Case studies - Systems we have built for service businesses

Each one is listed by business type, not by name, until the client says otherwise.

Salon

110

availability tests passed before launch

A phone agent that books in English and French on the live Square calendar

The problem: Bookings came in by phone in English and French. Every call needed someone free to pick up, work out who could do the service, and find a slot in Square by hand.
What we built: A Vapi phone agent wired through Make.com to the salon’s Square calendar. It answers in either language, checks which staff can do the requested service, reads live availability and offers up to four same-day slots. When a request is unclear it asks instead of guessing, and on any error it hands the caller to a person.
Result: The agent takes booking calls day and night, in both languages, without anyone on the floor picking up the phone. Before launch its availability checks passed 110 automated tests, with operator phone checks on top.
Stack: Vapi, Make.com, Square

Glass-repair shop

Voice + text

calls, texts and picture messages answered by one receptionist

A voice and text receptionist on Twilio, scoped to a small budget

The problem: The shop needed its calls and texts answered, and every option it had looked at was outside its budget.
What we built: A Twilio receptionist that picks up voice calls and handles SMS and MMS, so customers can call or text, picture messages included. We scoped it as a production MVP so it fit the budget the shop already had.
Result: A working receptionist in production, inside the existing budget, with the client kept up to date on every deliverable.
Stack: Twilio Voice, Twilio SMS and MMS

“He was able to get us to an MVP within a reasonable timeframe and kept in contact with us on deliverables.”

Mike B., glass-repair company

HVAC company

Every lead

texted back automatically and routed in the CRM

Every web lead texted back and routed to the right person

The problem: Web leads waited for a reply until someone was off a job and free to call back, so follow-up depended on who remembered.
What we built: An automation that catches each web lead the moment it arrives, texts the lead back, and routes it to the right person in the CRM with the next step already set, so nobody on the crew has to stop work to answer a form.
Result: Every lead gets a reply while they are still thinking about the job, and the owner can see where each one sits in the CRM.
Stack: Workflow automation, SMS, CRM

“He grasped the technical requirements quickly, delivered clean results, and handled feedback professionally.”

Kieu N., HVAC company, Arizona

Plumbing company

Relaunched

the Retell callback agent, with its call flow rebuilt

A missed-call callback agent cleaned up and put back in service

The problem: The company already had a Retell voice agent calling back missed calls, but its conversation flow needed cleaning up before it could be trusted with real callers.
What we built: We cleaned up the Retell flow behind the missed-call callback, tested it, and relaunched it.
Result: The agent is back in service, calling missed callers back before they ring the next plumber.
Stack: Retell AI

Supply-chain operations team

emails sent without a person approving them

Supplier email checked for risk, with replies drafted for a person to send

The problem: Every supplier email had to be read, matched to its purchase order and judged for risk by hand, and the risky ones depended on someone spotting them.
What we built: A Make.com workflow with OpenAI inside. Each supplier email is matched to its purchase order, run through an AI risk check and logged to Google Sheets. For high-risk issues it drafts the reply in Gmail for a person to review and send. It recognises a replayed email, so a duplicate never triggers a second draft or a second log line.
Result: All 7 test cases passed before handover with zero emails sent automatically. The team reviews drafts instead of writing replies from scratch.
Stack: Make.com, OpenAI, Gmail, Google Sheets

“Outstanding to work with. His attention to detail, communication skills and perfectionist mentality made for a smooth project experience.”

Brian G., supply-chain operations

Advisory firm

1 click

from an Airtable record to an editable report

Fit scoring and a finished research report from one click in Airtable

The problem: Scoring a prospective client and writing up the research was a manual job for every record, and nothing stopped the same client being worked twice.
What we built: An Airtable, Make.com and Claude API signal tracker, then a paid add-on on top of it. One click scores the client’s fit, skips anything already processed, picks up to six research signals and writes an editable Google Doc. The link is logged back to the Airtable record.
Result: The end-to-end test logged zero new AI errors, and duplicates are skipped automatically instead of being caught by eye. The firm came back for a second build.
Stack: Airtable, Make.com, Claude API, Google Docs

“Patient, clear and professional throughout the process. Communication was good, he was responsive, and he was willing to go through the details with me rather than just rush the job.”

Advisory firm client

Sales and support teams

Every call

ends with a transcript, a score and coaching feedback

Internal build, not a client project.

Roleplay calls scored by an AI agent, with coaching notes for the manager

The problem: Roleplay practice needs someone to play the customer and someone to listen and score, so most reps get little of it.
What we built: The rep picks a scenario in a dashboard, calls a Twilio number and talks to an ElevenLabs agent playing the other side. When the call ends, the rep and the manager both get a transcript, a score and coaching feedback.
Result: A working build, demonstrated end to end. The same loop can score a live AI receptionist’s calls as part of the monthly Run plan.
Stack: Twilio Voice, ElevenLabs

Clinics

front-office jobs in one screen

Our own product, not a client project.

The clinic front office in one screen, from appointments to billing

The problem: Clinic front desks run appointments, patient records, reminders and billing in separate tools, and staff re-enter the same patient in each of them.
What we built: The front end of Cliniqbit, a clinic-management platform: appointment scheduling, patient records, automated reminders and billing in one interface, so the front desk works from a single screen.
Result: One place for the four jobs a clinic front desk does all day.
Stack: Web app

Book a call

A short call on how calls and leads reach your business today, where they slip, and whether a front-office system is worth building for you. If it is not, we will tell you.

Book a Call hello@lambyte.com

Rather show us first? Email a Loom of your current setup to hello@lambyte.com and we will reply with what we would fix first. Prefer a form? Send your details.