Applied AI / Capability
Document data extraction.
Pull fields from invoices, forms and reports into your systems, with validation and a review queue.
Describe the task
When to bring us in
Sound familiar?
- Our team types the same fields from PDFs into our system every day.
- Documents arrive in a dozen different layouts from different suppliers.
- We tried an extraction tool, but couldn’t tell which results to trust.
The cost of leaving it
The same fields get typed from PDFs every day, and mistakes travel quietly into reports and payments. Adding a new supplier layout means more manual work, not less.
Why us
What we bring to document data extraction.
A person stays in charge
Models work on bounded tasks inside your systems, and anything that matters goes to a named person to confirm.
Tested on your examples first
We measure a model on real cases from your work before it reaches staff, so you see where it fails while the stakes are low.
We say when AI is the wrong tool
Not every task should be automated. When conventional software or a person is the better answer, we tell you.
What the work includes
Read the document once, then check what matters.
We define the fields you need and the rules each one must pass, such as formats, totals and matches against existing records. Extracted values are validated before they go anywhere. Anything that fails a check or falls below an agreed confidence level goes to a review queue, where a person confirms it against the source.
Field and rule definition
The fields to extract and the validation each must pass before it is accepted.
Extraction pipeline
Text and layout read from scans, PDFs and images, then mapped to the defined fields.
Confidence and validation checks
Format, total and cross-record checks that decide whether a value needs a person to look.
Review queue
A screen where reviewers see the source next to the extracted values and correct them.
Use cases
Projects we take on.
Supplier invoices into an accounting system
Line items, VAT and totals read from supplier invoices and checked against purchase orders before they are posted.
Onboarding forms for a regulated firm
Client details pulled from application forms into the onboarding record, with missing or conflicting answers flagged for follow-up.
Certificates into a property record
Safety certificates read for dates and results, so the ones about to expire show up before they lapse.
What you get
Outputs you can review.
- Field definitions and validation rules
- Evaluation set of representative documents
- Extraction service connected to your systems
- Review queue for flagged documents
- Accuracy report by field
Technologies we use for this
Where it matters
- Health & social care Staff can look up care policies and procedures with the source shown; decisions about a person’s care stay with people.
- Property & field services New repair requests, with their descriptions and photos, can be sorted by trade and urgency before an operator confirms dispatch.
- Fintech & professional services Statements and application forms can be read into your systems field by field, and any value below a set confidence goes to a reviewer.
What’s changing
Shifts to plan for.
July 2026
EU high-risk rules move to 2027
The AI Omnibus entered into force on 27 July 2026, moving obligations for Annex III high-risk systems, such as those used in employment, credit scoring and education, to 2 December 2027.
August 2026
EU AI transparency duties apply
Article 50 obligations, such as telling people when they are dealing with an AI system, apply from 2 August 2026.
February 2026
New UK rules on automated decisions
The Data (Use and Access) Act 2025 widened when significant decisions can be automated, provided people are told, can contest the outcome and can ask for a human review.
Process
How the work runs.
Collect real samples
We gather documents as they actually arrive, including poor scans, rotated pages and handwritten notes.
Agree the answers
Your team marks the correct values on a sample set, which becomes the yardstick for accuracy.
Tune checks per field
Thresholds are set field by field, so a doubtful total goes to review while a clear reference number goes straight through.
Connect and watch
Accepted values flow into your system, and review rates are tracked so a change in incoming documents gets noticed.
Cost
What shapes the estimate.
Layout variety
Ten suppliers with ten layouts take more work than one standard form.
Scan and image quality
Faxed, photographed or handwritten documents need more tuning and review than clean digital PDFs.
Where the data goes
Writing into a system with a good API is quicker than one that needs file imports or a workaround.
Ways to work together
A first, fixed stage
Discovery, a prototype or a technical review with agreed outputs, so the next decision is made with something concrete in hand.
Delivery in phases
The build split into releases, each with a scope you review and approve before it starts.
Ongoing engineering
An agreed scope for support and improvements after launch, with response expectations written down.
Common questions
Before you get in touch.
How many example documents do you need?
Enough to cover the layouts and quality you actually receive, including the messy ones. We agree a representative set with you at the start.
Where are our documents processed?
We agree which models and hosting options are acceptable for your data, including cloud region and retention, before any documents are processed.
What affects the effort?
The number of document types and layouts, scan quality, how many fields are needed, the validation rules and the systems the data goes into.