You can cut the month end close roughly in half without sending a single journal to the cloud. Mickai runs the whole close on your own hardware: Plutus reconciles the ledger, Daedalus builds the schedules and the board pack, and Aletheia drafts the audit working papers, replacing BlackLine and close management licences and the manual matching labour around them, with every step sealed into a signed audit record that no data egress or API fee ever touches.
Why agentic AI financial close is the 2026 back office story
Close automation is the story finance is telling itself this year. Vendors are shipping agentic close features, and industry reporting puts the direction of travel plainly. A January 2026 Deloitte study found that most finance organisations have already deployed AI somewhere in their operations, and analysts expect almost every finance function to run at least one AI solution by the end of the year. One widely cited survey found only a small share of finance leaders using agentic AI today, with close to half expecting to adopt it within twelve months.
The pull is obvious. Industry reporting attributes cuts in close cycle time of roughly 40 to 55 percent to close automation, along with large reductions in reconciliation errors. The catch is where the work runs. Most agentic close tools are cloud services, which means your general ledger, your bank data and your intercompany balances are sent to someone else's model, metered per seat and per token, with the audit trail living on their infrastructure. For a CFO, a financial controller or a shared service centre, that is the wrong trade.
What the close costs you today
A modern close stack carries three cost centres. First, the software: a reconciliation and close management licence, typically BlackLine or a comparable tool, priced per seat, per account and per module. Second, the labour: controllers matching thousands of lines by hand each period, chasing exceptions and rebuilding schedules in spreadsheets. Third, the hidden meter: when the close runs on a cloud copilot, every run carries a per-token API fee and moves your ledger off site, which is data egress you pay for in money and in risk.
Then comes the audit. External audit prep is its own scramble, days of controller time assembling evidence, tracing balances and reconstructing who changed what and when. None of that is billable value. It is the tax you pay for a close that was never designed to leave evidence behind it, and it lands on the same team that has just spent a fortnight getting the numbers to agree.
How Plutus, Daedalus and Aletheia run the close on your own hardware
Mickai runs the close as a set of studios. A studio is a ready-made application for one business function, sitting inside a single system you own and control. Plutus, our accounting studio, is the sovereign CFO stack: it ingests the trial balance, the sub-ledgers and the bank statements from your own systems, reconciles accounts, matches transactions and flags the exceptions with reasons. Because it runs on your hardware, nothing leaves the building and there is no per-token meter counting in the background.
Daedalus, our documents and spreadsheets studio, turns the reconciled numbers into real work product: reconciliation schedules with live formulas, flux and variance analysis, and a board-ready close pack, generated on screen in seconds. The finance team receives spreadsheets it can open, trace and check, not a black box that asks to be trusted. The whole sequence is designed so the machine does the matching and the drafting, and people spend their time on judgement.
- Ingest the trial balance, sub-ledgers and bank statements from your own systems, with no data sent off site.
- Plutus matches transactions, reconciles accounts and scores the exceptions, so people review the handful that need judgement rather than every line.
- Daedalus builds the reconciliation schedules, the flux analysis and the board pack as live spreadsheets and documents.
- Aletheia drafts the audit working papers and control assessments to ISA (UK) 230, with independence preserved.
- Every action is sealed under post-quantum cryptography into the Open Audit Record, so the evidence is ready before the auditor asks for it.
- A human controller reviews and signs off the exceptions, keeping judgement and accountability with the finance team.
Sealed working papers, before the auditor asks
The reason on-premise matters is not only cost, it is evidence. Every action Plutus, Daedalus and Aletheia take is sealed under post-quantum cryptography into the Open Audit Record, a signed, tamper-evident log of what happened, on what data, and who approved it. Aletheia, our audit studio, drafts the working papers and control assessments to ISA (UK) 230 with independence preserved, and the sealed record sits underneath them as the source. The Assistant that drives all of this runs on your own brain, a model built on your own data, on your own hardware.
To be precise about what that buys you: the system does not hold a certificate and does not grant one. It produces evidence that supports an examination. Your auditor still forms their own opinion, and your controllers still sign off. What changes is that the evidence is a by-product of the close, captured as the work happens, rather than reconstructed weeks later from memory and email. That is where the audit prep hours come down, and it is a mechanism, not a promise.
What you replace, and what you save
Every row here follows the same pattern: a licence or a manual step, replaced by a studio that runs on hardware you already own. The meter that disappears is named in each case.
| What you run today | What it costs you | With Mickai |
| --- | --- | --- |
| BlackLine reconciliation and close licences | Per seat, per account and per module subscription, every year | Plutus reconciles and closes on your own hardware, so the subscription meter stops |
| A close management tool such as Trintech | Annual close-orchestration licence and support fees | Plutus orchestrates the close on premise, with no external subscription |
| Manual reconciliation and matching labour | Controller hours matching thousands of lines each period | Plutus matches and scores exceptions, so people review only what needs judgement |
| Cloud AI close copilots | Per-token API fees and data egress each time the ledger leaves the building | Runs on device, so no egress and no per-token meter |
| AuditBoard or Workiva working-paper tools | Per-seat audit and working-paper subscription | Aletheia drafts working papers to ISA (UK) 230, sealed, with no separate licence |
| Audit prep scramble | Days of controller time assembling evidence for external audit | The Open Audit Record captures the evidence as the close runs, cutting prep hours |
The money that used to leave as subscription, egress and per-token fees stays in the building, and so does your ledger. That is the whole trade: the same close, faster, on infrastructure you own, with the evidence already written.
Frequently asked questions
Does agentic AI financial close mean our ledger leaves the building?
No. Plutus runs on your own hardware. The trial balance, the sub-ledgers and the bank data stay on premise. Nothing is sent to a cloud model, so there is no data egress and no per-token API fee. The close is fully offline.
What exactly does Mickai replace in a close stack?
The reconciliation and close management licence, typically BlackLine or a comparable tool such as Trintech, and the manual matching labour around it. Plutus reconciles and closes, Daedalus builds the schedules and the board pack, and Aletheia drafts the working papers, all inside one system you own.
Will auditors accept work produced this way?
The system produces sealed evidence that supports an examination, it does not grant a certificate. Every action is written to the Open Audit Record under post-quantum cryptography, and Aletheia drafts working papers to ISA (UK) 230 with independence preserved. Your auditor still forms their own opinion, on evidence that is easier to trace.
How much faster can we actually close?
Industry reporting puts close automation at cutting cycle time by roughly 40 to 55 percent, depending on the starting point. We do not publish a single savings figure of our own, because your gain depends on how manual your current close is. The mechanism is the same either way: the machine matches and drafts, your controllers review the exceptions.
