Internal knowledge search
Ask questions in plain language and get answers with links to the source documents.
Who it’s for
Who this is for and the problems it solves
Typical clients
- Teams that rely on large policy, technical or contract libraries
- Finance and operations teams processing invoices, forms and reports
- Support and sales teams who need fast, accurate answers from internal knowledge
- Healthcare organisations handling clinical and lab documents
Common problems
- People spend hours finding the right document or clause
- Data from invoices and forms is keyed in by hand
- General-purpose chatbots give confident answers with no source
- Knowledge is spread across drives, wikis, email and PDFs
Use cases
What we can build
Examples of our capabilities, not descriptions of specific client projects.
Invoice and form processing
Extract fields, validate them and send them for review or straight into your systems.
Contract and policy review
Find clauses, compare versions and flag items for a person to check.
Report processing
Turn scanned or PDF reports into structured data.
Scope
Scope, deliverables and what we need from you
What we deliver
- Document inventory and access-permission mapping
- Ingestion pipeline: OCR, parsing, chunking and indexing
- Retrieval and answer generation with source citations
- Extraction schemas and validation rules
- Search or chat interface, or an API for your own app
- Evaluation set and monitoring
Data & integration requirements
- A representative sample of documents, including difficult ones
- Who should be able to see which documents
- Example questions or fields to extract, with correct answers
- Where the output should go (your app, ERP, spreadsheet or database)
Delivery
How we deliver and evaluate
Discovery
Understand the workflow, the data and what success looks like. Agree scope and measurable targets.
Prototype & evaluate
Test the approach on a sample of your data against agreed criteria before a full build.
Build & integrate
Develop, connect to your systems and test with your team, with regular demos.
Deploy & support
Release to your chosen environment, hand over documentation and agree ongoing support.
We build a test set of real questions and documents with known answers, measure retrieval and answer accuracy against it, and review failure cases with you before launch. Answers show their sources so users can check them.
Evidence
Relevant work
Kubit’s product
BioHarmony Lab Report AI
Our own platform’s lab-report feature uses OCR and biomarker parsing to turn lab PDFs into structured data. Feature stages are team-reported; request a demonstration to confirm current availability.
Options
Ownership, deployment and support
Ownership
Ownership of the code, prompts, evaluation sets and any model weights we produce for you is agreed in the statement of work, subject to the licences of any third-party models or components used.
Deployment
Your own cloud account (AWS, Azure or Google Cloud), your on-premises environment, or hosting managed by Kubit’s.
AI models
Hosted models from major providers, open-source models run in your environment, or models fine-tuned on your data, chosen per use case for accuracy, cost and data handling.
Support
Post-launch support and maintenance are agreed per project, including monitoring, updates and response times.
FAQ
Questions about this service
What is RAG?
Retrieval-augmented generation: the system first finds the relevant passages in your documents, then writes an answer based on them and cites where they came from.
Can it respect document permissions?
Yes. We design retrieval so people only get answers from documents they are allowed to see.
Does it work with scanned documents?
Yes, using OCR. Quality depends on the scans, so we test on your hardest examples early.