Engineering for US operators

Operate cleaner.
Grow smarter.

Custom software, integrations, and human-reviewed AI for distributors, field service teams, multi-site clinics, and job shops.

Get a free workflow snapshot

What we build

Four levers. One commercial goal.

Faster quotes, fewer manual handoffs, and cleaner margin visibility across the systems you already run. Compare all services →

Vertical deep-dives

Where off-the-shelf AI breaks

Four operational bottlenecks where unreviewed chatbots guess wrong — and how we wire structured data, ERP lookups, and human sign-off around them. All industries · Workflow blueprints

Inside sales coordinator at a dual-monitor warehouse desk cross-checking Outlook RFQs against Excel pricing 01

Specialty industrial distributor

Quotes locked inside senior reps' heads

A veteran counter rep prices a 60-line contractor quote in 20 minutes. A newer rep spends half a day cross-checking PDFs and still misses discontinued part interchanges, LTL freight minimums, and Account 1847's negotiated multiplier.

Why AI fails

Out-of-the-box chatbots can summarize a catalog sheet, but they cannot reconcile ten years of customer rebates split across Outlook threads, Prophet 21 free-text notes, and branch spreadsheets — or tell which manufacturer supersession requires an adapter kit.

Oruvyn move

We build an inbox-to-ERP quoting queue: incoming PDFs, spreadsheets, and emails are parsed into line items, matched against a read-only ERP replica with confidence scores, and staged in a single review screen where reps approve low-confidence substitutions before pushing to the ERP.

First-draft quotes drop from 4+ hours to under 25 minutes, wrong-part returns fall sharply, and new counter hires quote accurately in weeks instead of years.

Two dental front-desk coordinators in gray scrubs reviewing dual monitors and insurance paperwork 02

Multi-location dental group

Insurance math no chatbot can finish

Front-desk coordinators spend 15–25 minutes per patient untangling remaining annual maximums, crown waiting periods, posterior composite downgrades, and dual-coverage rules before quoting an out-of-pocket co-pay.

Why AI fails

Clearinghouse 271 responses routinely conflict with carrier faxback sheets, plan years reset mid-treatment, and CDT codes map differently per payer. Asking a raw LLM to 'read the benefits PDF' produces confident co-pay guesses that blow up at checkout.

Oruvyn move

We connect Open Dental, Dentrix, or Eaglesoft schedules to clearinghouse feeds and payer breakdown history, compute deterministic patient-portion estimates, flag missing clauses in yellow, and hand front-desk staff a one-page co-pay worksheet 48 hours before chair time.

Pre-visit verification time drops by more than half, surprise balance write-offs shrink, and front-desk coordinators stop burning out on hold with payers.

Service dispatcher on headset coordinating technicians around clipboards and a magnetic schedule board 03

Residential & light-commercial HVAC

Dispatch boards that scramble by 11 a.m.

An 8 a.m. no-cool call turns into a condemned heat exchanger by 10:30, a van missing a universal TXV, a 10-year parts warranty lookup, and four afternoon homeowners waiting on updated arrival windows.

Why AI fails

Standard route optimizers minimize drive minutes while ignoring who is EPA-certified for R-454B, which truck actually stocked capacitors after yesterday's surge, or whether an aging install qualifies for manufacturer goodwill vs. replacement turnover.

Oruvyn move

We layer a live dispatch decision view onto ServiceTitan, Housecall Pro, or FieldEdge — combining call tags, equipment age, tech certifications, and truck stock — while automated helpers draft customer ETA texts and supply-house pick lists for dispatcher approval.

Fewer dead-end truck rolls for missing parts, higher single-visit completion during summer peaks, and a board that runs cleanly even when the lead dispatcher is off.

Fabrication estimator measuring a machined metal part with digital calipers beside engineering blueprints 04

Custom fabrication job shop

Estimating from half-dimensioned prints

A buyer emails a marked-up PDF print, two smartphone photos, and 'quote 25 pcs in 304 stainless by Friday.' Your senior estimator has to price laser time, brake setups, weld distortion scrap, and outside anodizing without underbidding.

Why AI fails

Vision models can read title blocks, yet they have zero grasp of your press-brake tooling changeovers, current grain-finish bottlenecks, plate nesting yield, or which repeat accounts submit unpaid rev-B changes mid-run.

Oruvyn move

We build a structured estimating workbench: incoming drawings extract BOM dimensions and tolerances, pull comparable historical jobs from JobBOSS or E2, enforce required risk checkboxes, and convert won quotes straight into traveler packets for the shop floor.

Quote turnaround speeds up without gut-feel underbids, actual-vs-estimated margin tightens, and shop travelers match what sales actually priced.

How we build

Shadow. Scope. Ship.

Full engineering method & pricing →
01 Shadow the real desk We inspect actual RFQs, dispatch boards, spreadsheets, and ERP fields from a busy Tuesday before proposing architecture.
02 Draw the sign-off line Software parses, matches, and drafts; your people approve substitutions, pricing overrides, and customer commitments.
03 Ship a 30-day pilot One production workflow wired into your existing tools, measured on turnaround time and error rate before phase two.
Operations engineer on laptop beside industrial wiring assembly
Assembly & floor
Commercial service van stocked for field calls
Field service
Machinist checking tolerances on the shop floor
Fabrication shop
Fulfillment and order operations desk
Order desk

Free workflow snapshot

Send one workflow that is burning hours or leaking margin. Within 2 business days, Caleb replies with a 1-page teardown of what to fix first and whether AI belongs anywhere near it.

  • Ranked bottleneck map: which steps lose hours or margin vs. which ones to leave alone
  • Architecture call: off-the-shelf connector, custom web tool, process change, or AI review queue
  • 30-day pilot scope: concrete systems touched, human sign-off gates, and rough build band
Turnaround: 2 business days Cost: $0, no retainer pitch

Rather send two messy spreadsheets or sample RFQs straight over? Email jarvis@oruvyn.com.

Step 1 of 3 · Your company

Company details

Who builds the work

Most 15-to-200 person companies do not fail at AI because they waited too long. They get burned because the work that moves margin lives half in their ERP, half in shared inboxes, and half in one senior employee's memory.

When you request a Snapshot or scope a build with Oruvyn, you work directly with an engineer who reads your real spreadsheets, quotes, and dispatch logs before writing a line of code. Sometimes the fix is an AI-assisted review queue. Sometimes it is a tight internal web app or two clean API integrations with zero model calls.

Caleb Njiiri · Founder, Oruvyn · jarvis@oruvyn.com · How we work & ownership terms

Caleb Njiiri, founder of Oruvyn Caleb Njiiri · Founder, Oruvyn

FAQs

What kind of company is Oruvyn?

Oruvyn is an operational software and AI engineering partner for US small and mid-sized businesses. Depending on the bottleneck, we ship custom internal web apps, ERP/CRM integrations, deterministic workflow automation, or human-reviewed AI queues.

Who is the Free Workflow Snapshot built for?

Owners, COOs, and ops leads at 10-to-250 person US companies where quoting, dispatch, eligibility checks, or order entry is eating hours or leaking margin. You send one workflow; Caleb replies within 2 business days with a 1-page written teardown.

What if AI is the wrong tool for our problem?

We tell you plainly in the Snapshot. Roughly half of the highest-return fixes we scope are two-way API integrations, database cleanup, or a focused internal screen with zero language-model calls.

Which industries do you work with most?

Specialty wholesale distributors, residential and commercial HVAC/plumbing/electrical contractors, multi-location dental and medical clinics, custom fabrication job shops, and B2B ecommerce operators.

What do engagements cost and how long does a first pilot take?

Every engagement starts with the free Snapshot. Focused 30-day pilots typically run $6,500–$15,000 fixed-scope (for example, an RFQ inbox parser + review queue). Full custom workbenches or multi-system integrations typically run $18,000–$45,000+ in milestone phases.

Who owns the code, prompts, and data?

You do, 100%. Upon milestone payment you own the Git repository, database schemas, infrastructure configs, and prompt templates outright — no recurring license fees to Oruvyn and no vendor lock-in.

When is Zapier or Make enough vs. building custom software?

When you are syncing clean fields between two modern SaaS apps under ~2,000 runs/month and a failed run is low stakes, Zapier or Make wins on speed. When intake arrives as messy PDFs, pricing rules live in ERP tables, or staff need a review screen before money moves, a custom service pays back fast.

More context in resources or request a free AI snapshot.