Hungarian subsidiary of an international insurance group
AI Applications & Integration
RAG-based knowledge search for claims handlers
Our role: End-to-end delivery · from ingestion to integration into the claims application
DetailsArea 02 — AI Applications & Integration
We bring AI built on your enterprise knowledge and data into day-to-day operations.
We don't build demos; we build AI that people use every day. It draws on your own documents and data, handles access rights, has measurable quality, and is built into your existing systems.
Area 02
Sound familiar?
Head of AI / AI CoE
From PoC to production
- Several GenAI PoCs run in parallel, but none has reached production and measurable business impact.
- There is no common evaluation framework; answer quality can only be judged subjectively.
- The RAG solution works in the demo but is inaccurate on the full document estate with real permissions.
- AI agents get stuck when they have to fit into existing workflows and core systems.
CISO / Compliance / Risk
Risk and compliance
- Staff use unapproved AI tools, and nobody can see what data goes into them.
- Running use cases lack risk classification and documentation under the EU AI Act and GDPR.
- It is unclear who is accountable for a wrong model answer: the business, IT or the vendor.
- Prompt injection and data leakage controls are not part of the development process.
CIO / Business leader
Value and operations
- GenAI running costs grow, but the return cannot be broken down by use case.
- Pilots started from an innovation budget; production operation has no owner and no budget.
- AI access to core system data and permissions has not been solved.
- Everything depends on a single model provider; switching and version-upgrade risk is unmanaged.
AI Applications & Integration
How we help
01
Knowledge search and RAG
Answers from your own documents and data, with source citations and access control.
02
Impact analysis
We map the connections between systems, data and processes, so you can see what a change affects.
03
Process automation
We hand individual steps of customer service, admin and internal processes to AI agents, with human approval.
04
From pilot to production
We prepare an existing AI prototype for enterprise use: quality measurement, security, monitoring, cost control.
05
Integration
We build the AI feature into your existing applications, APIs and workflows.
06
Quality measurement and guardrails
Test questions put together with you, measurable quality thresholds, checked outputs and safety constraints.
AI Applications & Integration
How we start
Scoped project
Enterprise knowledge search or impact analysis
One well-defined use case, with data sources, access rights and quality expectations agreed in advance.
- What we start with
- One type of question and the related documents and data sources
- What you get
- A working system with measured quality in your own environment, handed over to your team
- Who works on it
- Experienced AI and data engineers, with your experts
Assessment
Can the pilot go to production?
We review the existing prototype – data, quality measurement, security, integration, cost – and tell you what it needs to go live.
- What we start with
- An existing AI prototype or pilot
- What you get
- A gap list, risks and a step-by-step plan to production
- Who works on it
- An experienced AI engineer and architect
Technology
RAG (answers grounded in your own sources), knowledge search, impact analysis, process automation, system integration.
AI
Language models, RAG, embeddings, vector databases
Agents
Multi-agent systems, Google ADK
Data
Databricks, Spark, Airflow, Python
Integration
REST APIs, microservices, event-driven systems
Case studies
All case studies Hungarian energy supplier
AI Applications & Integration
AI-based impact analysis for data warehouse changes
Our role: Joint delivery with the client's data team · knowledge graph and query interface
DetailsFAQ
AI Applications & Integration
Which language model do you use?
That depends on the use case, the cost and the data handling rules. We build the system so that the model can be swapped later.
How do you prevent the AI from exposing sensitive data?
We handle access rights at the level of the source systems: everyone only gets answers from sources they already have access to. All of this runs within our ISO/IEC 27001 certified information security management.
How do we know it works well?
At the start we put together a set of test questions and quality thresholds with you, and measure the system against them – before and after rollout.
We already have a pilot. Can you continue it?
Yes, it is often the best starting point. We review it and make it production-ready instead of starting from scratch.
The first step
Work with us
Bring a specific data or AI problem.
In 30 minutes, one of our experienced engineers talks through the situation with you and what it would take to solve it. Not a sales pitch.
Data platform and AI feasibility assessment
For one selected business need, we look at which data and integrations are needed, what is missing, and in what steps it makes sense to deliver it.
Enterprise knowledge search or impact analysis
We deliver one well-defined use case, with data sources, access rights and quality expectations agreed in advance.