AI diagnostics that return the exact verified OEM part — not a guess.
Company overview, full product walkthrough & technology brief — for partners, the trade, and investors. Built by a service-industry owner for real field-service companies.
A field tech's whole day is a chain of judgment calls — what failed, which exact part fixes it, what to quote. Get the part wrong and the job doesn't just slip: it doubles. A second truck roll, a callback, an apologetic phone call, and a day of margin gone. The metric that captures all of it is first-time fix rate (FTFR) — and a single wrong-part trip drags it down hard.
Wrong part = drive back to the supply house, then drive back to the customer. Two extra trips against one billable visit.
A misdiagnosis means an unhappy customer, a callback, and the risk of a one-star review that costs far more than the job.
Senior-tech knowledge lives in a few heads. Newer techs guess, restock wrong, and the shop eats the difference job after job.
The core gap: raw AI can name a likely component, but it cannot reliably hand a tech the exact part number for the exact model in front of them — and a part number that's almost right is worth nothing at the supply counter. That last mile, from "probably the drain pump" to "this specific verified OEM part fits this specific model," is exactly where the money is lost — and exactly where ServiceLens lives.
A general model can tell you "it's the drain pump." Ask it for the part number for the exact model in front of you and it confidently invents one — roughly 0% right on the exact OEM part. ServiceLens returns the exact verified OEM part for the exact model — so techs never order the wrong part, and never drive back.
Plausible-sounding, frequently wrong. No tie to the actual model. No fault-code spec. No test step. The tech finds out at the supply house — or worse, at the customer's home.
Every answer is grounded in a verified corpus and dual-checked against that model. You also get the fault-code → cause → test-spec, so the diagnosis is provable before anyone drives.
Every screen that follows is the real, shipping product — not a mockup. We'll walk the whole loop a tech actually runs: from the cockpit, to the diagnosis, to the money shot (the part-first result), to the estimate, the saved-job flywheel, Ask Moe, and the connected office. For each screen we call out the feature — and why it beats the API-wrapper competition.
mobile-first + desktop dispatch
scan → symptom / code
part-first & verified
auto quote
the flywheel
chat + diagrams
CRM & integrations
One home, two postures. On the phone it's a guided, thumb-reachable cockpit a tech opens at the curb. On desktop it's a dispatch board for the office — same data, same verified knowledge base, no second login.
Desktop "Home" — dispatch cockpit: today's activity, the live knowledge base, and recent diagnoses with confidence.
On a phone the cockpit collapses to a thumb-first layout with a bottom tab bar — Home, Diagnose, Scan, Jobs, Menu — so the most-used actions are always one reach away. The dark navy theme isn't just brand; it's battery-friendly for an all-day device and easy to read in a dim basement or a bright driveway.

Mobile "Home" — the field cockpit, thumb-first with a bottom tab bar.
Diagnosis starts where the model lives: the nameplate. Snap the plate to capture the model, pick the trade, drop in a fault code (the fastest path) or describe the symptom in plain English — then run a free prescreen or go straight to the full breakdown.
"Full Diagnosis" — trade, model (with nameplate scan), fault-code fast path, symptom description, prescreen vs full.
The same trade selector, scan, code field, and symptom box collapse cleanly onto a phone — so a tech standing at the unit runs the exact diagnosis the office would, without learning a different screen. Voice is deliberately absent: a tech tapping a quiet screen looks professional in a customer's home, where a voice assistant would not.

Mobile "Full Diagnosis" — the same flow at the unit, silent and thumb-first.
This is the screen the whole product exists to produce. We don't bury the part under paragraphs of prose — we lead with it. The first thing a tech sees is the exact OEM part to order, badged Verified for this model, with the cause, the test, the priced parts list, the estimate, and the close-the-loop write-up all underneath. This is the moat, on screen.
"Diagnosis Report" — part-first: the verified part to order leads, then cause, priced parts, estimate, and the logged outcome.
On the phone, that same part-first report stacks into a single scroll: the verified part and Order part first, then cause, the priced parts list, the auto-estimate ($95.00 · Build estimate), the full write-up, and the outcome-logged card. One screen carries the job from "what's wrong" to "ordered, quoted, and closed."
No scrolling past prose to find the number — it's the first thing on screen.
Order part · Build estimate · Download PDF · Ask Moe — all from the report.

Mobile "Diagnosis Report" — the entire part-first job in one scroll.
Diagnosis flows straight into money. From "what you found," ServiceLens builds a priced estimate — parts (with OEM pricing grounded in the verified corpus), labor, and tax — that becomes a quote you can share or push to your CRM. The numbers a shop charges most are pre-filled from saved defaults, so a quote is seconds, not paperwork.
"Build the estimate" — job details, parts, labor & tax (pre-filled from saved defaults), then Compute estimate.
The economic loop closes here. A tech walks in, diagnoses, prices, and quotes — without a back-office round trip. One avoided second-trip pays for the whole seat, many times over.
Every diagnosis is saved, scored, and re-openable — a searchable library of the shop's work. But it's more than a log: when a tech closes a job and records what actually fixed it, that confirmed fix re-grounds the next diagnosis. The corpus gets smarter with every job, across the whole platform.
"Diagnosis History" — every job saved, scored, and re-openable, with Download PDF and Build estimate on each.
part-first, grounded & verified
the verified OEM part, first try
what actually fixed it, rated
confirmed fix enters memory
grounded in a real, confirmed fix
Sometimes a tech doesn't want a report — they want to talk it through. "Ask Moe" is a conversational senior-tech assistant that doesn't just answer in prose: it draws the decision tree. Moe renders flow and test diagrams so a tech can follow the call branch by branch — silently, on screen, no voice required.
"Moe the Tech" — walks a "fridge not cooling, fan silent" call and renders the decision tree as a flowchart.
ServiceLens is the diagnostic brain — and it plugs into the rest of the shop's stack. Settings is where a shop sets estimate defaults, manages the subscription, invites techs, and connects the CRM and field-service platforms it already runs. Solo techs get a built-in CRM; teams push jobs straight into the tools they live in.
"Account" — preferences, estimate defaults, CRM integrations, subscription, and invite-a-tech, all in one place.
Not a GPT wrapper. ServiceLens is a proprietary system from the Sophia XT AI lab. Compositional Routing is the production grounding engine today: it routes a query to the exact verified data across multiple corpora, then dual-verifies — input and output — every recommended part against that model's corpus before it ever reaches a tech.
Both the inbound model identity and the outbound part are checked against the corpus. Each part is flagged Verified (corpus-confirmed for this model) or Confirm-fit — killing hallucinated part numbers at the source.
Proprietary NCN routing + LCM architecture is the Sophia XT lab's research direction extending the engine — the foundation we're building toward, not a live production claim. Compositional Routing is what grounds answers today.
Works across leading model APIs. The defensibility is the verified corpus and the routing + verification layer — not any single vendor's model, so we ride every model improvement without being captive to one.
Most "AI for trades" tools are a prompt around a general model. That's easy to build — and easy to copy. The hard, slow, expensive thing is a verified, exact-model parts-and-codes corpus per trade. That's the part competitors don't have, can't quickly clone, and the part that makes the answer right. Each new trade we finish widens the moat instead of just adding a feature.
Thin or no parts list. No model-level verification. Hallucinates part numbers. Forgets the job when the chat closes. Fast to ship — and just as fast for the next team to copy.
Millions of verified OEM parts mapped to exact models, codes mapped to cause and test, dual-verified per answer, and a flywheel that compounds with every closed job. The data is the defensibility.
Verified, exact-model parts-and-codes data is slow, unglamorous, and expensive to assemble. That difficulty is the moat — it's why competitors run a thin list or none at all.
We don't trust the model's first answer — we check it against the corpus and badge it. A part labeled "Verified" means corpus-confirmed for that exact model, not "the AI sounded sure."
Every closed job re-grounds the next. A wrapper is a snapshot; ServiceLens is a system that gets more accurate the more it's used — a gap that widens over time.
We're honest about the map: appliance is live, deep, and exact-model-verified today. Other trades are expanding — each at its own stage. This is a roadmap, not a completed claim. The strategy is to go deep before going wide, then repeat the playbook trade by trade.
The expansion playbook: prove the depth in appliance, then port the same corpus-and-verification machine to the next trade. Because the engine is trade-agnostic, each new trade is a data project, not a rebuild — and every finished trade widens the moat rather than just adding a feature.
These are precise figures from an internal benchmark that anyone on the team can re-run — not rounded marketing numbers. We report the strict floor alongside the best case, and we keep accuracy and confidence as separate measures on purpose.
The conservative lower bound — exact component match under strict scoring.
The everyday calls a tech actually makes resolve at the top of the range.
The model knows what it knows — confidence is calibrated, so techs can trust the flag.
Read it straight: 81% is the strict floor, not a ceiling; 100% is on common faults specifically; 86% is calibrated confidence, a separate measure from accuracy. We keep these distinct on purpose — conflating them (writing, say, "81% confidence") would be exactly the kind of overclaim this product is built to eliminate.
Comparative and defensible — we don't claim to beat everyone at everything. We claim a specific, field-tech-first position: exact-part grounding, no voice, a large verified-OEM-parts corpus, and a self-improving flywheel, priced for SMB.
| Capability | ServiceLens | Aquant | Aiventic | FSM (HCP / Jobber / ServiceTitan) | Manufacturer tools |
|---|---|---|---|---|---|
| Exact verified OEM-part grounding | Core moat — verified vs confirm-fit | Industrial focus | Leaner corpus | Not a diagnostic engine | One brand only |
| Target user | Field tech & SMB shop | Enterprise / industrial & medical fleets | Field service | Dispatch & invoicing | That brand's techs |
| Interaction model | Silent text + diagrams | Complex enterprise UX | Voice-heavy | Forms / scheduling | Portal / lookup |
| Office-to-tech prescreen | Yes — part before dispatch | Varies | — | — | — |
| Self-improving job flywheel | Yes — confirmed fixes re-index | Knowledge mining | — | — | — |
| Auto-estimate & quote | Built in | — | Limited | Strong | — |
| Pushes into FSM / CRM | HCP + Jobber wired | Integrations | Some | is the FSM | — |
| Brand / trade breadth | Brand-agnostic, multi-trade | Industrial verticals | Field-service | Trade-agnostic ops | Single brand |
| Positioning | Affordable · field-first | Enterprise · premium | Mid-market | Operations backbone | OEM support |
Comparisons reflect each vendor's primary public positioning; capabilities evolve. ServiceLens is complementary to FSM platforms — the AI diagnostic brain they lack, that pushes work into them.
Defensible, comparative positioning — not absolute "we beat everyone." Against each category we hold a specific, ownable advantage.
Aquant targets enterprise, expensive, complex industrial and medical fleets. ServiceLens is affordable and field-tech-first — appliance/HVAC/SMB, silent with diagrams, and a stronger office-to-tech prescreen that gets the part right before the truck rolls.
Aiventic leans voice-heavy on a leaner corpus. ServiceLens is no-voice, with a larger verified-OEM-parts corpus, exact-part grounding, a self-improving flywheel, and built-in auto-estimate — the depth and the close-the-loop, not just the conversation.
These run scheduling, dispatch, and invoicing — not diagnosis. ServiceLens is the complementary AI diagnostic brain they lack, and it pushes work directly into them (HCP + Jobber wired today; ServiceTitan is an FSM we push into, not a live integration claim).
OEM tools are locked to one brand. ServiceLens is brand-agnostic and far broader — one tool across makes, models, and trades, instead of a different portal per manufacturer.
A diagnostic tool lives or dies on whether a tech will actually use it on the job. Every product decision here is a field decision — made by someone who has run the truck.
Built phone-first for an all-day device — the dark navy theme is easy on the eyes in a dim basement and easier on the battery in a long shift.
Deliberate, not missing. A tech tapping a quiet screen looks professional in a customer's home; a voice assistant would not. Text + diagrams keep the tech in control.
Conversational senior-tech mentorship that draws the decision tree — Mermaid-style flow and test diagrams a tech can follow branch by branch.
The office gets a full dispatch view of the same data — pre-diagnose, assign, and track jobs without a separate tool or a second login.
The free prescreen lets the office pre-diagnose and recommend the part before dispatch — so the truck rolls stocked right the first time.
EN · ES · FR · DE in the interface — built for the real, multilingual makeup of a modern field-service crew.
No enterprise procurement, no per-feature upsell maze. One flat price per tech, a free trial to prove it, and a value math that's hard to argue with: avoid a single wrong-part second-trip and the seat is paid for many times over.
Everything in the walkthrough — diagnose, the verified part, auto-estimate, history & flywheel, Ask Moe, and connected CRM. One price.
Priced for the SMB field-service shop, not enterprise procurement. Scale is just seats.
A real diagnostic engine — grounded in a verified OEM corpus, dual-checked per model, and getting smarter with every closed job — built by someone who has actually run the truck.
Exact verified OEM part for the exact model.
Dual verification flags Verified vs Confirm-fit.
Every closed job re-indexes into memory.
$34.99/tech/mo · silent · mobile · CRM-wired.
ServiceLens by Sophia XT — a proprietary system from an AI research lab. $34.99 per tech / month. Never order the wrong part. Never drive back.