inQubator: TenderBot Experiment

Tenders

Every agency (and I would guess most companies) have to respond to tenders. It's how we win work and make money.

Whilst each tender is slightly different, there are always similarities between them. Many of the questions will be identical (especially for public sector tenders that have a specific structure) and they ask for the same evidence.

We were spending days filling these in. There had to be a better way.

The solution

Luckily we already had a repository of reusable content that had been built up over a few years. This was in the form of a SharePoint list and a set of case studies (but theoretically could have been any format). 

Using this data we created a bot that would ingest a tender RFP (and supporting documentation) and then spit out suggested answers. Of course, they need some tweaking and curating, but it has turned days into mere hours.

Who better to ask than the bot itself whether this was successful:

"I am useful because I:

• Transform Quba’s delivery history into structured, persuasive tender responses
• Provide defensible evidence using real project outcomes
• Align technical detail with procurement evaluation criteria
• Reduce bid preparation time while increasing quality and consistency
• Ensure every claim is grounded in documented experience

In competitive tender environments, clarity, credibility and evidence win contracts.

My role is to ensure Quba presents all three."

This was a good case of where using AI proves the "good data in, good data out/bad data in, bad data out" mantra - if we didn't already have the data this experiment would have been a bust.