AI & automationDocument & data AI
The PDFs read themselves. The numbers add up without a spreadsheet marathon.
Invoices, work orders, contracts, inspection reports, and the data trapped inside them. AI reads them, pulls out what matters, checks it against your rules and puts it where it belongs. Then a dashboard shows you what the data says while it is still true.
If any of this is you
- Supplier invoices keyed in by hand and matched by eye.
- Contracts skimmed for the one clause that matters.
- Field reports nobody reads until something goes wrong.
- A monthly dashboard that is out of date the day it is built.
What is included
Reading, checking, routing and reporting.
Nothing is written into a system of record without passing your rules, and anything uncertain goes to a person first.
- Document intelligence
- Reads, extracts, classifies and routes information from contracts, invoices, reports and PDFs, at any volume.
- Invoice and work order processing
- Supplier invoices matched to orders, coded and queued for approval. Exceptions flagged with the reason.
- Contract review
- Key clauses, dates, obligations and risks pulled out and summarised for a person to decide on, in minutes rather than an afternoon.
- Report summaries
- Field and inspection reports summarised, anomalies flagged and routed to the right team the day they are filed.
- An assistant that knows your documents
- Staff ask a question in plain language and get an answer that cites the policy, manual or contract it came from.
- Dashboards on live data
- What is happening now and what is coming, built on data from your systems rather than a spreadsheet someone updates.
For your technical lead
- Models
- OpenAI and Anthropic APIs, chosen per document type
- Retrieval
- Your documents indexed so every answer cites its source
- Validation
- Extracted values checked against your rules before anything is written to a system
- Human review
- Confidence thresholds route uncertain items to a person
- Data
- Stays in your tenant and your accounts
- Dashboards
- Built on live connections to your systems, not exports
What it costs
Free to find out. Fixed price to build. Monthly to keep it working.
How it works
Start with the pile that hurts most.
Data preparation is a real step, not a footnote. Most of the work in the first weeks is getting your documents and data into a shape the AI can rely on.
Week 1, free
Assessment
We look at what arrives, how much of it, and what it costs to handle by hand. You see the top three document flows in writing.
Weeks 1–3
Design and data preparation
Extraction rules, validation rules and the systems it writes to. Sample documents cleaned and structured. Fixed price agreed.
Weeks 3–6
Build and test
Built and run against your real backlog. Accuracy measured against what your team would have done, before it goes live.
Week 6 onward
Launch and improve
Live with human review on anything uncertain. Thresholds tightened as accuracy is proven, and retrained as your documents change.
What it costs
Free to assess. Fixed price to build. Monthly to keep it accurate.
Documents change format, suppliers change templates, policies get rewritten. The retainer keeps extraction accurate and the assistant current.
What the assessment involves- AI readiness assessment
Free
One week. We map where your team's time goes and hand you the top three places automation pays back, in writing. No obligation.
- Build
Fixed price, quoted in writing
Scoped and priced after the assessment, before any build starts. A typical build reaches launch in about seven weeks.
- Ongoing improvement
Monthly retainer, from a scope agreed at launch
Monitoring, retraining and changes as your business and data change. This is where the return compounds, and it is why we treat AI as an engagement rather than a project.
The assessment is free. Builds are quoted at a fixed price before work starts. Ongoing improvement is a monthly retainer. Final pricing depends on the systems involved, the state of your data and the integration scope, and is confirmed after the readiness assessment.
Fair questions
What finance and operations teams ask us.
- Can we trust what it extracts?
- Every value is checked against your rules, anything below a confidence threshold goes to a person, and every extraction is traceable to the page it came from. Accuracy is measured against your team's own work before launch.
- Our documents are a mess.
- That is normal, and it is why data preparation is a step in the process rather than something we hope you have done.
- Is our data used to train someone else's model?
- No. We use commercial API terms that exclude training, your data stays in your accounts, and we put that in writing.
- We do not have a data team.
- You do not need one. You need the people who know what a correct invoice looks like, for about two hours during design.
Start with a free AI readiness assessment.
One week, no obligation. We look at where your team's time goes and hand you a short written report on where AI would pay back first.
- A map of where the hours go in your business, by team
- The top three automation opportunities, ranked
- A fixed-price proposal for the first build, or an honest 'not yet'
Prefer to talk? Get in touch