Process and scenario analysis
We describe the agent’s tasks, inputs, decisions and exceptions, and agree which steps require human approval.
AI services
AI agents for business
We design and deploy AI agents for business – from an AI assistant for your company that answers your team’s questions, to agents that complete tasks in your ERP, CRM and other systems. With permission control, human oversight and cost monitoring.
When a company needs AI agents
An AI agent is more than a chatbot. It can understand an instruction, plan the next steps, access company data and complete a task in a system: update a CRM record, prepare a quote, change an order in the ERP or produce a report.
Building AI agents makes sense where employees perform repetitive work that still requires a bit of “intelligence” – reading a document, making a simple decision, moving information between systems. Classic automation cannot cope with non-standard cases, while manual work is expensive and error-prone.
We design agents to work securely: they only access the data they need, important operations can require human approval, and every action is logged.
Typical tasks for AI agents
- Handling enquiries and preparing quotes based on CRM data.
- Entering orders and documents into the ERP.
- Classifying and routing emails and tickets.
- Preparing reports and summaries from multiple sources.
- Helping employees find information in procedures.
AI agent development – service scope
We build AI agents on proven frameworks and connect them to your systems via APIs and the Model Context Protocol (MCP).
AI assistant for your company
We build an assistant that answers your team’s questions based on company documents and data, and shows the sources of its answers.
Task-performing agents
Agents work in ERP, CRM, ticketing systems or email – creating records, updating statuses and preparing documents.
MCP and API integrations
We give agents controlled access to company tools and data, with permissions limited to the necessary minimum.
Security and oversight
We design roles, limits, operation approval, action logging and protection against unwanted instructions.
Quality and cost monitoring
We measure agent effectiveness and model usage (e.g. in Langfuse) to keep quality high and costs predictable.
How we build AI agents
We start with one well-defined process. Once the agent proves its effectiveness, we extend its scope or build further agents.
Consultation and process selection
We choose the process with the greatest potential and a clearly defined business outcome.
Agent design
We define tasks, tools, data access, escalation rules to a human and success criteria.
Prototype and testing
We build a prototype and test it on real examples from your company, measuring the accuracy of its actions.
Integration and rollout
We connect the agent to your systems, roll it out to a selected team and train users.
Monitoring and growth
We analyse the agent’s actions, improve quality, optimise costs and extend its scope.
Technologies
Frameworks and models for AI agents
We match models and frameworks to the task. We build agents with frameworks such as Microsoft Agent Framework and the OpenAI Agents SDK, and give them access to company tools and data through the Model Context Protocol. We monitor agent quality and costs in Langfuse.
- OpenAI
- Anthropic
- Microsoft Agent Framework
- OpenAI Agents SDK
- Model Context Protocol
- Langfuse
- Python
- Node.js
- PostgreSQL
- Redis
Example use case
An AI agent that prepares sales quotes
The sales team of a B2B company receives requests for quotes by email. Sales reps manually check the customer in the CRM, price lists and availability in the ERP, and then put the quote together in a template.
An AI agent reads the request, identifies the customer and products, retrieves commercial terms and availability, and prepares a draft quote and a CRM task. The sales rep reviews the quote and sends it with one click. Unusual requests are flagged by the agent for manual handling.
An illustrative scenario showing the typical course and scope of such a project.
Frequently asked questions
How is an AI agent different from a chatbot?
A chatbot answers questions. An AI agent also plans steps and completes tasks in your business systems – for example, creating a CRM record or preparing an ERP document – within the permissions it has been given.
Can an AI agent work without human oversight?
It can, but it should not always do so. We design important operations to require employee approval, and every agent action is logged.
Which systems can AI agents work with?
Any system that provides an API or database – including ERP, CRM, WMS, ticketing systems, email and document repositories. We use the Model Context Protocol to expose tools to agents.
Which AI models do you use?
We match the model to the task and security requirements – including OpenAI and Anthropic models, run in the cloud or in your own environment.
How do you control the cost of running AI agents?
We monitor token usage and answer quality, match models to tasks and use caching and limits. See our AI cost optimisation service for details.
Experience since 2009, now powered by AI
Since 2009
17 years of building enterprise-class web applications, custom business systems and mobile apps.
Industry know-how
Many projects for automotive, logistics, healthcare, food service and catering, construction and finance.
From idea to maintenance
Consulting, design, development, integrations, rollout, training and ongoing growth in one place.
Data security
Solutions designed in line with GDPR and the EU AI Act, with access control and transparent costs.