AI for accountants.
Built by a studio that has shipped to a real ledger.

The Xero invoice agent we built — the one that scoped the work, drafted the proposal, and pushed an invoice a real client paid — is in production today. The finance agents we build for clients run on the exact pattern: bookkeeping reconciliation in Xero or QuickBooks, accounts payable that codes invoices against your GL, accounts receivable that chases without ruining the relationship, contract terms extraction that surfaces every renewal date. Human-in-the-loop on anything that hits the ledger.


Four finance agents.
Each one wired to your real ledger.

Not chatbots answering finance questions. Agents that log into Xero, read your bank feed, code your invoices, chase your debtors, and surface what needs your attention. Audit trail on every commit, kill switch on every agent, ownership stays with you.

AI bookkeeper

Reconciles Xero, QuickBooks, or Sage transactions. Categorises expenses against your chart of accounts. Generates and sends invoices on agreed triggers. Flags anomalies for human review. The Xero invoice agent is live with a paying client. The productised version is Signet Bookkeeper.

Accounts payable automation

Reads inbound invoices via OCR, matches against POs, codes against the GL, routes for approval, posts to the ledger. Three-way match where the data supports it, exception flagging where it does not. Cuts AP processing time by 70-90% on routine invoices.

Accounts receivable chasing

Tracks aging, drafts dunning emails in your tone, sends them on cadence, escalates after rules you set. Voice agent option for hands-on follow-up calls (built on the same stack as our Voice AI practice). Reduces days sales outstanding without the awkward human conversations.

Contract terms extraction

Reads your supplier contracts, customer contracts, and finance leases. Extracts payment triggers, renewal dates, price escalators, termination windows, indemnity caps. Pushes the data into your ERP or contract registry. Finance team and legal team work from the same single source of truth. Often pairs with our AI for Legal work.


One AI agent, paid by a real client.

In one session, an AI agent did this end to end:

  1. Built the deliverable. A working piece of software for a client problem. Deployed, tested, live.
  2. Wrote the proposal. Scoped the work, set the price, drafted the statement of work.
  3. Pushed the invoice to Xero. Created the contact, issued the invoice, sent it via the Xero API.

A human client paid it. Three jobs that used to cost an employee a week, done in a single session, with a human at the approval gates. The Xero agent is live, productised as Signet Bookkeeper, and runs in production today.


Wired to the tools you already pay for.

We do not add another finance tool to your stack. We wire AI agents into the ledger, ERP, and bank feeds you already trust.

Ledger

Xero QuickBooks Online Sage Intacct NetSuite Microsoft Dynamics

Banking

Plaid TrueLayer Open Banking UK Wise Stripe

Document AI

Azure Document Intelligence AWS Textract Google Document AI Mistral OCR

Reasoning and orchestration

Claude API OpenAI MCP Function calling Audit logging

Human-in-the-loop, by default.

Finance is not the place to let an AI commit autonomously to your ledger. Every action the agent takes is logged with the source data, the rule it applied, the confidence level, and the human approval (if required). Your audit trail is stronger after we wire the agent in than before, not weaker.

Approval gates are configurable per workflow. The bookkeeper agent might auto-categorise transactions under £100 against a learned chart of accounts and queue everything above for human approval. The AP agent might auto-pay matched invoices from approved vendors and route everything else for two-eye review. You decide where autonomy ends.

We provide a kill switch on every agent. If the output drifts, you stop the agent in one click. We deploy in your tenant with your model provider keys, so your data does not feed any vendor's training. SOC 2 and ISO 27001 audit-friendly by design.

Audit Per-action trail with source
HITL Configurable approval gates
Tenant Your Azure or AWS
Kill One-click stop on every agent

You might want a finance AI agent if...

01

You are a finance lead at a mid-market company quietly considering hiring a second bookkeeper, AP clerk, or AR analyst.

02

Month-end takes longer than it should because someone has to chase data from five systems before the close.

03

Days sales outstanding has crept up and your AR follow-up is inconsistent across debtors.

04

You have a pile of supplier and customer contracts and nobody has surfaced their renewal dates and price triggers in one place.

05

You are an accountancy practice and your team is doing the same reconciliation work on every client every month.

06

You evaluated a finance AI vendor and the demo did not match the reality of your data, your chart of accounts, your tooling.


Three ways to start.

Scope

Finance AI Discovery

One week, fixed

We map your ledger, ERP, document flow, and approval rules. Pick the first agent shape, design the integrations, scope the build. You leave with a written specification, an effort estimate, and an honest read on whether you should buy a vendor or build with us.

  • Process and tooling map
  • Build-or-buy recommendation
  • Audit and approval design
Build

Finance Agent Build

4 to 8 weeks

One agent built end to end against a defined workflow (bookkeeping, AP, AR, or contract extraction). Goes live with full audit trail and human approval on every commit until your team trusts it. We hand over runbooks and ownership.

  • Production agent in your tenant
  • Audit trail and HITL approval
  • Runbook and handover
Run

Managed Finance Agents

Monthly retainer

We monitor accuracy, retrain as your chart of accounts and processes evolve, push fixes when output quality drifts. Quarterly audit-readiness review. Model and infrastructure costs at cost.

  • Output monitoring and accuracy QA
  • Quarterly audit-readiness review
  • Updates as your processes change

Adjacent capabilities.

Voice AI

Voice AR-chasing agents. Voice receptionist for the finance inbox. Same studio, voice-first delivery on the same finance stack.

Data AI

Document extraction, RAG over financial policy docs, knowledge agents that cite source. The plumbing under most finance AI work.

AI for Legal

Where contract terms extraction crosses into legal review and obligation tracking. Finance and legal share the data; both sides benefit.


Finance AI that has shipped, for teams that want to close the books.

We work with finance and accounting teams across the UK, US, and Australia. The Xero agent is live. Yours can be next.