Dealer Service Bay Case Study

Industry

Automotive Service

What We Built

Service ops platform, voice advisor

Status

Live working product

Stack

React · Gemini Live · Prisma

Sterling Motorcars is a demonstration store created by Agentify AI, an independent shop not affiliated with or authorized by Cadillac or General Motors. Every screen below is the actual product; the recovery and set-rate figures are published industry benchmarks (DemandLocal and Impel) modeled onto the demonstration store, not a customer's audited results.

The Problem

The phone rings at 7pm. Voicemail answers. The chain shop books it.

Every missed call to a service department is a booked bay that never happens. Industry studies put the recoverable money from missed service calls near $853,000 a year per store, and most stores recover none of it: after-hours callers hit voicemail, declined services scroll away in yesterday's tickets, and nobody can see the money leaving.

The fix is not another receptionist. It is a service lane that answers its own phone, prices every call it catches, and shows the store the money it almost lost.

The Solution

What Agentify AI Built

One store, two workspaces, and a voice advisor on the phones: the desk runs the day, the garage runs the bays, and Gus catches every call the store would have missed.

1. Run the Lane

One toggle, two workspaces: a light Front Office for the desk and a dark Shop Floor for the garage, every control sized for a gloved hand. The advisor runs the whole day from a phone: check a car in, move it bay to bay, assign a technician, mark it picked up.

2. The 7pm Call

At 7:04 PM the store is dark and the phone rings anyway. Gus answers, quotes the brake job from server-priced services, checks real availability on the 20-minute slot grid, and books the bay. The appointment lands on tomorrow's schedule, flagged Booked by Gus.

3. The Recovery

The recovery queue replays the calls the store never picked up: 14 missed calls caught, called back within minutes, and priced, including $434 from one call before the doors opened. A follow-ups list then ranks every declined service by what a call tonight is worth.

Inside the Product

The Actual Interface

Captured from the live product as it runs today: every screen is the real interface, and every booking Gus makes is validated by the server against the same slot grid the board reads.

The Dealer Service Bay lane board: ten bays and six color-blocked lanes from Scheduled to Ready for Pickup, with the recovered ticker at the top — The Lane, Live · 30 repair orders, 10 bays

Six lanes from Scheduled to Ready for Pickup, run from a phone: check in, advance, assign, pick up. A Watch mode plays the rush on its own.

Gus, the voice service advisor, answering the after-hours line at 7:04 PM and booking an appointment on tomorrow's schedule — Gus, After Hours · Booked by Gus, server-validated

7:04 PM: Gus answers the dark store's phone and the booking lands on tomorrow's schedule against real slot availability.

The recovery queue: missed calls caught, called back within minutes, and priced, totaling $3,710 saved this week — The Recovery Queue · $3,710 saved this week

Calls caught before open and called back in five minutes, each one priced so the store can see what it almost lost.

The follow-ups list: declined services ranked by predicted recoverable value, $5,730 in total — The Follow-ups · $5,730 predicted recoverable

The money the customer already said no to: every declined service ranked by its predicted value if Gus calls tonight.

The two Dealer Service Bay workspaces side by side: the light Front Office desk view and the dark Shop Floor garage board — The Product · Two workspaces, one design system

The light Front Office and the dark Shop Floor share one design language, with every stage color passing WCAG AA contrast in both themes.

The after-hours call in progress: Gus quoting a brake job and booking a real appointment while the store is closed

The AI Moment

The store is dark. Gus books the bay.

At 7:04 PM a customer with squealing brakes calls a closed service department. Gus answers, quotes the job from server-priced services, reads real availability on the slot grid, and books the first open bay on tomorrow's schedule. The appointment is a real row in the store's database, validated by the same server the lane board reads, and the morning advisor finds the ticket already written.

$853,320

Per lane a year, industry benchmark

33%

Set rate lift, industry benchmark

The Voice Agent

Gus, Under the Hood

Gus runs on four tools validated by the store’s own server. Prices come from the service catalog, times come from the live slot grid, and every booking is re-checked at the moment it is written, so he cannot promise what the schedule cannot keep.

The Data He Reads

A priced catalog of ten common services, oil change through factory maintenance, each with a range and a duration.
Real openings on a twenty-minute slot grid, computed from bays, technicians, and business hours.
A customer file looked up by phone number: the vehicles on record, their mileage, and up to two previously declined services.
Store facts he holds without looking: hours, the shuttle, the lounge, and that open safety recalls are always free.
From the caller he collects the minimum needed to book or hand off: a name, a ten-digit callback number, and the vehicle.

The Answers He Gives

Price ranges rendered by the server, with “plus tax and shop supplies” on the first quote and “the exact number comes after inspection” after a range.
Up to three real appointment times, never a fourth, and never one a tool did not return.
Triage that sorts urgency: a steady check-engine light books the flat diagnostic; a flashing one gets “stop driving, tow it in tonight.”
A booking confirmed with its date and time, and “nothing due today.”
At most one gentle mention of a service the caller previously declined, then he drops it.

The Guardrails

No price, time, or availability he did not just receive from a tool in this call. No memory answers, no guessing.
No diagnosis over the phone, no warranty rulings, no financing. Off-menu work becomes “let me have Sandra call you first thing,” with a callback number captured.
The server re-validates the slot at booking time. If it was just taken, he apologizes and searches again rather than double-booking.
Never speaks, requests, or confirms an email address; a phone number is the only contact detail he handles.
Every row he writes is flagged as a demonstration booking, so the day resets without touching the store’s history.

The one thing he writes: a real appointment. It lands on the Tomorrow board flagged Booked by Gus, opens a scheduled repair order, and moves the recovered-revenue ticker. Everything else Gus does is read-only.

Try It Yourself

Five Calls Gus Gets Right

Gus can only quote prices his server returns and only offer times that are actually open on the schedule. Open the app, tap Ask Gus, and run these five calls.

1 The 7pm Booking Say: “My brakes are squealing and I need to get in tomorrow.” He asks the one clarifier that matters, squeal or grind, quotes brakes at $434 to $529 per axle plus tax and shop supplies, offers three real openings from tomorrow’s schedule, and books the one you pick. The appointment lands on the Tomorrow board flagged Booked by Gus. 2 The Menu Price Ask: “How much is an oil change?” Between $89 and $109, and he explains the difference: the higher number is the package with a tire rotation. About 45 minutes, and the express lane is quickest before nine. 3 The Triage Say: “My check engine light just came on.” One question: steady or flashing. Steady gets the flat $165 diagnostic that applies to the repair if approved. Flashing gets told to stop driving and tow it in tonight. 4 The Memory Say: “Can you look me up? Four six nine, five five five, zero two zero four.” He finds the caller on file: a 2019 ATS at about 74,100 miles, and gently mentions the front struts she passed on last month. Once, then he drops it. 5 The Honest No Ask: “My transmission is slipping. What would that cost?” No made-up number. Transmission work is off his menu, so he offers to have Sandra call first thing in the morning and takes a ten-digit callback number instead.

Open Dealer Service Bay and ask Gus

Every booking Gus makes is a real row in the store’s database, flagged as a demonstration booking so the day can be reset. Prices and times come from the server, never from the model.

The Film

Sterling Motorcars, the Film

A day in the service lane, filmed like a commercial: the morning walk-in, the front counter, the car coming down off the lift.

Poster frame of the Sterling Motorcars film: a small dealership service lane at the start of the day

Tap to play · produced by Agentify AI for the demonstration store

The Outcomes

What Changed

Repair Orders, One Living Day

Ten bays, six color-blocked lanes, and thirty repair orders move through one day at the demonstration store, with the recovered ticker running at the top.

Missed Calls, Caught and Priced

Every call the store never picked up is caught, called back within minutes, and priced: $3,710 saved this week on the recovery queue.

$434

From One After-Hours Call

One caller with squealing brakes reached Gus instead of voicemail, and the booking was on the schedule before the doors opened.

The demo ticker closes at $2,735 a day. Annualized, that models the industry benchmark: $853,320 per lane.