Using AI to identify sustainable procurement criteria in public tenders

Manually checking thousands of tender documents for sustainability criteria is virtually impossible. Artificial intelligence can take on this task – provided it knows what to look for. A BFH research project is investigating how standardised procurement criteria can improve AI-supported sustainability detection.

Anyone who checks a tender for sustainability relies on specialist knowledge: Which criteria are relevant? Where are the sustainability aspects that are not immediately apparent? This knowledge is sector-specific – what applies in construction may not apply to IT procurement.

This is precisely the challenge facing artificial intelligence: without clear guidelines, it lacks the context to reliably identify sustainability criteria. A research project at BFH is therefore investigating how standardised procurement criteria can improve machine recognition.

Existing recommendations as a starting point

For many service and product groups, there are already national and international recommendations on technical specifications, suitability and award criteria that show how sustainability aspects can be integrated into tenders. Examples of such recommendations are the Swiss Sustainable Procurement Toolbox and the EU GPP Criteria. Taking these into account saves time. If the selected criteria are semantically consistent with these recommendations, AI can more easily identify them as sustainable. If the criteria deviate more from the recommendations, recognition becomes more difficult. Following the official recommendations therefore supports both the integration of sustainability into tenders and reliable recognition by AI.

Closing gaps: new criteria for underserved sectors

For some service and product groups, there are currently insufficient recommendations on sustainable procurement criteria. In such cases, AI falls back on general, cross-sector recommendations. While this is possible, it leads to less accurate results, as sector-specific characteristics are not taken into account. This makes it all the more important to develop new recommendations for these groups. This can be done through official guidelines, but also through the exchange of practical experience or the joint development of industry-specific recommendations. This makes it easier for procurement agencies to integrate sustainability more clearly into their tenders and helps suppliers to make their services visibly more sustainable. This also benefits the AI, which can then identify and map sustainability ambitions more reliably. Investing in new recommendations and revising existing ones is worthwhile in order to embed sustainability in public procurement in the long term and make it more visible.

Implications for practice

Three key findings can be derived from the research project:

1. Use existing recommendations: Where sector-specific criteria already exist, using them saves time and increases the visibility of sustainability ambitions in tenders.
2. Develop new recommendations: Many service and product groups lack appropriate recommendations – a gap that should be closed by industry initiatives or official guidelines. Existing recommendations should also be regularly updated, as markets and requirements are constantly changing.
3. Change perspective: Clear wording improves machine recognisability. Sustainability criteria that are clearly and standardisedly formulated can be identified more reliably by AI – contributing to greater transparency in public procurement.