A focused on-site sprint that leaves your finance team with a deployed AI agent — built on your actual data, running before we leave.
Most AI programmes teach your team about AI. This one builds something with them. We arrive with a clear sprint scope, work directly in your environment, and leave a working agent running on your real data.
The team doesn't just watch. They build alongside us — so when we leave, they own it and can extend it themselves.
A pre-session call maps your workflows, data sources, and PoC target. Day 1 starts with a clear objective, not a discovery exercise.
No synthetic demos. We work in your environment — your GL, your ERP, your reporting stack. The agent is live the moment it's done.
Every sprint closes with full documentation, a prompt library, and a roadmap of the next agents worth building.
Three phases. Variable duration. The structure is fixed — the time allocated to each phase scales with your team's size, starting AI maturity, and the complexity of the PoC target.
We map your team's recurring workflows against a finance-specific AI readiness matrix. The output is a prioritised list of use cases — with a clear decision on which PoC to build, and which workflows should not use AI unsupervised.
We build the PoC agent live, in your environment, on your real data. Your team builds alongside us — not watching. We use Claude Cowork for non-technical participants and Claude Code where deeper integration is needed. The agent runs before this phase ends.
We hand over everything — documentation, prompt library, knowledge base — and run a structured session to make sure the team can operate, extend, and maintain the agent independently. We also set the data architecture direction for moving from Copilot mode to autonomous operation.
Duration scales to your team. A CFO and two analysts can complete all three phases in 2 days. A finance function of 10+ typically runs the full 5 days. We agree the timeline on the discovery call — before we arrive.
We scope the PoC together during the Map phase. Below is the full menu — pick one for the sprint, use the Embed session to sequence the rest.
| Balance sheet reconciliation agent | Reconciles BS lines to sub-ledgers, flags variances, drafts commentary |
| Accrual agent | Identifies missing accruals based on historical patterns and open POs |
| Month-end close tracker | Tracks task status, chases owners, flags blockers in real time |
| Intercompany reconciliation agent | Matches interco balances across entities, surfaces mismatches |
| Journal entry reviewer | Flags unusual entries vs. historical patterns before sign-off |
| Audit trail generator | Auto-documents the reasoning behind journal entries and adjustments |
| Variance commentary generator | Actual vs. budget narrative, board-ready, in your house style |
| Revenue forecasting model | Driver-based forecast that updates automatically as actuals come in |
| Rolling forecast updater | Ingests actuals, rolls forward assumptions, outputs a refreshed forecast |
| Financial model | Three-statement or unit economics model connected to live data |
| Budget formats agent | Builds and populates budget templates from prior-year actuals and targets |
| Scenario modeller | What-if on key drivers — headcount, FX, churn, price — instant output |
| Performance dashboard | Live KPI dashboard connected to your data sources, refreshes automatically |
| KPI tracking agent | Monitors KPIs against targets, alerts on breaches, drafts commentary |
| Report integration & share | Assembles reports from multiple sources and distributes to stakeholders |
| Board pack generator | Pulls from data sources, populates templates, drafts narrative sections |
| Management reporting agent | Auto-populates standard management packs on close, every month |
| ESG reporting agent | Collects non-financial data, maps to frameworks, drafts disclosures |
| Cash flow forecasting agent | 13-week rolling forecast, pulls from bank feeds and AR/AP ledgers |
| Covenant monitoring agent | Tracks debt covenants against actuals, flags risks before breach |
| FX exposure tracker | Monitors multi-currency positions, flags concentration risks |
| Payment terms optimisation agent | Analyses AR/AP terms against cash position, recommends adjustments |
| Pricing anomaly detector | Flags deals outside standard terms before they hit the books |
| Customer profitability agent | Allocates costs to customers, ranks by margin contribution |
| Churn early-warning agent | Finance-side signals — payment delays, contract shrinkage — surfaced early |
| Cost centre overspend alert | Monitors actuals vs. budget by cost centre, alerts owners on breach |
Every sprint is scoped before we arrive. One 30-minute call is enough to design the session, pick the PoC, and agree the timeline.