Systems tailor-built to last
Designed around the process you already have, not around what a product expects it to be.
An operational system designed around a process the company already has and cannot bend to fit software. We take it end to end: architecture, engine, interfaces, integration into the landscape that is already there, production, and running it afterwards.
People arrive here when off-the-shelf did not fit, when the ERP went in and the work is still being done in Excel beside it, or when the integrator they asked will not take it on.
A real number needs two weeks of looking first.
A fixed price on a system built from zero rests on how much of it we have built before. On the parts we have built many times the estimate is solid — we know their shape and what getting them wrong costs. The parts specific to your process are the reason you are here, and an estimate made before reading them is a guess. A guess that comes in low is paid for by the system rather than by us.
So the sequence is: a paid discovery of 2 weeks, and then a fixed price on the build. The discovery is what turns the specific parts into known ones. The discovery is what you buy the right to a real number with. It is credited against the build within 60 days.
If a number is needed this week, this sequence cannot produce an honest one. That is a real constraint, and it is worth knowing in the first call rather than the fourth.
What that includes
Calculation engines whose results can be explained — where the number the system produces can be traced back through every input that made it, because someone will ask, and "the model said so" is not an answer in a business that carries the consequence.
Excel processes moved into real software without losing the logic the user relies on — including the logic nobody wrote down, which is usually the part that matters and always the part that is discovered in week three.
Integration with ERP and industry systems — including the ones with no API, the ones two versions behind, and the ones whose only export is owned by one person.
Multi-agent architectures for production environments — with boundaries of responsibility that hold when a component is confidently wrong.
Operation under constraints on bandwidth, uptime and data residency — narrow links that drop, sites with no engineer present, and jurisdictions that decide where the data may sit.
Migration between infrastructure or model providers without downtime — planned at design time, because it is nearly impossible to retrofit.
Running it afterwards — the handover leaves you able to run it without us, and we stay on because that is useful rather than because it is the only way it keeps working.
How the work runs
- Understand — 2 weeks. Read access to your systems, 6–8 hours of someone who knows the process, real data samples, your non-functional constraints, and one named person who can decide.
- Design — architecture, the integration map, decomposition into milestones with week estimates, and the list of assumptions that would reopen the estimate if they break. That list is not fine print. It is the honest part.
- Build — in milestones, at a price fixed after the discovery. Built for the conditions it will meet: a link that is narrow and drops, and a site with nobody technical on it.
- QA — a fixed set of cases with known outcomes, run per milestone and again before release. It is what tells you that a change to one part did not break another.
- Integrate into your landscape — into the systems as they are configured today. Integration is planned first, because it is the phase that decides the schedule.
- Hand it over — repository, documentation, configuration and prompts, with no lock to us and none to a model provider.
- Support — a retainer priced by the volume handled. The person answering is the person who built it.
What a discovery actually produces
| System | Direction | Interface | Owner | Flag |
|---|---|---|---|---|
| ERP · Dynamics NAV 2016 | read+write | OData v4 | client IT | at risk |
| 2016 exposes OData partially. Writes go through a staging table, +1.5 weeks against a current-version assumption. | ||||
| Item master · AS/400 | read | nightly CSV | client IT | at risk |
| No API. The extract is owned by one administrator who retires in March. Documenting it is in scope. Replacing it is not. | ||||
| Mail · Exchange 365 | read | Graph API | client IT | ok |
| Pricing · Excel on share | read | file watch | sales | at risk |
| Structure changes without notice. We version every copy we read and fail loudly rather than silently. | ||||
| Bank · MT940 | read | SFTP | finance | ok |
| Carrier portal | read | none | external | at risk |
| No API and no contractual right to scrape. Out of scope. Manual export agreed with operations. | ||||
| Label printer | write | ZPL over LPR | warehouse | ok |
ASSUMPTIONS THAT REOPEN THE ESTIMATE
- NAV stays on 2016 through the build. An upgrade mid-project changes the write path and the estimate with it.
- The AS/400 extract keeps its current column set.
- Two named people are available four hours a week each.
- The carrier portal stays out of scope.
DECOMPOSITION
- Milestone 1 · intake and item resolution
- 4 weeks
- Milestone 2 · calculation engine + explanation trail
- 5 weeks
- Milestone 3 · NAV write path via staging
- 3 weeks
- Milestone 4 · operator interface and handover
- 3 weeks
- estimate
- 15 weeks
Illustrative data.
Seven systems, four flags, four assumptions and 15 weeks. Every flag is a place this project could slip, named before anyone signed anything. This is the deliverable — not a slide deck about our approach.
What it costs
| Package | Duration |
|---|---|
| A2 · Discovery | 2 weeks |
| A3 · Build | from 8 weeks |
The discovery gives you: a written feasibility opinion · the solution architecture · the integration map with every connector named · decomposition into milestones with estimates in weeks · the assumptions that would reopen the estimate · and a fixed-price proposal for the build.
It does not give you a prototype, code, or access to our framework. It is 2 weeks of engineering judgement, and it is priced as such.
Credited against the build within 60 days. If you stop there, the document is yours — another supplier can build from it.
What is sold is a system that runs, not people who are available. If what you need is capacity rather than a system, that is cheaper elsewhere, and we will say so on the call rather than take the work.
This is the line where we have the most to show
AgriCore — grain trading. Closed testing on live deals at partner desks. Ten signed LOIs. Our strongest traction anywhere in the company.
LogiSea — maritime fleets. Three vessels in installation. Built to run over a link that is narrow and drops, on hardware already on board.
Two systems, two industries with nothing in common, both built from zero, both integrating into landscapes we did not choose.
What we have not published yet is a sample discovery — the real document behind the map above. It is on the list, and until it exists, the map on this page is an illustration rather than an artefact.
What this line assumes
If a product fits your process, buy the product. It will be cheaper, faster and better supported than anything we would build. We would rather tell you that in the first call than in month four.
Your ERP stays as it is configured. We build beside it and integrate into it.
Every build starts with a discovery — including for a smaller scope, and including for a client we like.
What is sold is a system that runs, not people who are available. Scope sets the price; retainers are priced by volume handled.
A system built from zero usually contains agents, and stands on a data layer that had to be built first.
Forty-five minutes. Bring the process — the one that does not fit anything you have looked at. We will tell you whether it needs building, and roughly what would be hard about it.