Should you use AI in tender writing?
Ryan May, Tender Specialist (Perth)
Everyone is talking about it, and everyone has an opinion on it. Importantly, a lot of people are also using it. But when is it appropriate to use AI? And is it even helpful for tender writing? This blog will walk you through the benefits and the shortcomings of this new technology to help you understand how it may assist you – or why you should be careful when using it. For reference, know that I will be talking about Copilot here, though many of the points discussed are relevant for other AI programs. Let’s jump into it!
AI for understanding
In theory, AI is an excellent telling tool – I’ll circle back to the ‘in theory’ part. Ask it to tell you about something and it will give you comprehensive feedback, often providing additional details you didn’t realise you wanted. Because of this, AI can function as a useful aid when reading through tender documents, and in my opinion, this is where it shines brightest. An AI like Copilot can be asked a simple question, such as, “When is this application due?” and it will return information on not just the due date, but also the day and time, and can even provide information on what time in your time zone the application is due. This is dependent on what information is available in the document you’re looking at, though Copilot is capable of looking through additional items such as emails, the internet or other documents in your SharePoint to gather further information for you, if you ask it to do so.
AI can also be asked to provide a summary of sections or the entirety of complex documents, helping you to quickly grasp the meaning of questions asked in tender documents. It is rapid and presents this information in an understandable way. This can help you to create more targeted and specific answers that increase your chances of winning.
Back to that ‘in theory’ caveat though – AI is great for all of the above, when it works. Unfortunately, I found it hit and miss. Half the time I’d be provided with fantastic, robust information, and would have no issues asking for Copilot to draw from multiple sources to perform its miracles. The rest of the time it would give me half of what I needed, missing important details, and refuse to draw information from accessible documents on SharePoint. Is AI great for understanding? Sure! Is it reliable? Not so much.
AI for doing
AI can also come in handy when it comes to doing things… in theory. I know! Once again, issues with reliability and consistency rear their head. Depending on what you plan to do with it however, it may still be capable of speeding up your processes and increasing overall efficiency.
The most well-known feature of AI is its ability to create written content, which can be useful in tender writing – to a point. While not my preferred use of AI, a first draft concocted by an AI can be a useful starting point for those with a phobia of the blank page. This is where a human is best positioned to take over. Continuous iterating of this draft with an AI will not help you to target the client’s needs and your win themes, as AI doesn’t know how to do this at a professional level. The human element is a necessity here to ensure the response covers all the required themes.
This is the same when it comes to asking AI to write emails or create tables in Excel. It speeds up both processes, and in the case of emails can create solid content that may only require minor revisions. In the case of Excel, AI is helpful in guiding formatting by providing you with clear instructions on how to complete complex formatting tasks, even going as far as to provide functional lines of code you can input and use. Further, Copilot can be used in other programs like Word to create tables populated by content from other documents, such as a tender schedules document. The contents of this table can then be copied and pasted into an Excel document, such as a tender management plan, speeding up the creation of management tools that require headings, descriptions of questions and word limits.
Of course, this is only when it works. My experience taught me that AI will just as often create incorrect tables as it will correct ones, causing frustration as I attempted to refine its creation into something useful. I often felt that in the time it took to achieve a usable AI table I could have made my own! And that was when it chose to function. If I attempted to ask Copilot to make changes to an existing table in Excel it often shied away, stating its inability to alter anything with merged cells – and making it functionally useless for my purposes.
The shortcomings
Beyond the problems of consistent performance, there are other significant issues that anyone getting involved with AI must know about. Confidentiality, for starters. While the paid version of Copilot will keep your data secure, other versions of AI as well as the free version of Copilot are not confidential and as such should not have confidential information entered into them. Doing so is effectively the same as putting that information on the internet for everyone to see. AI also suffers from bias in that it will pick up on biases in the material it learns from. It will also hallucinate, which is a fancy way to say it gets things wrong sometimes. For these reasons it is important you check the output of your AI assistant to ensure you aren’t pushing false or biased content into your tender submissions.
AI also requires a lot of hand holding. Specificity is key for effective responses from your AI, and I found that when I made more specific requests I had greater success. This can be more difficult than you might expect though. It helps to speak to the AI as though it is a child that needs very clear directions to provide you with the results you want. It will get things wrong, sometimes repeatedly, and until you learn effective prompting techniques it will likely feel quite ungainly. This is particularly evident when trying to get it to perform more complicated actions. My advice to those committed to adopting AI right now? Have patience and use it where it makes sense.
AI cannot do everything for you – don’t expect it to. When it works, however, it can be a useful aid that increases your efficiency and provides much needed tender support. In my opinion, it isn’t where it needs to be yet for it to be worth getting tangled up in. But despite this, I am drawn to it. Keep your finger on the pulse – it may not be ready yet, but with enough time I believe AI will become the useful, reliable tool it already promises to be.
Want some help with a tender beyond what an AI can provide? Tender Plus offers professional tender services such as tender management, tender coordination and tender strategy, to name a few. Don’t hesitate to get in touch – we look forward to working with you!
Did you enjoy this article? You might like some of our other blogs, like Brace for impact: 5 dos and don’ts for impactful tender writing and 10 reasons to use a tender writing service.