The order of operations, and why it is the whole argument.
Nine minutes. Your own numbers. No call required.
An artificial intelligence system is a multiplier. Placed on work that matters, it returns its cost in months. Placed on work that does not, it returns a demonstration and a maintenance bill.
Establish where time, margin and information are being lost. Price each loss in the organisation's own hours and rates. Then decide what to build.
This is why every engagement begins with the Cortex AI Audit, and why the Cortex Diagnostic that follows it is paid work with a deliverable. The picture it produces is the asset. What gets built afterwards is downstream of it.
If the diagnostic finds less recoverable value than it costs, the fee is refunded.
Every finding in a Cortex audit maps to one of three lenses. The lenses are what make an audit repeatable, and what stop an interview turning into a wish list.
Hours consumed by work that produces nothing a customer would pay for. Status meetings, manual assembly of reports, chasing an update that already exists somewhere in writing. It is the easiest lens to price, because people can usually tell you the hours once they are asked precisely.
The same output produced more than once, by more than one person, because the first version could not be found or could not be trusted. Onboarding written from scratch for every client. Answers given again to a question answered last month.
Decisions that stall because the number, the file or the approval is somewhere else. This is the lens that most often goes unpriced, and it is usually the largest, because nobody counts the hours a decision spent waiting.
Every finding on a Cortex roadmap is tagged with its lens, and the register is grouped by lens, so the register shows a pattern. The register is published with every assumption shown.
Where projected payback on a proposed engagement runs past twelve months, Cortex proposes the smaller engagement and shows the arithmetic behind the decision.
This is enforced in the system that produces the roadmap, so it does not depend on the discipline of whoever is presenting it.
It has already happened. The same two figures that produced the 62,000 euro finding for a 31 person agency ruled out our largest engagement for that company: payback would have run past sixteen months, so it was modelled and not proposed. The roadmap said so, in writing, on the page the client read.
The Cortex AI Audit. Six steps, your own numbers, no call.
Three weeks. A structured examination of how your organisation operates, delivered as the Cortex Diagnostic.
Thirty minutes with the person who runs the business, against a question script written from that company's audit answers. It is free, it follows the report, and it is where the shape of the organisation, the commercial picture and the interview list are established. The Cortex Diagnostic begins after it.
Four to eight conversations, one hour each, with the people who carry the work rather than the people who describe it. Structured interviews surface what documentation never records: the decisions, exceptions and judgment that determine whether a system will hold in production.
Every number said out loud in an interview is captured with the quote that carries it. Each figure is built from headcount, hours and loaded rates supplied by the business, and each one appears with its basis attached. A drain nobody can put hours to is recorded as unpriced, and stays out of the total.
The register becomes a plan.
Four axes. Value, the annual figure and the confidence behind it. Feasibility, how much has to be true for it to work. Cost, to build and to run. Risk, what happens if it is wrong.
The scores produce a sequence. Some findings are worth more but depend on something else existing first, and a roadmap that ignores dependency is a list.
Leading findings are demonstrated with working proof of concept systems. A working demonstration answers one question: whether the thing behaves the way the roadmap claims when it meets the company's actual data.
The roadmap arrives as a presentation and as a living page in the Cortex Portal, with the register, the sequence, the scopes and the payback on each. The roadmap is yours and can be used with any builder.
We build what the roadmap prescribes, then we keep the number in it true.
Most of it runs on infrastructure the company already pays for. A business with four to seven systems is usually using a fraction of each, and connecting what exists wins on cost, on timeline and on the only measure that matters afterwards, which is whether people use it. Delivery runs on a template catalogue: a proven base system, plus the custom work a specific company needs on top. Training runs until the systems are in production and in daily use.
Then adoption is measured. Value captured is tracked against the projection the build was sold on, and both numbers stand in the Portal, current, all year. The roadmap is revisited as the technology advances: a finding that was infeasible in March is often routine by September. Every build carries a continuity agreement, because systems built on models that change every few weeks decay when nobody is watching them.
Every figure is built from three inputs the business supplies: how many people, how many hours, and the loaded hourly cost of those people. The arithmetic is shown alongside the number, and the assumptions are labelled as assumptions so they can be substituted. No benchmark figures and no industry averages are used at any point.
Findings are reviewed with the client before the roadmap is presented, and a figure that cannot be defended in that conversation is removed. The register is more useful when it is short and correct.
The interviews take one hour per person and are scheduled around the business. Total time consumed across a 31 person company is typically around six hours. Nothing is installed and no system access is required.
A Cortex audit runs three weeks from the first conversation to the roadmap presentation. Most of that is scheduling. The interviews themselves consume around six hours of your team's time in total, spread across the people who know how the work is really done.
No system access is required for the audit. It is conducted in conversation, supported by whatever documents a company chooses to share. Access arrives later, during implementation, and only to the specific systems a build touches.
Nine minutes. Six steps. A number built from your own hours and rates. Nothing to install, and no call unless you ask for one.
Your score appears on screen. The written report arrives by email.
