About the organisation
Sector
Life sciences & biotechnology
Size
International scale-up with worldwide sales and support
Challenge
Multiple versions of the truth
Desire to use AI on business data
Doubts about data quality
Solution
dgl Business Solution (DAGONGIN Agentic AI platform)
The technical foundation was in place, the semantic layer was missing
LUMICKS has its technology in order: modern infrastructure, systems that talk to each other. What was missing was the layer above it that records what a concept means and where it is located: the ontology. Clear definitions, with owners who decide on them.
Data from Salesforce, Power BI and Excel tell what has happened, but not what the organisation understands by it. Specifically:
- Core concepts such as prospect, campaign and order amount were interpreted differently across systems and departments – multiple versions of the truth
- In management reports, the same question kept coming back: is this figure correct, and according to which definition?
- Definitions did exist, but scattered across documents, dashboards and people’s heads – recorded nowhere, without an owner, and therefore unusable for software
- The desire to use AI on the company’s own business data was there. But AI on raw data inherits that ambiguity – and building it manually takes months that were not available
How this affected the organisation
- Finance reconstructed and verified figures instead of interpreting them
- The ambition to use AI remained an ambition – there was no semantic layer
What was at stake
Making decisions based on figures without consensus means deciding with a blind spot. And AI on raw data gives structurally unreliable answers: the model has no reference for what is correct.
On top of that comes external pressure: GDPR, the EU AI Act and CSRD require an organisation to be able to show how a decision was made. That requires recorded definitions with a responsible owner. And a scale-up has little margin to reverse wrong decisions: it has to be fast and right.
“We could disagree for a long time about what ‘order amount’ actually meant. Now it’s simply written down – and someone has put their name to it.”
Michiel Hoedemakers - IT Business Analyst
Why dgl?
A solution that starts from the business, a true business solution and not just another technical tool.
Dgl’s approach revolves around one principle: let AI do the work AI is good at, and let the business decide what is true.
AI works at machine speed, the business stays in control. Taking inventory and making proposals takes hours; what is correct is decided by the organisation – not by the model.
AI does the work nobody wants to do
Nobody wants to pick apart Excel files, dig through Power BI models or reconstruct report definitions. Slow, tedious, error-prone – exactly where AI excels. The DAGONGIN software reads dashboards, reports, data models and metadata, with existing definitions and taxonomies as a starting point, and proposes domains, definitions, business rules and ownership.
The business stays in control
Every proposal is reviewed by the business. The agent guides the user step by step: it makes a proposal, explains where it comes from and asks for a judgement – approve, refine, or reject with an explanation. That is human in the loop in practice: the machine executes, the human assesses and thus stays in control.
What DAGONGIN does
From source systems to answers that are correct.
Salesforce, Power BI en Excel
The software reads fields, tables, relationships, dashboards and reports; existing taxonomies and definitions are included as input.
Meaning, recorded
The agent proposes domains, definitions, business rules and ownership: what a concept means and where it is located. The business approves. The result is machine-readable.
Question, report and dashboard
AI combines the figures from the source systems (read-only) with the meaning from the approved layer. Every answer, every report and every dashboard delivers reliable outcomes. Fast and traceable.
The added value of the middle step. An AI model that runs directly on raw data guesses at the intent. An AI model that runs on a validated ontology knows the intent – and can show which approved definition the answer is based on. That is where reliable AI for business data begins
What dgl delivered
An ontology in a short timeframe
Existing taxonomies, data models and reports form the input. From these, the agent builds up the definitions: what a concept means and where it is located in Salesforce, Power BI and Excel. The business reviews the definitions. By using the platform’s AI functionality in this phase, the ontology layer is created in a short timeframe.
From ontology to reporting and dashboard
The first reports for Sales work directly with the approved layer. A report definition that took weeks of work in Power BI is ready via the agent in ten minutes. Because AI does not have to derive the business rules again, analyses and dashboards follow at the same pace.
Data quality control
DAGONGIN monitors data quality and flags where the ontology is missing or ambiguous. This makes it visible at a glance which figure is fully reconciled and which figure still needs attention from an owner.
A blueprint for what follows
The ontology layer for Sales serves as a blueprint for the underlying data structure and for every subsequent domain – not a one-off project, but a layer that grows with the organisation, maintained by the owners themselves.
The result
From technical foundation to working reports. By making smart use of the platform’s AI functionality, delivered in a short timeframe.
The business value
From technical foundation to working reports. By making smart use of the platform’s AI functionality, delivered in a short timeframe.
Speed. A report definition that took weeks of work in Power BI is ready via the agent in ten minutes. That pattern applies more broadly: complex development work that previously required specialists and lead time is now done by AI at LUMICKS. The question shifts from ‘when can we start’ to ‘what do we tackle first’.
Reliability and ownership. Every answer is based on a definition that the business itself has approved. Ownership lies where the knowledge is: with the business, not with IT. This makes data-driven management possible, and makes it demonstrable afterwards where a figure comes from.
Scalability. What has been built for Sales is the blueprint for the rest of the organisation. Every subsequent domain builds on the same layer instead of starting over.
A simpler landscape. Because reports and dashboards work directly with the approved layer, part of the existing reporting landscape becomes redundant. Expensive tooling and licences can be phased out.
In short
✅ Within a few weeks, a working ontology layer, built via the agent and approved by the business
✅ Analysis of existing systems, reports and metadata at machine speed – the business only needs to validate the proposals
✅ Reliable reports for Sales based on approved definitions, delivered within the same project
✅ One shared, machine-readable meaning of core concepts across Salesforce, Power BI and Excel – AI and reporting tools read from the approved layer instead of from raw sources
✅ Answers to data questions are traceable to an approved definition with an identifiable owner
✅ Assured data quality that is directly visible in reports thanks to clear labelling, explanations and quality scores. Including action-oriented recommendations.
Do you want reliable AI for business data, without giving up control?
Do you recognise the doubts about the reliability of AI outcomes? Discover what an ontology layer validated by the business can bring your organisation.