Product teardown · essence only It listens to the call and collapses five screens into one. A real-time assistant for salespeople on live calls. It transcribes as they talk, recognizes what's being discussed, pulls the matching inventory and policy, and puts the answer on screen at the moment it's needed. The agent stops searching and goes back to talking. 62K+Calls processed per day 22.5KAgents and admins on it 75+Dealer organizations live 3Platforms: web, tablet, phone 01The load it removes The job was never selling. It was tab-switching. What an agent held open before, and what replaced it. Before — five-plus lookups, mid-sentence Inventory system Pricing tool CRM record Policy docs Competitor lookup → After — one surface, updating itself Transcript Live assistant Customer record 5+Apps open at once to answer a basic question 25–90Calls a day, none of them memorable 20–60Store policies to recall on cue 02The mechanism Six steps, all inside the length of one call Nothing here waits for the conversation to end — except the last step. 01 Listen Audio streams in as the call happens. 02 Transcribe Speech becomes live, readable text. 03 Recognize Vehicles, intents and objections get named. 04 Retrieve Matching stock and policy are fetched. 05 Surface The one relevant answer appears on cue. 06 Summarize Hang-up writes the notes and follow-ups. Step five is the whole product. Retrieval is commodity; choosing the single thing worth showing at second forty-three of a live conversation is not. 03The layout that worked Two panels hold still so one can move The arrangement arrived at after two rejected alternatives. Anchored Transcript Scrollable, referenceable record Somewhere to look back to Moves Live assistant One suggestion at a time Policy reminders on cue Action items captured as they land Anchored Customer record Persistent through the whole call Prior visits, preferences, payment Fixed sides, live centre — the reason the screen stays legible at conversation speed. 04What the failed versions taught Every rejected prototype produced a rule Two explorations were built and killed. Four rules survived them. What was tried Why it broke The rule it produced A chatbot — one clean answer, nothing else The feed moved at talking speed with nothing to anchor to Fast is not usable. Give the eye a fixed point. Dropping the customer profile to save room The agent had to hold the context in their head instead Never make the human the memory. Vehicle options as rich cards Ate the screen and fell apart past a handful of trims Dense and scannable beats pretty and paginated. Trusting the model's output as final A wrong read mid-call had no escape hatch If it can be wrong, it must be correctable — live. 05Four surfaces One engine, four places it shows up Each surface exists because a different moment demanded it. Desktop The live copilot The three-panel call screen. Inventory, policy and action items assemble themselves while the agent talks. Auto-summary at hang-up Dashboard Alerts, routed by role Voicemails, unhappy customers, same-day appointments. Managers choose which alerts reach which agent. 75+ types · 1-hour callback window Phone Showroom capture The same listening, untethered. Open, find or create the customer, record the in-person conversation. Hands the summary to finance Internal Policy authoring Each organization runs a different playbook, so policies are configurable: trigger, wording, scope, on or off. Per-organization rules 06The non-obvious part The hard problem was consent, not accuracy Recording competitive, autonomous salespeople for the first time. Coaching, not surveillance Managers get review screens framed as helping the agent win, not catching them out. Framing decides adoption Pay the agent first Solve today's annoyance before asking for tomorrow's data. Value has to land before trust is requested. Utility precedes buy-in Precision over volume The right sentence at the right second, not a dashboard of everything that might apply. One answer beats ten Strip it all back and the product is one bet: an agent who never has to look something up will out-sell an agent who does. Everything else is plumbing. Provenance Distilled from a single public product-design case study published in 2026, covering roughly eight months of work on a vertical AI assistant for automotive retail. Product and company names are omitted by request. Figures — call volume, seat count, organization count, alert types, and the workload ranges — are the case study's own; the pipeline, the rules table, and the three principles are my compression of its narrative, not labels it uses.
