The truth is in four places
The estimate lives in one tool, the schedule in another, the invoice in a third, and the part everyone actually relies on lives in one person's head. Reconciling them is a weekly job nobody signed up for.
One system your whole operation runs on.
Centralux is a founder-led AI engineering practice. We replace the spreadsheet-and-group-chat sprawl behind quote-to-cash with production software you can audit — engineered so your data stays isolated, traceable, and yours.
- Single-tenant by default
- Every action leaves a record
- Flat fees — you own the code
The situation
The work is fine. The system underneath it isn't.
Most operations we meet aren't badly run. They're well run on top of tooling that was never designed to hold the whole job.
Every handoff is a person retyping
Approved estimate to work order. Completed job to invoice. Payment to the books. Each hop is manual, each one loses a detail, and the cost only shows up later as a slow collection or a job billed short.
You can't ask your own data a question
"What did we actually make on that job?" should take a sentence, not an afternoon of exports. When the answer is expensive to get, you stop asking — and you start running on feel.
What we do
Four services. One system underneath.
Each one is a scoped engagement with a fixed price and a working deliverable at the end. Most operators start with one and add the next when it pays for itself.
Operations centralization
We map how work actually moves through your business, then build the one place it should live — with the existing tools you keep either integrated or retired on purpose.
- A written map of your current workflow and its failure points
- A single operational system your team works in daily
- Integrations to the tools worth keeping
Quote-to-cash automation
Estimate, approval, schedule, completion, invoice, payment — wired end to end, so an approved job becomes a scheduled job becomes a sent invoice without anyone retyping it.
- Automated handoffs between each stage
- An audit trail on every state change
- Alerts where the pipeline actually stalls
Talk-to-your-data interfaces
Ask your own business a question in plain English and get an answer drawn from your live records — with the query and the source rows shown, so you can check the work instead of trusting it.
- A private assistant scoped to your data only
- Sources shown with every answer
- Read-only by default; writes are explicit
Data-integrity engineering
The foundation under the other three: isolation, provenance and auditability designed in from the first table, not retrofitted after something goes wrong.
- Single-tenant data isolation
- Provenance recorded on every automated action
- Access rules you can read and verify
The difference
Light at the center. That's the whole idea.
The name is central + lux: one place where the truth of your business lives, and enough light on it that you can check it yourself. Most AI work fails on the second half.
Your data is not in a shared pool
Every engagement gets its own isolated store. No shared multi-tenant tables, no cross-client bleed, no quiet second life for your records as someone else's training material.
Every automated action is attributable
When a system acts on your behalf, the record says what acted, on what input, at what time, and on whose authority. Automation you can't reconstruct afterwards is automation you can't defend.
Answers arrive with their sources
Anything the system tells you can be traced back to the rows it came from. That is the difference between a tool you trust with an invoice and a demo that sounds confident.
An AI system you can't audit is a liability wearing a leverage costume.
How an engagement runs
Three phases. A price before each one.
No open-ended retainers to get started, no discovery that bills for months and ends in a slide deck. Each phase is scoped, quoted and approved before work begins.
Phase one
Map
We sit with the people doing the work and trace how a job actually moves from first call to money in the bank. You get a written map of the workflow, the leaks in it, and what fixing each one is worth. It stands on its own even if you stop here.
Phase two
Build
We build the system against the map, in working increments you can use as they land — not one reveal at the end. Data isolation, provenance and access rules are part of the build, not a later hardening pass.
Phase three
Hand over
Code, credentials, accounts and documentation transfer to you. We train your team in the system they'll actually use. Ongoing support is available and optional — never a condition of the handover.
- Pricing
- Flat fee per phase, agreed in writing before the phase starts. You get the number first.
- Ownership
- The code, the data and the accounts are yours. Nothing is held hostage to keep you paying.
- Lock-in
- None. No proprietary runtime you can only rent from us, and no exit fee for leaving.
Built for
Operators who are past the point of patching it
- Field-service, trades and contracting businesses running real volume
- Owners who already know which handoff is costing them money
- Teams whose data is sensitive enough that where it lives matters
- Anyone who wants working software, not a strategy deck and a vendor list
Not a fit if
We'd rather say so on the first call
- You want a chatbot bolted onto the website and nothing behind it
- The goal is a headline about adopting AI rather than a fixed operational problem
- Nobody internally can make a decision or grant access to the systems involved
- The budget assumes an off-the-shelf subscription, not engineering
Bring the messiest part of your operation. We'll tell you what it would take.
A discovery call is a working conversation, not a pitch: forty-five minutes on how your jobs actually move, and an honest read on whether this is worth building.
No cost, no obligation. If it isn't a fit, we'll say so on the call.