Credit Assessment
The full credit file, assembled and analyzed: spreading, policy checks, and a committee ready memo where every figure is cited to its source document.
How an engagement runs
Three steps, the same for every offering on this page.
1
Pick
A readiness check, a short list of candidate use cases, and an honest view of which one pays back first. Then a pilot on files you have already decided, against success criteria agreed up front.
2
Build
Inside your premises, on your data. Your policy is encoded as verifiable checks, your IT stands up and runs the system, and you evaluate output against your own files, not a demo of someone else's configuration.
3
Own
Handover, the playbook, and the knowledge to extend it. Your team runs the system and takes it to the next workflow. Nothing stays dependent on us.
First AI Use Case
For the organization whose board has decided to move on AI and whose first working use case does not exist yet. We turn the directive into one live system: we select the use case with you, build it on your premises, and hand it over to your team.
The output is not a strategy deck. It is a running system, the playbook behind it, and a team that knows how to take it further.
Proof: Credence, our AI credit investigator, is a first use case taken all the way to a working system. See what a finished one looks like.
What's included
- AI readiness check: data, systems, people, and constraints
- Use case shortlist with an honest payback view
- Pilot on files or cases you have already decided
- Build inside your premises, on your data
- Your policy encoded as verifiable, auditable checks
- Handover, playbook, and training for your team
Financial Services AI
Where we start with banks and lenders: three document heavy workflows we know deeply. Credit assessment runs on Credence. KYC and CBE report augmentation run on the same engine, reconfigured for a different workflow: the same evidence and citation layer, the same on premise deployment, the same rule that the analyst makes the final call.
Proof: Credence is running in a fully Arabic configuration. Watch the three minute demo.
KYC & Customer Due Diligence
Verification of customer documents at intake, before the back office: completeness, required attachments, signatures and stamps, consistency with the application.
CBE Report Augmentation
The AI layer around regulatory reporting: circular tracking, impact against your own policies, and the narrative sections. We do not touch core report generation.
Custom AI Development
For the use case that does not fit a ready made engine. Retrieval pipelines over your documents, agents that carry a workflow end to end, and LLM applications for your staff or your customers, taken from prototype to a system your team can run.
We measure before we ship. Every retrieval and generation step gets an evaluation set built from your own data, so you know what the system gets right, what it gets wrong, and by how much.
Proof: how we built a retrieval evaluation framework for a credit investigation agent. Read the five experiments.
What's included
- RAG pipelines over your documents, in Arabic, English and Dutch
- Agentic systems that run a workflow end to end with human sign off
- LLM applications for internal teams and customer facing use
- Evaluation sets and quality measurement on your own data
- Vendor neutral model choice: Azure OpenAI, Anthropic, or open models
- Deployment on premise or in your own Azure or AWS tenant
MCP & AI Integrations
Your people already use ChatGPT, Claude or Copilot. The question is how those assistants reach your data without a copy of it leaving the building. We build Model Context Protocol servers on top of your core systems, with access control, versioning and an audit trail, so an assistant can answer from live data with the same permissions the user has.
Proof: Plekundig MCP, the first MCP server built for the Dutch real estate market, is live in production. Read how we built the secure gateway.
What's included
- Custom MCP servers on your core systems and data stores
- Fine grained access control, versioning and audit logging
- OAuth gateway and enterprise grade authentication
- ChatGPT, Claude and Copilot connected to your data
- Schema aware and schema free data access for conversational AI
- Governance: what an assistant may read, and what it never sees
What a first engagement looks like
Four things that are true of every engagement we take on.
The first conversation is freeThirty minutes. We tell you honestly whether AI can take the workflow.
The pilot runs on decided filesYou see what AI does on your own numbers before any commitment.
Built with your team, not for themYour officers put the policy in, your IT runs the system.
You own the resultThe playbook, the logic and the system stay with you.
Tell us where the work piles up
Where the files come back, where people read documents all day. We will tell you honestly whether AI can take it, and then take it to reality.