Tender Response Generation

Downer Group is an integrated services company active in Australia and New Zealand. Listed on the Australian Securities Exchange and New Zealand Stock Exchange as Downer EDI, Downer is an ASX 200 company.

Tendering for contract works is a time consuming and repetitive process. With finite resources, companies must selectively choose which projects to allocate resources and time in crafting a response and potentially miss out projects that they simply do not have the resources to bid for.

Automated responses are drafted using artificial intelligence, leveraging advanced capabilities in large language models like ChatGPT. The knowledge for these responses are drawn from past bids and reformulated to target the new project. The bid team then review the outputs and make adjustments before submitting a response.

Proven results in weeks, not years

The tool reduces the time required to prepare a response to each tender, allowing the bid team to tender for more projects, resulting in increased revenue through volume

With the advances in generative AI solutions such as Open AI’s ChatGPT and GPT-4, the main challenge remaining is providing context to the model from existing unstructured documents. Using Open AI’s embeddings, the content from the unstructured document library was broken down into short statements and their embeddings were calculated and stored within a searchable vector database.

When a new tender is presented, parameters about the current tender (such as the industry, number of words, etc.) are provided as input into the model together with relevant statements from the document library through cosine similarity searches. Prompt engineering is then used to generate an appropriate draft response for human reviewers to validate.