The finance studio: sovereign AI for the office of the CFO

Why the CFO has been left waiting

Finance was supposed to be an early winner from artificial intelligence. The work is dense with structured data, repeatable judgement and clear right answers, which is exactly what a capable model should handle well. Yet the office of the CFO has been one of the slowest places to adopt it, and for a very good reason. The material a finance team touches is the most sensitive a company holds. Unpublished results, the cash position, salary tables, the model behind an acquisition, the covenant tests that decide whether a facility stays open. None of that can be pasted into a public chat window, and no responsible finance leader would allow it.

So finance teams have been stuck with a choice that is really no choice at all. Send your numbers to somebody else's cloud and hope the contract holds, or stay on spreadsheets and manual review and watch the calendar. We think that is a false trade off, and we built the finance studio inside Mickai to remove it entirely.

What the finance studio actually is

Mickai is a Sovereign Intelligence Operating System, a SIOS. It is not an app you log into and it is not a service that reaches back to a data centre somewhere. It runs on the customer's own hardware, on premises and air gapped when required, with zero data egress and no public cloud round trip. The finance studio is the part of that system tuned for the work a finance function does day to day. It draws on 50 specialist brains, 25 domain and 25 operational, coordinated under deterministic governance so that the same inputs and the same policy produce the same result every time.

The point worth sitting with is where the data lives. When a finance team asks the studio to reconcile a ledger, model a scenario or draft a commentary, the source data never leaves the building. The reasoning happens next to the numbers, on the company's own machines, inside its own perimeter. There is no external endpoint to trust, because there is no external endpoint at all.

The work it takes on

We designed the studio around the tasks that eat a finance team's month, the ones that are too important to rush and too repetitive to enjoy. A few of them:

• Close and reconciliation. Matching transactions across ledgers and bank feeds, surfacing the exceptions that need a human, and drafting the narrative that sits on top of the numbers.
• Forecasting and scenario work. Building and stress testing models, running sensitivities on revenue, cost and cash, and explaining in plain language what moved and why.
• Board and investor packs. Turning a clean set of figures into a coherent commentary, with the same story told consistently across the deck, the notes and the covenant schedule.
• Controls and anomaly review. Flagging duplicate payments, unusual approvals and transactions that break a policy rule, before they become a finding.
• Regulatory and audit preparation. Assembling the evidence trail an auditor or regulator will ask for, already organised, already signed.

None of this replaces the judgement of a controller or a CFO. It removes the hours between having a question and having a defensible answer, and it does so on data that was never allowed to travel before.

Every action leaves a signed record

Finance runs on evidence, so the studio was built to produce evidence as a matter of course. Every action the system takes is written to a cryptographically signed audit record, what we call the Open Audit Record. It captures what was asked, what data was touched, which brain acted, and what came out. That record is signed with post-quantum signing, ML-DSA-65, so the trail holds up not only against today's scrutiny but against the decryption capabilities the next decade will bring.

For a finance leader this changes the conversation with the audit committee and the external auditor. Instead of reconstructing after the fact who did what and when, the answer is already there, tamper evident and complete. The memory the studio builds up, the context of past closes and past decisions, belongs to the customer. It is theirs to keep, theirs to inspect and theirs to delete. We do not hold a copy and there is nowhere for one to go.

“A finance team should not have to choose between capable AI and control of its own numbers. Sovereignty is the feature that makes every other feature usable.”

Governance that a controller can trust

There is a reason we insist on deterministic governance rather than letting a model improvise. Finance is a domain where the same question must produce the same answer, where a rule is a rule, and where surprises are almost always bad news. The 50 brains do not run loose. They act inside policy that the customer sets, with the boundaries of each task defined before it starts and the result checked against those boundaries at the end. When something falls outside policy, the studio stops and asks rather than guessing. That is the behaviour a controller needs and the behaviour a regulator expects.

The protection underneath

Everything above rests on a body of work we have filed to protect. Mickai has 104 filed UK patent applications, carrying approximately 2,340 claims, building toward examination and grant. Those filings set out the full specification, the claims and the figures behind the architecture the finance studio depends on, from the governance layer to the audit record to the way sovereign memory is held. We mention it here not as a badge but because a finance leader is right to ask whether the thing they are trusting with the general ledger is a passing idea or a serious, documented system. It is the latter, and the record of it is on file.

A signal worth reading

We are early and we say so plainly. The proof point we can point to is our own public signal. On Crunchbase our founder now ranks number 2, and the company Heat Score has reached 94 out of 100, having climbed from single digits. That is attention, not revenue, and we will not dress it up as anything more. It tells us that the idea of AI a company can own, applied to the most sensitive work it does, is landing with the people watching this space closely.

Where this goes next

The office of the CFO has spent two years watching other functions experiment while it held the line on data it could not risk. We think that wait is ending, not because the risk got smaller but because the architecture changed. When the intelligence comes to the data instead of the data going to the intelligence, the trade off that kept finance on the sidelines simply disappears. That is the shift the finance studio represents, and it is the reason we built Mickai to run inside the walls rather than beyond them. The next quarter close, and every one after it, can be faster, cleaner and fully evidenced without a single number ever leaving the building. We are looking forward to helping finance teams prove that for themselves.