KriterioBot, built on Generative AI, identifies and suggests criteria, weighting, and prose for both EU tenders and smaller bids, based on previous tenders. This saves Energinet valuable time on each tender.
The company: Energinet
Energinet is responsible for Denmark’s electricity and gas transmission network and plays a central role in ensuring the most stable, sustainable, and cost-effective energy supply.
Central to that mission is a lot of time-consuming tenders and procurement of goods and services, all of which must comply with national regulations, documentation requirements, and EU legal requirements. To ensure open competition, transparency, and the best possible terms, a structured and efficient tendering process is crucial.
The wide range of Energinet’s procurement creates a significant need for uniform, consistent, and precise sets of criteria to evaluate suppliers and bids.
Energinet has therefore established the project ‘KriterioBot’ as one element of a larger strategic initiative, where the potential of Generative AI is explored across Energinet.
The engine in KriterioBot is based on previous tender materials, which Energinet’s buyers can drag-and-drop into a simple web interface built by us. This ensures that KriterioBot is always based on the best possible material.
Using the refined tender history, carefully designed prompts, and OpenAI’s powerful large language models, KriterioBot identifies and suggests criteria and weighting for both large EU tenders and smaller purchases. This increases efficiency as well as quality and helps ensure that the criteria comply with the correct standards, rules, and formats.
Theme/object
The buyer enters basic information about the desired purchase
Prompts/instructions
The buyer can ask KriterioBot to focus on, for example, sustainability, ISO standards, or to comply with specific format requirements or text lengths.
Product group filter
By limiting the source material to specific product groups, one can guide KriterioBot’s focus in a particular thematic direction.
Language
Danish or English – KriterioBot supports both languages.
Creative or source-true
The buyer can specify whether KriterioBot should be source-true or imaginative. If source-true is chosen, only results based on the source material are presented. If imaginative is chosen, KriterioBot generates new, potential criteria on its own.
Draft
KriterioBot produces a draft of the criteria set, which the buyer can use for inspiration. There will always be a competent buyer who prepares the tender material. KriterioBot is a capable assistant in the startup phase.
Sources
KriterioBot references all the documents it bases its answers on, so the buyer can always see the context and find more information.
An efficiency that reduces work pressure and frees up time for more value-creating tasks, while standardization ensures that quality remains high.
The KriterioBot project is an element of the first GenAI project at Energinet. The development has taken less than 6 months from startup to launch, including testing and fine-tuning of initial issues.
Now that the fundamental systems, understandings, and technical designs are in place, the path is laid to address the GenAI potential in more parts of Energinet, which will lead to further efficiencies and innovation in the future.
With KriterioBot, Energinet has taken an important step towards a more digital and automated future.
Anja Overgaard
Commercial Lead Manager, Strategy & Sales
Inspari – a valantic company