From specification to approved comparison report. Without the manual work.

Dundir removes the administrative work before the decision and documents the reasoning process after it. The procurement officer reviews and decides. The system calculates and records.

How it works

You enter a material specification
In plain language, the same way you would describe it now.
The AI layer translates the specification
The system converts the specification into structured procurement requirements: required certifications, delivery windows, budget limits and weighted objectives.
The optimisation engine calculates the best supplier combination
The system runs against your supplier database and finds the mathematically demonstrable best combination based on your hard requirements and weighted objectives. Not a suggestion. A calculated answer with full reasoning trail.
You review the output and decide
You see a ranked comparison with source data, certification status and full reasoning trail. You decide. The system records your decision and the reason.
The audit-ready report is generated automatically
Ready for internal approval, external audit or public procurement review. No extra work afterwards.

Why local deployment

Three consequences of local deployment that are directly relevant for municipalities and construction companies:

Your procurement data never leaves your own server.
There is no dependency on external APIs or cloud providers.
The system complies with GDPR and municipal IT policy.

Sensitive procurement information, supplier pricing agreements and contract data stay where they belong: with you.

Every output shows the reasoning process

Which data sources were used, how each supplier was assessed, why one supplier ranks higher than another. The system records it explicitly.

For municipalities this is a legal requirement. For construction companies it is the difference between a report a manager signs and one they send back.

Why this doesn't work with generative AI

In an internal Skanska experiment, a generic enterprise AI system gave different numerical outputs for identical inputs, with errors up to €8,354 on a single tender line. Generative AI generates an answer. It cannot reproduce the calculation. Dundir's optimisation engine is deterministic: the same result for the same input, every time.

Hamppi 2025, Aalto University.

The system is trained on your own procurement history

Dundir is trained on your organisation's own procurement history, not a generic model. The result: the system knows your preferred suppliers, your certification requirements and your previous decisions. Every retraining after a new procurement cycle refines the objective function further based on your own data.

When a senior procurement officer leaves, that knowledge remains available to the whole team.

What the system doesn't do

Dundir does not make procurement decisions. The procurement officer reviews the output and decides. The system does not replace the procurement officer.

It removes the administrative work so the procurement officer can do the actual work: judge, maintain relationships and escalate. The decision and the responsibility remain with you.

What an implementation looks like

Intake and process mapping
We map your current procurement process: which systems, which suppliers, which decision criteria.
Data ingestion and fine-tuning
The system is trained on your ERP procurement history and supplier data. This is what makes the output specific to your organisation.
Deployment on your infrastructure
Locally, on your own servers. No cloud connection required.
Handover: you own the system outright
After handover, no subscription is required for the system to function. Maintenance and retraining are taken on when useful, not because the system would otherwise stop.

Curious what manual procurement costs your organisation?

Take the scan

Five questions. Specific to your team size and volume. No email required to see the result.