How CostCtrl works.

A deterministic engine computes every number. An AI layer speeds up building the model and reads the result back in plain language. Below: what each part does, how we get you live, and how CostCtrl compares with what you use today.

The allocation engine, and the AI on top of it.

CostCtrl is two systems working together. Our allocation engine applies consistent rules to every cost and every customer. On top of it, an AI layer reads the output the moment it lands and turns it into the analysis a finance team would otherwise wait weeks, and pay an analyst, to produce.

The flow, end to end

Inputs

Financials, ops data, master data

Engine

Cost pools → activities → outputs

Attribution

Every $ allocated, every rule explained

AI layer

Patterns, insights, board-ready output

What the engine does.

The engine is the heart of CostCtrl, and it is deterministic. It maps your revenue streams, activities, cost pools, products, and customers into a single operating model, then attributes every cost to where it was consumed, using driver logic applied consistently across the whole business.

There is no guesswork and no approximation in this layer. Every attribution carries the rule that produced it, and every number can be traced back to its source. This is the part you can take to a board meeting and defend line by line.

What the AI does.

The AI sits on top of the engine and does one job: it removes friction. It accelerates the work of building and maintaining the model, and it reads the engine's output to surface insights in plain language: what's driving a result, where margin is leaking, what a scenario would do.

It does not compute the numbers and it does not invent them. Through the MCP port it can also drive the model, changing inputs and asking the engine what happens, and the same rule holds: the figures come from the engine. That separation is what lets you move fast and still trust the result.

Generic AI, on its own

Produces confident answers you can't verify
Numbers drift, or are made up
Nothing you'd put in front of a board

CostCtrl: engine + AI

Every figure computed by the deterministic engine
Each number traceable to its rule and its source
The AI reads and explains, it never invents

The allocation engine is what makes it reliable. The AI is what makes it fast. Apart, each one disappoints. Paired, they do the job neither can do alone.

Closed loop

The port that lets your own AI drive the model.

The MCP server adds a standard port to the engine. Plug in the AI you already pay for (Claude, ChatGPT, Copilot or your own) and it can create a scenario, change inputs, run the calculation and read the result. The agent changes inputs. The engine computes the consequences. The agent never types a number into the P&L.

Build

From raw exports to a first calculation, with everything the agent could not place with confidence listed for the finance team to decide.

Diagnose

Actuals land, the routine recalculates, compares to plan and to last month, and reports the drift. Every month, not once per project.

Tune

Set a target and constraints. The agent searches and the engine computes every candidate. Excel's Goal Seek, for the whole business.

Audit

A sceptical second opinion on the model itself: driver rates, proxy drivers, sensitivity, with the rule behind every number.

How the closed loop works

How we work with you

From first conversation to a model you run.

One agreement from the start, and you stay in control of how much help you want along the way.

Setup

We build your model on your own data and get you live, typically in 3 to 6 weeks. The build is engineer-led and accelerated with our own AI tooling. It ends with quantified improvement opportunities, each with a financial target, presented to the person who owns the budget.

The first three months

A one-off setup fee covers the build, and the monthly platform fee starts at sign-up. You can cancel at any time in the first three months, so you see the result before you commit to a year. The model you run is the one we built, with no restart.

Run it and grow it

After month three the plan runs on an annual commitment, and the model is yours to grow. Add product lines, refine drivers, run new scenarios as the business changes. Your finance team can drive it day to day, with our team alongside for as much or as little support as you want.

As autonomous or as hands-on as you like. Some teams run everything themselves after go-live. Others keep us close for ongoing analysis. Either way, the model belongs to you.

How CostCtrl differs, in one paragraph each.

Vs. BI tools

BI dashboardsCostCtrl

BI tools show you what's in your data. CostCtrl tells you what your data means once costs are properly attributed. A dashboard can display that answer. It cannot produce it.

Vs. generic AI / DIY

A chatbot and an analystCostCtrl

General-purpose AI will happily produce a confident answer that nobody can check. CostCtrl's numbers come from a deterministic engine grounded in 25 years of cost-modelling practice, with every figure auditable. The AI speeds up the work. It does not supply the result.

Vs. traditional ABC / TDABC

Enterprise cost toolsCostCtrl

The same methodology the best consultants use, ABC and TDABC both, built into the product and ready to run. CostCtrl supports both calculation methods, so the model fits the business rather than forcing the business to fit one method. The rigour, without the multi-month engagement or the consulting invoice.

Vs. consulting projects

A deck and a deliverableCostCtrl

A consulting project ends. CostCtrl is a live model your finance team owns, updates monthly, and stands behind in the next board meeting.

A short history of knowing what things cost.

The methodology behind CostCtrl isn't new. The rigour was always sound. The cost of putting it to work was the problem.

1990s

The ABC movement, and why it stalled.

Activity-based costing arrived in the late 1980s and dominated the cost-accounting conversation through the 1990s. The premise was simple: allocate costs by the activities that drive them, not by a flat percentage of revenue. The implementations were not simple. Models took years to build and broke within a quarter. By the 2000s, most companies had given up.

2000s

TDABC: the evolution of ABC.

Time-Driven Activity-Based Costing, developed by Kaplan and Anderson, evolved ABC into something more powerful and more scalable, modelling cost through the time each activity consumes. A real step forward, but a more demanding one: it asks more of a company's data, its systems, and the maturity of the team running it. For most businesses, that demand kept a rigorous model out of reach.

2010s

Enterprise cost platforms calcified.

A handful of enterprise platforms made cost modelling tractable for very large companies. They required six-figure implementations, dedicated consultant teams, and year-plus deployments. The mid-market couldn't afford them.

Now

Why now.

We paired the proven methodology with our own allocation engine and an AI layer. The engine computes every number. The AI builds and reads the model alongside your team. Work that used to take months (schema generation, data integration, driver inference, insight synthesis) now takes weeks, at a price the mid-market can justify.

Want a walkthrough?

Book a 30-minute call with Miguel. Bring your questions. He'll bring the methodology and twenty-five years of seeing what works.

Book a 30-minute call →