How we work, and how we charge.
Four steps, no slide deck, and you keep the numbers at every stage.
A scoped build, a monthly run fee, and an outcome component tied to what the system handles. We’re paid on what the system does, not on hours.
Tell us the problem. We’ll tell you whether we’ve solved one like it.
- 00:05You describe the specific problem: AP backlog, order volume, margin leak, the spreadsheet nobody can retire.
- 00:15We tell you honestly whether we’ve solved one like it, and show the case if we have.
- 00:25A realistic first 90 days: what we’d scope, what it would cost, what we’d expect to measure.
- 00:30You leave with either a next meeting booked, or a clear reason we’re not the right fit.
Outcome: a next meeting booked, or a clear reason we’re not the right fit.
Live in four weeks. You keep the numbers.
We run a pilot on your own documents, on your systems as they stand: orders, confirmations, invoices or remittances, whichever process we start with. The measure is simple: what proportion records without a person touching it, and what happens to the ones that don’t. The gap is usually discipline rather than software, such as purchase-order coverage, order formats, or a customer or supplier who never sends what they should, and we will show you where it is.
Connect
Your ERP or accounts package, your inboxes and your bank feed, read-only to begin with. Nothing changes in the ledger.
Rules record
Deterministic matching runs on your real documents. You see what fits first time, and what doesn’t.
AI resolves
The agent works the exceptions: reads the correspondence, contacts the customer or supplier, recommends. You approve.
Measure
The proportion that recorded itself, the hours it returned as a monthly figure at your cost per hour, and which customers, suppliers or document types are the gap. The numbers are yours either way.
Outcome: the proportion that recorded itself, the hours it returned as a monthly figure at your cost per hour, and where the gap really is. No long-term contract.
A scoped build at a transparent cost.
Four to twelve weeks, depending on how many processes we take on first. We take the first process live before we start the second. Where a standard pattern fits, such as statements, cash application or order intake, it is live in weeks. Where your process has logic of its own, such as haulage line-matching, multi-entity consolidation or customer-specific pricing rules, we build it against your systems.
Deterministic where we can, AI where we must.
Classical software and rules for the logic that is rule-based, which is most of it. The AI is reserved for the judgement calls. That keeps cost, latency and surprises low, and keeps the audit simple.
Confidence gates, not blind autonomy.
Every recommendation carries a confidence score and the evidence behind it. Below the threshold, the agent says it isn’t sure and routes the case to a person with everything known so far.
Resolution memory.
When a person resolves an exception, the pattern is kept. Next time the agent recommends sooner, and patterns that prove reliable are promoted to rules. The return compounds rather than plateaus.
A full log.
Every input, every step and every output, by rule, agent or person. Regulated businesses in healthcare, finance and aviation need it to deploy at all; everyone else finds it useful the first time an auditor asks.
On your infrastructure or in our cloud, same architecture.
Cloud models where your data and your regulator are content with them. Local models on your own hardware, behind your firewall, where they aren’t. Same platform, same log; only where it runs changes.
Nothing intelligent posts to your ledger. Rules post. The AI recommends. People approve.
The same team that built it runs it.
A monthly fee for the platform and the people who run it, plus the outcome component: per document, per exception, or a share of what you don’t spend. Processes change, customers and suppliers change, agents need tuning; that work doesn’t end at go-live, and it isn’t handed to a junior support team. If you host it and want your own team running it day to day, the same fee covers the platform and our people behind them.
Three ways it runs: in our cloud, run by us; on your infrastructure, run by us; or on your infrastructure, run by your team, supported by us. Your people approve in every case.
Where it helps, one of ours sits inside the business.
Where it helps, one of our people sits inside your business part-time: operational leadership on the COO side, technical leadership on the CTO side. Usually as part of a run engagement, and available on its own.
Fractional COO.
Operating plan, process, team structure and change, owned by a senior operations leader in your leadership team.
Fractional CTO.
Strategy, architecture, vendors, security and board communication, owned by a senior technologist in your leadership team.
It only pays if your team works with it.
Adoption is a people problem before it is a technology one. Sylvia Stafford leads this side of the work: how the team is shaped around the work we now run, and the sponsorship that carries it.
Operational restructuring.
Around the process we now run, because the shape of your finance and operations team shouldn’t look the same afterwards as it did before.
Role redesign.
Which roles change and how, communicated honestly to the people in them. Most teams we meet have a vacancy they can’t fill, not a person they want to lose.
Training.
For the people who approve what the agent recommends, the managers who read the numbers, and the leaders who steer what we take on next.
Executive alignment.
Sponsorship from the top through the months it takes to land, which is where most change programmes stall.
Process review first.
Where the process itself needs work before automation adds value. Automating a broken process just breaks it faster.
Four things we won’t sell you.
We don’t replace your ledger.
The ERP or accounts package stays the system of record.
We don’t run production.
No MRP, routing, work orders or stock valuation.
We don’t become your IT department.
We run the processes we take on, and connect to the systems around them.
We’ll tell you when the answer is no.
If a process isn’t ready to automate, or if a rule would do the job without an AI, we’ll say so.
Start with thirty minutes.
No slide deck. Tell us the process; we’ll tell you whether we’ve solved one like it before, and what a realistic first 90 days would look like.