Sustainable AI in Tender Writing: What Bid Teams Should Ask About Energy Use

Sustainable AI in Tender Writing: What Bid Teams Should Ask About Energy Use

Many organisations that bid for public sector work have net zero commitments, carbon reduction plans and ESG reporting obligations. As AI becomes part of the bid process, it is reasonable to ask what environmental footprint those tools have and how to account for it.

• Running and training AI models uses electricity and, in data centres, water for cooling
• The footprint depends on the model, the hardware, the data centre and how the tool is used
• Retrieval based approaches focus the model on a smaller set of relevant documents
• Published per query figures vary widely between sources and change as models and hardware change, so ask vendors for their own data
• Include AI tools in the same supplier questions you already ask about carbon and sustainability

Why AI has an environmental footprint

AI models run on servers in data centres. Training a large model and answering each request both use electricity, and data centres may also use water for cooling. Independent estimates of the energy used per request differ considerably, depending on the model, the hardware, the data centre's energy mix and the length of the task, and they change quickly as the technology develops. For that reason this article does not quote per query figures; ask each vendor for current data relevant to your use.

How retrieval based tools work

Retrieval augmented generation (RAG) is an approach where the system first retrieves relevant passages from a defined set of documents, such as your past bids, policies and case studies, and then uses a language model to draft an answer from them. mytender.io uses this approach so that drafts are grounded in your own content and evidence.

Retrieval also shapes how much work the model is asked to do: it is given a focused set of relevant material rather than an entire document library. As with any AI system, the overall footprint still depends on the models, infrastructure and usage involved.

Practical steps for bid teams

1. Use AI where it adds value. Focus it on tasks such as analysing tender documents and drafting from your evidence, rather than repeated trial and error.
2. Reuse approved content. A well maintained bid library reduces the number of drafts needed for each answer.
3. Ask vendors for data. Ask where their services are hosted, what is known about the energy sources of those data centres, and whether they report on emissions.
4. Include AI in your reporting. If you report scope 3 emissions or prepare a carbon reduction plan, include material digital services in the same way as other suppliers.

Questions to ask any AI vendor

• Where are your services hosted, and which cloud providers do you use?
• What do your hosting providers publish about renewable energy and water use?
• How do you limit unnecessary processing?
• Can you provide information to support our own carbon reporting?

For more on choosing AI tools for bids, see our guide to general purpose AI versus specialist bid tools and our guide to the best AI bid writing software.

Sources and Further Reading

• Environmental Impact of Artificial Intelligence - Wikipedia
• Retrieval-Augmented Generation - Wikipedia
• Data Center Environmental Impact - Wikipedia

---