Software and AI for financial services
We build AI in financial services for the operational work behind every account, policy and ledger: onboarding checks, claims, alerts and bookkeeping. People with the licence and the mandate make the decisions, and every step is recorded for auditors and supervisors.
Industries in financial services
The work, and where software fits.
Banks, insurers and accounting firms process large volumes of documents and transactions under close supervision. Much of the work is gathering facts from several systems, checking them against rules and writing down why a decision was made.
Software helps most in that gathering and writing. We build document processing, investigation agents, policy search and integrations with core platforms, each with evals, access controls and an audit trail. Analysts, handlers and accountants review the output and keep ownership of the outcome.
How we improve financial services.
Onboarding and intake documents
Identity papers, company extracts, claim files and invoices read, cross-checked and structured before a person reviews them.
Investigations with the evidence attached
Agents that assemble data for AML alerts, claims and approvals, and draft a note that cites every source.
Search over policies and conditions
Internal policies, insurance conditions and tax guidance searchable by question, with answers tied to the approved version.
Integrations around core platforms
APIs and pipelines connecting core banking, policy administration and accounting packages to new products and reporting.
Evals and audit trails for supervisors
Test sets, decision logs and documentation that let risk, compliance and internal audit see how each tool behaves.
Related work.
- Financial services
Open-source document parsing API and MCP server for financial documents
210+API endpoints- Document parsing
- OCR
- MCP server
- Finance operations
AI invoice approval agent that cites the policy clause behind every decision
100%of automatic decisions cite a policy clause- AI agents
- Human in the loop
- RAG
- Insurance
How LLM evals let an insurance claims assistant ship weekly at 94% accuracy
94%answer accuracy, tested on every release- LLM evals
- Langfuse
- Promptfoo
Further reading.
LLM evals: how to test AI features before every release
A practical approach to LLM evals: build a test set from real cases, combine code checks with model grading, and block releases that regress.
5 min read
Human-in-the-loop AI agents: design patterns for production
The patterns we use to run AI agents safely in production: approval checkpoints, confidence thresholds, tool permissions, audit trails and fallbacks.
4 min read
Document parsing with LLMs: extracting data from invoices and forms
How to turn invoices and forms into reliable structured data with LLMs, schemas, validation and a review queue for the cases that need a person.
4 min read
Common questions.
We design software for ICT risk management, incident logging and resilience testing, and supply the contract terms, security information and exit plans your third-party register needs. Your organisation keeps responsibility for its DORA framework.
In what we build, a qualified person makes those decisions. The software gathers information, drafts and flags, and each decision is logged with the evidence it was based on.
Yes. We use EU-hosted model providers or deploy inside your own cloud tenant, and we can mask personal and account identifiers before text reaches a model.
We build the workflows, AI and integrations around the core platform and agree interfaces and hosting with your vendor and IT team. Code and documentation are handed over so they can support it too.