Skip to content
DisruptX

Services · AI and automation

We add AI where it saves real time: assistants that answer from your own data, document processing, and automation of repetitive back-office work. Every feature is tested on real examples before it ships, and we will tell you when simpler software would do the job better.

Start a project All services

What we do

Assistants, document processing and workflow automation, added where they save real time and measured before they ship.

AI is useful for some jobs and a costly distraction for others. We start with the workflow and the number that would prove it worked, and we will say so when a rule, a form or a report would do the job better. When AI is the right tool, we build it with the same testing, logging and cost controls as the rest of the product.

Assistants in your product

Search and chat features that answer from your own data, with sources shown and a clear reply when the data doesn't say.

Document processing

Extraction and classification for contracts, invoices and forms, with a review step for anything uncertain.

Workflow automation

Repetitive back-office steps handled by automation or an agent, with a person approving anything expensive to get wrong.

Measured quality

A test set of real examples agreed before the build and scored on every change, so quality can't slip quietly.

Cost you can forecast

Caching, smaller models for simple requests and budgets per feature, so running cost is known before launch.

Data boundaries

Clear rules on what leaves your systems, providers that don't train on your data and self-hosted options where policy requires it.

What you get

Deliverables, not just hours

  • A written view on whether AI is the right tool for the job
  • A test set of real examples and a quality score for every change
  • A cost per request estimate before the build starts
  • Logging and approval steps for anything expensive to get wrong

Tools we use

Chosen for the job and your team

  • Node.jsAPIs and real-time services
  • PythonData work, integrations and AI pipelines
  • FastAPITyped Python APIs
  • PostgreSQLOur default database, for good reasons
  • RedisCaching, queues and rate limits
  • GraphQLWhen many clients need flexible data
  • OpenAI and AnthropicHosted models for most features
  • Open-weight modelsLlama and Mistral, when data must stay in house
  • pgvectorSearch over your data without another database
  • LangGraphMulti-step workflows that can pause for approval
  • EvalsReal examples scored on every change
  • LangfuseTraces and cost per request

Related work

A project like this

Contract page with highlighted clauses beside a list of extracted terms and confidence scores

Professional services · AI document tool · 11 weeks

Contract review that keeps an analyst in charge

A services firm had analysts reading 300-page supplier contracts to find the same twelve clauses every time. The work was slow, expensive and easy to get wrong late on a Friday.

We built a review tool that extracts the clauses with citations back to the source page, scores its own confidence, and routes anything uncertain to an analyst queue. A set of real contracts is rerun on every change to catch quality slipping.

-70%
Analyst hours per contract
100%
Extractions with citations
Weekly
Automated quality runs

Questions

Common questions

Which AI models do you use?

We choose per task based on test results and cost: usually hosted models from OpenAI or Anthropic, or open-weight models when data must stay inside your own infrastructure.

Will our data be used to train AI models?

No. We use providers and settings that exclude your data from training, and we document exactly what data leaves your systems before anything is built.

How do you make sure the AI gives correct answers?

We agree a set of real examples and expected results before the build, score every change against them, show sources with each answer, and send uncertain cases to a person.

More services

Often combined with

Next step

Have a project in mind?

Tell us what you're building or what's holding you back. We'll come back with an honest approach, a rough timeline and a price range.

Start a project