies, and remembering every follow-up by hand.

Customer texts, appraisals, pending offers, and shipping coordination competed for the owner’s attention.

  1. Built an inbox for customer texts and photos, with an AI agent that checks the photos and drafts replies in the owner’s voice.
  2. Added learning from the owner’s edits, so the drafts sound more like him over time. A draft goes out only when he taps send; routine replies come from his approved scripts.
  3. Built a backup AI voice agent and another for the sale-event line. Added texting during calls to the main-line voice agent and have revised its instructions since.
  4. Connected offer follow-up and return shipping, and added alerts for unusual activity and an AI assistant that answers questions from the business’s documents.
  1. SMS inbox and photo intake
  2. AI texting agent
  3. Owner review queue
  4. AI voice agents
  5. Offer follow-up
  6. FedEx returns
  7. Alerts and AI assistant

Keep judgment with the team

Let the AI check customer photos and draft replies, and keep the owner’s review before any AI-written draft is sent.

Connect the whole journey

Give intake, offers, follow-up, and returns a shared workflow instead of another isolated inbox.

What the project delivered

689customer inquiry recordsFrom 491 different callers, counted from the intake sheet on 4 September 2026.
2,076historical offers managedPast offers, managed in the same connected buying workflow.

Checks behind the work

1,627automated software tests passedRecorded in September 2026 across 63 test files. These checks help detect unintended changes; they are not a guarantee that every problem has been found.
v74main-line voice agent versionThe main-line AI voice agent reached its 74th published revision.

Say hello.

Tell me what you'd like to build or improve.

Quick navigation

Find a page.

Use the arrow keys to choose a result, Enter to open it, or Escape to close.