TLDR: AI tools, particularly Large Language Models (LLMs) like ChatGPT, are transforming the way we interact with technology to create content. Recent advancements, such as OpenAI's ChatGPT-4o, have made these tools more user-friendly, making technologies more accessible for users. For proposal and bid development, AI tools assist in research and content generation, though with necessary precautions: safeguarding privacy, validating outputs, and clearly marking AI-generated content. Prompt engineering, particularly using structured models like our CLEAR methodology, enhances the quality of AI responses making them more relevant to users. Despite AI's potential though, concerns remain about dependency, security, and maintaining unique, high-quality proposals.
Most of us will be aware of the rise of Artificial Intelligence (AI) tools in the last few years; specifically Large Language Models (LLMs)* such as ChatGPT or Microsoft’s CoPilot (other platforms are available). The technology is largely free to use and an effective alternative for traditional search engines where we’re looking to get succinct responses to the queries we make.
The use of AI in our jobs, we believe, is a fantastic benefit and rather than a threat to our future employment, can aid us in being more effective day-to-day. It is also becoming more useable and accessible, especially for those with very little experience of the technology. This was reinforced earlier this month (May 2024) when OpenAI released ChatGPT-4o, which seeks to make interactions more intuitive and natural. It also brings together the use of audio and video along with textual interaction. You can see more here:
Tools like ChatGPT are impressive and one can foresee the acceleration of capabilities across all similar platforms in the months and years to come in an endeavour to bring the technologies to users whatever their capabilities to use technology. It is probable too that AI tools will interface with other technologies to make things easier to use, in an attempt to enhance the customer experience, or logically, to reduce costs. Within the facilities management and business services sector where we largely operate, LLMs can help inform diagnostics for engineering and helpdesk functions, deliver contextual health and safety messaging, or keep service users informed about progress.
For those involved in proposal and bid development, AI tools are incredibly useful for researching topics or to aid in developing our understanding of themes or topics; but they must be used with some caveats. We’ve written about this previously here: https://www.keynegroup.com/AITools.php. By way of a reminder, our key takeaways from our research and use of multiple AI technologies include:
- Restrict use of AI tools where is a risk to privacy and confidentiality.
- Validate and edit the outputs from AI tools where used.
- Highlight where a response or communication has been informed by or generated by AI tools.
Our AI Policy is available on our website for those curious about how we protect our Clients from AI generated content in bids and proposals. Despite the advances on useability, and our forewarning on using AI platforms, the benefits do outweigh the challenges.
Using prompt engineering to get the best out of AI
We find that structuring our interactions with AI helps in generating more meaningful and relevant responses. This is often referred to as prompt engineering and can be a science in itself. The main point to recognise is that there is no single method or set of rules for how you communicate with AI. Finding what works best for you out of trial and error is how the vast majority of users interact and is how most of us at Keyne have developed our understanding of what is most effective for us. We’ve narrowed our prompts down to a single universal specific model in the most part, the CLEAR model. This is an acronym and stands for the following:
The CLEAR Model:
- Context and Purpose - Describe the context and purpose of your interaction, being clear about what you need.
- List Key Points and Details - List the key points or details you want included.
- Explain Specific Instructions - Give specific instructions to guide the AI, specifying the structure, end goal, or outputs.
- Articulate Tone and Style - Specify the tone and style of the response.
- Resources and Constraints - Mention any constraints or preferences.
Given the intuitive nature of the approach, we prefer to go through an example rather than describe each point of the model in detail. Using a fictitious tender question, let’s have a look at how we might generate a response using ChatGPT.
Tender Question: Provide a detailed overview of a contract Communication Plan to ensure effective interaction between all affected stakeholders. Word count: 500 words maximum.
Using the CLEAR approach:
| Heading | Our Prompt |
|---|---|
| Context and Purpose | Generate a formal proposal response covering how we would generate a contract communications plan to ensure effective interaction between all affected stakeholders. |
| List Key Points and Details | Refer to industry best practice when defining approach • Identify established communication frameworks to adopt to improve effectiveness • Relate to key elements of ISO 44001 – the standard for collaborative business relationships, citing parts of the standard that relate to communications • Propose alternative methods to reach a dispersed stakeholder audience • Propose different frequencies of interactions to suit importance of topics and criticality of service outputs |
| Explain Specific Instructions | The response should be in clear flowing paragraphs, avoiding use of bullet points where possible and structured to directly respond to the question without any recap. |
| Articulate Tone and Style | The tone shall be business formal, written in the third-person and have a score of between 60 – 70 on the Flesch Reading Ease scale. |
| Resources and Constraints | The length of the text shall be no more than 500 words and be written with UK English spelling. Avoid an introduction and conclusion. |
Our consolidated prompt for ChatGPT: Generate a formal proposal response covering how we would generate a contract communications plan to ensure effective interaction between all affected stakeholders. Refer to industry best practice when defining approach. Identify established communication frameworks to adopt to improve effectiveness. Relate to key elements of ISO 44001 – the standard for collaborative business relationships, citing parts of the standard that relate to communications. Propose alternative methods to reach a dispersed stakeholder audience. Propose different frequencies of interactions to suit importance of topics and criticality of service outputs. The response should be in clear flowing paragraphs, avoiding use of bullet points where possible and structured to directly respond to the question without any recap. The tone shall be business formal, written in the third-person and have a score of between 60 – 70 on the Flesch Reading Ease scale. The length of the text shall be no more than 500 words and be written with UK English spelling. Avoid an introduction and conclusion.
The response from ChatGPT can be viewed here: ChatGPT Output
Our Thoughts
On first reading through, the AI generated output sounds relevant and authoritative and the CLEAR approach has delivered us comprehensive and relevant narrative. Style aside, it appears to be a good foundation upon which to base a response.
There are some concerns though. Reference to clauses within ISO 44001 aren’t the most relevant to cite, and tools proposed might not be the best determinant for the frequency and urgency of communications on a typical contract. There’s also no mention to front-line operatives or specific roles (because we didn’t include them in the prompt) and generally, the text is vague and can be true of any contract, (again because we didn’t provide detail in the prompt).
Overall though, AI generated outputs do appear to provide a foundation on which to base the final response. With some editing and validation, for some questions, using the CLEAR approach with AI tools just might be beneficial.
But just because we can, does this mean we should?
What will differentiate Suppliers if we all use the same tools, how can Evaluators validate what is being submitted, and is there a concern that we will deskill the art of bid and proposal writing over time by becoming more reliant on this type of approach?
Additionally, building on the success of tools like ChatGPT and recognising industry concerns, specialist AI powered bid and proposal tools have become available that allow organisations to relate to their own library of content on which to base responses. This makes generated text less general in nature and helps organisations to bid more with less resources. These tools can be expensive for smaller Suppliers and still raise questions around security, and on the relevance, validity, and authenticity of outputs. They do seem to provide a good approach for transactional and repetitive works, but, in our opinion, less so where more bespoke responses are needed. As they become more affordable, and organisations generate their own libraries of content to reference, they may become more commonplace.
What do you think?
Do you have experience of using AI for your bids and proposals? Have you had challenges with evaluating AI generated responses within your procurements? Or are you worried for your career in the months and years to come? We’re genuinely curious as to what you think the future of AI for bids and proposals looks like.
Need help with adopting AI in the workplace? Need effective policies and routines to embed for your workforce? Drop us a note below.
NOTE: When we mention AI in this article, we specifically refer to Large Language Models (LLMs). Other types and subsets of AI exist and if you are interested in finding out more, we point you in the direction of the excellent IBM article at: https://www.ibm.com/think/topics/artificial-intelligence-types
A.I. DISCLAIMER: This content has, in whole or in part, been informed from the outputs of A.I. tools. Where a position is presented that is wholly based on the output of A.I. tools, this will be clearly identified. The main banner image of this article has been created with A.I.