From data silos to AI agents: data intelligence in medium-sized businesse

In the operational reality of many industrial companies, failures of critical components often still lead to manual, time-consuming processes: telephone calls, email correspondence and the manual comparison of Excel lists characterise everyday work. This ties up valuable capacity.

We are currently at a technological turning point. The era of purely descriptive analytics – i.e. viewing past events in dashboards – is being expanded by agentic AI. These are systems that not only visualise data, but also autonomously derive and execute options for action based on that data.

Scenario 1: Automated procurement processes

An AI agent in a modern data intelligence environment acts proactively. If a sensor detects an impending stock shortage, the system does not wait for human intervention. The agent checks the stock level in the SAP system, compares supplier prices in real time and evaluates delivery times based on historical logistics data. The result is a prepared order record that only needs to be approved by the purchasing department.

This relieves specialists of routine administrative tasks and speeds up the supply chain.

Scenario 2: Digitisation of expert knowledge in plant engineering

Another area of application is the preparation of quotations for complex specialised plants. Calculations often depend on the implicit experience of less experienced employees who refer to similar projects from the past. However, when this generation retires, a critical knowledge gap often arises, as this experiential knowledge is rarely documented in a structured manner and thus leaves the company.

An AI agent on the Databricks Data Intelligence Platform can make this implicit knowledge explicitly usable and preserve it. It analyses historical CAD drawings, technical specifications (PDFs), spreadsheets and project communications from recent years.

The system recognises patterns: ‘The requested module is 90 per cent similar to the 2018 project, although the material costs for steel are 20 per cent higher today when adjusted for the steel index.’

The agent generates a data-supported calculation draft. This does not replace the work of engineers, but it reduces research time from days to minutes and permanently safeguards valuable company knowledge against staff turnover.

The technical foundation: the data intelligence platform

As promising as these scenarios are, they often fail in practice due to the data infrastructure. An AI agent that accesses fragmented CSV exports and data silos cannot make reliable decisions (the ‘garbage in, garbage out’ principle).

Databricks addresses this problem with its lakehouse architecture (Data Lakehouse Architecture | Databricks), which serves as a central integration point:

• Uniform database: Structured financial data from the ERP and unstructured maintenance logs are consolidated in one place. This gives the agent the full context.
• Semantic understanding: The platform uses AI to index the structure and meaning of company data, reducing manual optimisation steps.
• Mosaic AI (Databricks GenAI Announcements at Data + AI Summit 2024 | Databricks Blog) and Agent Bricks (Agent Bricks | Databricks): Databricks provides a modular framework for developing, testing and monitoring agents in operation according to standardised patterns.

Governance with Unity Catalog: Compliance by Design

A key obstacle for AI projects in medium-sized businesses is data security. Technical measures must be in place to ensure that sensitive data (such as HR information or sales figures) is not processed without authorisation.

Databricks' answer is Unity Catalog (Unity Catalog | Databricks). It acts as a central governance layer. When implementing agent solutions, access rules are defined at this point: for example, a service agent may read product data but is not granted access to personnel data.

This governance is enforced centrally, regardless of whether access is by a data scientist, an analyst or an AI agent. This enables the legally compliant use of RAG (Retrieval Augmented Generation) in a corporate context.

Cost efficiency through serverless architecture

In terms of operating costs, Databricks offers a flexible model with Serverless Compute (Serverless computing | Databricks). Resources do not need to be permanently reserved. If an agent is inactive, there are no compute costs. During peak loads – such as checking thousands of documents at the end of the month – performance scales automatically. This makes enterprise technology calculable and economically efficient even for medium-sized businesses. (You can read more about this in our blog on FinOps: DBU optimisation & TCO calculation for DWH migration).

How adesso is shaping the path to the ‘Agentic Enterprise’

The technology is the basis, but integration into the existing system landscape (SAP S/4HANA, Salesforce, legacy systems) is crucial.

At adesso, we combine technical expertise with an understanding of processes for medium-sized businesses. We not only implement the platform, but also realise use cases with measurable ROI:

• Intelligent document processing: Automated validation of incoming invoices and delivery notes.
• Predictive maintenance agents: Autonomous fault reporting by machines, including proposed solutions and spare part identification for the technical team.

The Data Intelligence Platform is ready. Is your data strategy?

Deep dive topics: The series at a glance

This article outlines the overall strategic picture. In our blog series, we delve deeper into specific aspects of the platform for different roles. Here are the direct links to the deep dives:

• For IT management & SAP architecture: Liberating SAP data: Seamless integration into the lakehouse without replication effort
• For CISOs & compliance management: Unity Catalog and open source: Compliance & security for SMEs
• For innovation managers & AI engineers: Agent engineering instead of gimmicks: The path to secure AI applications with Mosaic AI
• For data engineers & IT ops: Databricks Lakeflow & Lakebase: Architecture simplification for data engineers 2026
• For CFOs & cloud managers: Databricks FinOps: DBU optimisation & TCO calculation for DWH migration

Conclusion

The Databricks Data Intelligence Platform provides the technological foundation – adesso provides the strategy for integration. Let's evaluate how you can modernise your data landscape and transform isolated solutions into a unified, intelligent platform. Our experts accompany you from architecture planning to productive AI deployment to automate processes and shorten innovation cycles.