Gable
Request a demo
Pre-seed · Washington, DC

The AI operating system for the blue-collar economy.

Diagnose a homeowner's repair problem. Price it. Hand the job to a contractor whose entire business runs on our software. The byproduct is the thing nobody else has: the trades' ground-truth dataset.

Read the deckRequest a demo
  • Shipped

    Homeowner assistant

  • Built

    Contractor OS

  • The destination

    Trade foundation model

Mission

The trades are losing their knowledge. We are the ones writing it down.

Plumbing, electrical, HVAC, carpentry, roofing, masonry — the real expertise lives in people's hands and heads, passed mentor to apprentice over years. It was never written down, and it is retiring faster than it is replaced.

  • Advocate for the trades

    This work is skilled, essential and well paid — and a generation was quietly steered away from it. Every surface we build treats a tradesperson as an expert, because reaching the next generation starts with how the work is talked about.

  • Teach the next generation

    The knowledge base is a curriculum. An apprentice gets a mentor available at 7am on an unfamiliar unit; a vocational school gets teaching material grounded in real jobs instead of a 1998 textbook.

  • Give the craft its hours back

    AI should amplify human expertise, not replace it. Automating a contractor's paperwork returns those hours to the judgment and craftsmanship only an experienced tradesperson has.

The most valuable version of Gable is not a better way to find a plumber. It is the system that keeps the trade's knowledge from walking out the door — and puts it in everyone's hands.

Who this is for

Homeowners · contractors · apprentices and new techs · vocational schools · property managers and insurers · manufacturers — and anyone curious how buildings work.

The insight

Home repair is not a marketplace problem. It is an intelligence problem.

Homeowners aren't looking for a contractor first — they're looking for confidence.

  1. 01

    What's wrong?

    Homeowners can't describe their own problem, so they call the wrong person. Roughly half of our own contractor calls ended in “we don't do that”, “we don't serve your area”, or “our minimum is much larger”.

  2. 02

    Can I fix it myself? What should it cost?

    Everyone else starts at the hiring moment. We start one step earlier — where trust is actually won, and before any existing platform gets involved.

  3. 03

    Who is actually qualified?

    A brain that knows what a job is, what it costs and how long it takes can be pointed at either side of the market. Only then is matching a solved problem.

The three phases

One brain, built in the order that can't be reversed.

The brain takes years. The workflow around it takes quarters. We built the brain first, on purpose — owning both sides of the market isn't the moat, the ordering is.

Phase oneDemandShipped & live

The homeowner assistant

Describe it in plain language, or send a photo. We own the whole decision, not a slice of it — diagnose, price, decide, hand off.

  • Diagnose

    Hybrid retrieval over a curated trade knowledge base, or a photo of the fault. Nobody has to pick a category.

  • Price

    A deterministic cost engine — itemized labor, materials and time, by location and size. Never a figure the engine didn't produce.

  • Decide

    DIY or hire, honestly: steps, tools, materials, difficulty and safety warnings.

  • Hand off

    A license-verified local pro, and a quote request that arrives already scoped — diagnosis, size, photos, estimate and contact travelling together instead of a name and a phone number.

And what it refuses to do: off-catalog jobs are called a “rough AI estimate” — in those words — rather than quietly invented. Contractors only pay for a lead once they accept it.

Phase twoSupplyBuilt

The contractor OS

Small trade businesses don't lose money for lack of leads. They lose it on unanswered phones, slow lead response, estimates written at 9pm, and 45-day collections.

  • Front desk

    Answers as the business, qualifies against trade, area and minimum job value, books the real calendar. Emergencies page a human.

  • Shared inbox

    Every message triaged to an intent, a reply drafted in the business's voice, filed under the right job.

  • Scheduling

    Slots that fit working hours and how long the job actually takes. Double-booking is a database constraint, not a hope.

  • Estimates

    The contractor's own pricebook first, our cost catalog for the gaps — and every line records which it came from.

  • Invoicing & payouts

    Deposits, progress bills and finals, hosted payment links, funds settling to the contractor via Stripe Connect.

  • Field copilot

    The same trade brain in the tech's pocket: the fault, the procedure, the code clearance, the part.

Retention inverts here. A lead-gen relationship is switchable on a whim; a system of record is not.

Phase threeThe destinationWhere this goes

The trade foundation model

An AI that knows everything in the trades. Anyone can read every manual. Nobody else can see what a plumber in Ohio actually charged, and how long it actually took.

  • Level 0 — Retrieval

    A curated corpus and hybrid RAG. Shipped.

  • Level 1 — Knowledge graph

    Entities, not documents. Next.

  • Level 2 — Specialist models

    Post-trained per task. The seam exists.

  • Level 3 — Ground-truth flywheel

    Labels from real jobs: the contractor's edits to our price, actual durations against predicted, what the tech found against what we diagnosed remotely, and the final invoice against the estimate — the cleanest label in the system. This is the bottleneck, and it's a product problem, not an ML problem.

  • Level 4 — The domain model

    Proven by a benchmark. The achievable version isn't a $100M pre-training run — it's the model with the best trade knowledge, judgment and ground truth, because it owns data that only exists where the work happens.

Only the contractor OS produces that data, generated as a byproduct of the contractor getting value — which is why we built the OS before we needed it.

Why now

The knowledge is walking out the door in exactly the decade we gained the ability to capture it.

Experienced tradespeople are retiring faster than apprentices replace them, and most of what they know was never written down. Models only just became good enough to read a symptom, a manual and a photograph.

  • $600B+

    US home services annually — where the assistant and the lead fee sit.

  • Trillions

    The construction, maintenance, repair and skilled-trades economy the model addresses.

  • 1–10 trucks

    The size of the overwhelming majority of trade businesses — too small for enterprise field-service software, and our beachhead.

Each phase opens its own market: contractor software and payments (recurring, higher-margin, a system of record nobody leaves); insurance and property intelligence (the same photo-to-cost judgment, made thousands of times a year); and workforce training, where the knowledge base is a curriculum that reaches people before they pick their tools.

Team

Two founders, married, who met at George Mason 13 years ago.

We had this problem ourselves, repeatedly. The product exists because we couldn't answer the first question — so the domain knowledge is first-hand, not researched.

  • Miao Zheng

    CEO

    Product, UX, customer discovery, growth

    • Lead Data Scientist at Booz Allen Hamilton — eight years building decision systems over complex data.
    • MS Data Analytics Engineering; MBA, Johns Hopkins.

  • Ray Han

    CTO

    Architecture and the majority of the code

    • Ten years at Workday, Salesforce and Capital One.
    • MS Computer Science; MBA, Johns Hopkins.
    • Runs his own renovations — Home Depot Pro Xtra status through DIY.
The deck

Everything above, in eleven slides.

No email gate — read it, and write to us if it lands. If you'd rather be walked through it, ask for a demo and we'll do it on a real job instead.

Open the deckDownload PDF

What's inside

  1. 02The problem — why half the calls end in “we don't do that”
  2. 03The insight — an intelligence problem, not a marketplace one
  3. 04Product one · demand — the homeowner assistant
  4. 05Product two · supply — the contractor OS
  5. 06Market — and why now
  6. 07Competition — who owns which piece
  7. 08Business model — the success fee and the subscription
  8. 09Where this goes — levels 0 to 4
  9. 10Mission — writing the trades' knowledge down
  10. 11Team — the two of us
Talk to us

Request a demo, or just say hello.

Both founders read everything that comes in, and both answer it. If you'd rather see it than read about it, ask for a walkthrough on a real job — a real symptom, a real price, a real hand-off.

What we're looking for

  • Design partners

    Contractors willing to run the OS on real calls, real estimates and real invoices.

  • Supply density

    Enough verified pros in a market to answer every lead a homeowner sends.

  • The corpus

    Licensed codes, the knowledge graph, and the benchmark that proves the model.

  • Both founders full-time

    The round that takes us there. We're raising pre-seed now.

What's this about?

See the homeowner assistant and the contractor OS on a real job.