The Lab.
Things we've actually built — live, usable, and measured. If we recommend something for your practice, a version of it has been grown here first. Step in. Touch everything.
Lab 004 · Live now · Newest exhibit
The Workbench.
Five AI tools that research your business the moment you enter — and remember you when you come back: a property scout with live web search, a profit leak detector that sizes your hidden revenue in dollars, a business coach, a marketing strategist, and a lead machine. Plus an AI receptionist answering a real phone line. Everything you make is emailed to you as a designed report.
Sit down at the benchAI Receptionist — call it live: (425) 494-4489
Deal Scout — live property deal sourcing
Profit Leak Detector — your hidden revenue, in dollars
Business Coach — direct answers, your context
Marketing Strategist — strategy first, then the copy
Lead Machine — who to reach and how
Case Studies
Each card says what was built and what was measured. Where a number isn’t in yet, it isn’t on the card.
~90
potential patients reached every month
A voice agent that books straight into the schedule
Challenge
After-hours and overflow calls were going to voicemail, and the patients on them were walking. Nobody could say how many, because a voicemail that never becomes a callback leaves no trace.
Solution
Mapped where patients were being lost, then designed and deployed a voice agent that answers every call, captures who is calling and when they want to be seen, and books straight into the schedule — handing off to staff when a caller needs a person.
Results
- Over 700 calls answered so far since going live in January 2026 — about 90 a month, each one a potential patient the front desk would otherwise have missed
- 5 appointments booked in the first 2 days live — $4,300+ in scheduled visit value
- Every appointment request lands in the front desk’s inbox automatically, with the transcript, no callback queue
- Answering live patient calls every day, with no staff added
745
of his bids, read end to end
A bid writer built from 745 of his own bids
Challenge
Every bid was typed by hand in Word: ten to fifteen minutes for a roof, up to an hour for the complicated ones, five to eight a week. ChatGPT and Copilot had both broken his formatting, so he’d stopped booking jobs on the spot.
Solution
Read all 745 of his bids from 2023–2026, rebuilt his estimating on his own rate card, and installed it inside his own ChatGPT. It shows the arithmetic first, refuses to price anything it has no rate for, and hands back a Word bid in his house format.
Results
- Pricing checked against 80 of his own roof bids: 4.7% median error, no lean high or low
- Installed in a 20-minute walkthrough with him driving; the first bid built live from a two-line description
- “This was a very easy process. I’m impressed with your professionalism and I’m excited to use my new AI helper.”
3
systems, one client, one builder
One billing company, three systems, one accountable builder
Challenge
A growing billing company with teams in two countries had a website that didn’t sell, client wins nobody had captured on record, and a phone carrier that blocked their texting and offered no support when it did.
Solution
Started with the website. Then ran the client interviews ourselves — research-grade, on camera — and cut them into testimonials. Then audited the phone stack against five carriers and scoped the migration. Each piece paid for the next conversation.
Results
- A physician client’s interview became a video testimonial — and that practice then signed a credentialing contract with Emerald
- Off AT&T and onto Nextiva, signed September 2026: 9 numbers to 12 lines, texting unblocked, five carriers compared on the client’s own call data
- Reporting across 7–8 billing systems mapped next, so team leads stop spending 1–2 days a month per client assembling it by hand
- “You understood the process completely right… that’s the right way to do.” And on the interview: “You asked the right question… that was really well made.” — Senthil, operations
11
days from blank repo to live product
A daily log with an AI coach that reads it back
Challenge
The founder had 2,000+ days of his own daily tracking, and a coaching group whose members dropped their reports into a group chat, where every insight got buried by the next message. The first version of the app was lost with its no-code host.
Solution
Rebuilt from a blank repo on our own stack in September 2026: a 60-second nightly log, an AI coach that reads the last 30 days and answers in plain words, field reports, a life audit, a tap counter for the day, a roster view for coaches, calendar reminders, segment email, and billing.
Results
- Rebuilt and live in 11 days, with all 235 of the founder’s own entries imported on day one
- 17 beta testers in the first 11 days; 5 of them logged this week
- 60 AI coach reads delivered since the rebuild — 42 on daily logs, 18 on field reports
1 hr
of training to run her own site
A wellness site the owner now runs herself
Challenge
Thirty years of nutrition guidance and a product line, on a site she couldn’t touch. Every change went through someone else, so nothing changed.
Solution
Rebuilt the site on a platform she can edit herself: her story, products, a downloadable guide, a quiz with personalized results, a booking link at every contact point, blog and newsletter. Then trained her to publish her own updates.
Results
- Kicked off February 19, live on her own domain March 27, 2026
- One hour of training, one session — she has published her own pages and newsletter since
- Monthly maintenance retired on purpose — the site is hers to run
Hers
editable, transferable, owned by the client
A broken WordPress site replaced with one that says what she does
Challenge
A WordPress site that had broken, with copy that listed credentials instead of the reasons her clients actually come in: depression, anxiety, feeling stuck.
Solution
Rebuilt from scratch: symptom-led copy her clients recognize themselves in, a 15-minute consultation one tap from every page, a testimonials page, her military and first-responder work, local SEO for Navarre — plus team access so she and her assistant can edit it and read the analytics themselves.
Results
- Kicked off March 4, 2026; first revision round shipped in May with team access handed over
- Google Business Profile and local search pointed at the new site
- Ownership moving into her own account — the way every build we do is meant to end
- “I love the upgrades, Thank You So Much!!!” — Joni, on the May revision
Lab 001 · Live now
What is voicemail costing you?
Put your own numbers in. The result is the annual revenue currently leaking through unanswered calls.
Your numbers
Uncaptured opportunity
30 missed inquiries / mo
Revenue leakage (monthly)
$45,000
Annual impact
$540,000
*Preliminary calculation from your inputs. An assessment refines these variables for your practice.
Walk the other exhibits.
Lab 002 · Live
The Museum of Sound
An immersive walk from the first sounds to the language of AI — built as an experience, not a page. This is what we make when we explore for its own sake.
Enter the museumLab 003 · You're inside it
This website
Museum-grade design system, classical architecture as structural language, every page a different room of one building — designed, written, and shipped with the same AI-assisted system we install for clients.
Exhibit ongoing. You're in it now.We run our business on what we sell.
Every system below is in production — not for a client, for us. Nothing here is a mockup, and every claim is literally true.
The discovery pipeline
Book a call and you experience the product: the conversation is ingested by our system and your AI Opportunity Snapshot — your #1 opportunity, sized in dollars — arrives in writing the next day.
Experience it: book the 20-minute call.
The second-mind CRM
An Obsidian + Claude system that reads our meetings and email, updates 90+ relationship records, flags who to follow up with, and keeps every promise dated. It runs this entire business, daily.
Ask to see it live on a call.
VM Voice, in production
The same voice agents we sell, answering for real practices right now — designed conversation by conversation.
Read the voice-agent case study above.The AI-training kit
A live training cohort built from a tested session structure — starter templates, first prompts, and a build-it-in-session format.
Cohorts run monthly.
Three paintings, and why the work matters.
These used to hang on the sales pages. They live here now, on purpose: the craft stays on every page; the wall text belongs in the Lab.
The point was never the technology.
A Sunday afternoon on an island in the Seine. Ordinary people doing nothing in particular — and a painter free to spend two years rendering them, dot by dot, because someone decided that was worth the time. It hangs in Chicago still.
That’s the world worth building toward: people whose hours belong to them, work made freely, and the good of it kept for whoever comes next. Automation is a means. The end is the time it hands back.
A Sunday on La Grande Jatte — 1884
Georges Seurat, 1884–86
Art Institute of Chicago · Public domain
Panini painted the whole room at once.
Every monument in Rome, gathered onto four walls so a visitor could stand in one place and see how it all related. Nobody could get that view by walking the city. Someone had to assemble it deliberately, and then you needed a guide who had already spent years in the room.
A project gives you one monument. The ongoing seat gives you the room — and someone standing in it with you every week, who remembers what you decided last quarter and why.
Ancient Rome
Giovanni Paolo Panini, 1757
The Metropolitan Museum of Art · Public domain
Everyone here is making something.
In 1880 Kyōsai painted a calligraphy and painting party: a room full of artists working side by side, watching each other, borrowing what worked. Nobody is spectating. That is the whole difference between a room and an audience.
Expansion OS is eight seats, and you watch seven other businesses get built alongside yours. Half the education is seeing someone else solve a problem you haven’t hit yet.
Calligraphy and Painting Party
Kawanabe Kyōsai, 1880
Cleveland Museum of Art · Public domain
Growing now.
Labs in progress. Each one goes live on this page when it's real enough to use — not before.
AI Readiness Score
An 8-question assessment that scores how prepared your practice is for AI — and exactly where the gaps are.
In development
Call the demo line
A live voice agent you can phone and try to stump — the AI receptionist answers at (425) 494-4489, linked from The Workbench.
Live now
3D equipment teardowns
Scroll-driven, interactive breakdowns of real clinical equipment — see every component of a machine as you move down the page.
Queued
Learn: AI from zero
A leveled AI-literacy track — three short courses from "what is AI" to prompting as a reusable business asset.
Recording
Guardrails, not just code.
Every lab and every client system runs inside the same discipline: find the loss, map it, build it, measure it — with the guardrails below as non-negotiables.
Standardized tone
Agents that sound like your best receptionist.
Data discipline
Your data never trains public models.
Human in the loop
Complex queries escalate to people.
Drift control
Continuous monitoring for alignment.
"If we recommend it, we've run it ourselves first."