Invoice validation
Every invoice checked against its contract and purchase order, with the expected total, variances and potential savings.
Let AI read the contracts, invoices and purchase orders, and keep every decision with your team.

What I build
Every invoice checked against its contract and purchase order, with the expected total, variances and potential savings.
Pricing, discounts, commitments and service credits pulled from contracts, each linked to its page and section.
PDF, Word, Excel and scanned documents read and turned into structured data, with OCR where needed.
Findings routed to the right person to approve or reject, with an audit log of who decided what.
Non-editable PDF reports with timestamps, versions, results and explanations, ready for auditors.
Revenue, collections and supplier accuracy in one view, from the same data the checks use.
AI, right now
Finance and procurement teams sit on piles of documents that decide real money: contracts, purchase orders, invoices. Checking each invoice against its contract properly is slow, so it mostly doesn’t happen, and overcharges go unnoticed.
AI can now read those documents reliably enough to do the first pass, as long as people stay in charge of the decisions. PayRight is built on exactly that rule, and OpsVanta shows the finance side of a full operations platform.
AI extracts the clauses and line items that matter, so people review findings instead of reading every page.
Pricing variances, quantity mismatches and missed discounts flagged before payment, not found in an audit later.
Each finding explained in a sentence an approver can act on, with a link to the source.
AI identifies, a person confirms, the system calculates. Nothing is approved or paid automatically.
Proof

RAG and AI chatbots
A UK contract compliance and invoice assurance platform. AI extracts contract clauses, people confirm them, and every invoice is scored against contract and PO.

MVP and SaaS development
An internal ERP built solo with Lovable and Claude Code on Lovable Cloud, running sales, projects, tickets, collections, finance and AI insights in one place.
Call
A 30-minute call about what you need, who it is for, and what success looks like.
Scope
A written scope: features, architecture, timeline, price range and risks.
Build
Working increments you can click through, not a big reveal at the end.
Ship
Live on your domain and hosting, with a handover so you know how it runs.
More in how I work, or see rag and ai chatbots.
Questions
No. It's a decision-support tool. AI extracts and compares, the system calculates, and a person approves or rejects every result.
Contracts, purchase orders and invoices as PDF, Word, Excel, CSV or JSON, including scanned copies, which go through OCR first.
Every extracted clause shows its source, people confirm clauses before they're used, and results are reproducible from the stored data. Before go-live, results are checked on a sample of your own documents.
Usually through the system's API or regular exports. Which systems, and how, is set in the scope after a call.
A 30-minute call is enough to tell whether it fits, what the first version should include, and how long it takes.