Can I draft a proposal using ChatGPT and a GPT?
Ask any proposal expert about their perspective on AI, and you'll encounter a spectrum of opinions. Some embrace AI tools like ChatGPT with enthusiasm, while others remain cautious, if not skeptical. In the middle, you'll find those intrigued by the potential but unsure where to start.
Let's first clarify that what you can do using ChatGPT depends on the model you are using: ChatGPT 3.5 (free version), ChatGPT 4 (subscription-based chat), or a specialized GPT.
How far can I go by using the free version? (If you tried this already fast forward to GPT-4 and GPTs sections)
Here are a few practical ways to use a free version of ChatGPT (ChatGPT 3.5) in proposal management:
- Igniting Proposal Positioning:
While ChatGPT doesn't reveal proprietary competitor information, it can swiftly provide insights into publicly available data about your competitors. This data can then inform your response strategy and position your proposal more effectively.
- Efficient Summarization of Information:
Responding to RFPs often involves condensing vast amounts of data, a painstaking task. Generative AI, when prompted correctly, can help with summarizing information for various proposal sections. ChatGPT can further identify key themes and requirements, helping companies to tailor their responses more effectively.
- Elevating RFP Responses
Proposal teams often tackle a multitude of questions in RFPs, which can be overwhelming. ChatGPT can handle the initial drafting, and "chain prompts" can fine-tune responses to cater to specific needs, making the content more relevant and consistent.
- Identifying missing information
By analyzing the RFP, ChatGPT can identify any missing information and can prompt the user to provide it. Improving the completeness of the response.
- Editing and Proofreading:
Finally, you can use ChatGPT to review and edit your response for grammar, spelling, and clarity. It can help you improve the overall quality of your document.
Example prompt for answering a question from an RFP:
„You are a [role responsible for answering this type of question] for a [describe company], targeting [describe the prospect company]. You offer [describe product or service] and you need to explain [describe what the prospect is wanting to know] in an RFP response. Can you answer this question?“
RFP Question:
[Paste the RFP question here]
First follow-up question: Can you enhance the response with the following data?
[Paste the updated information sourced from a Subject Matter Expert]
Second follow-up question: Can you address the specific pain points and requirements highlighted below?
[Paste the relevant details about the prospect]
Third follow-up question: The recipients of your response lack technical expertise. Can you simplify the explanation without compromising professionalism?
New possibilities with GPT-4 in Proposal creation
A significant evolution from ChatGPT 3.5 is GPT-4, introducing an innovative feature called Advanced Data Analysis. Exclusive to premium (paid) accounts, this feature allows you to directly upload data to ChatGPT. The primary advantage is the ability to submit files directly to the system, eliminating the need for you to explain complex data and make constant adjustments during regular conversations. Advanced Data Analysis supports various file formats, like text, image files, full documents (PDFs, code, and other data files), as well as audio and video. The performance of this feature varies based on the file type, but it's specifically designed for data files such as .csv and .txt.
GPTs for Proposals
On November 6th, OpenAI took a significant step forward by announcing custom versions of ChatGPT, referred to as GPTs. This allows users to create personalized instances of ChatGPT for specific purposes, expanding the range of applications beyond the default capabilities.
In terms of skills required, anyone can build their own GPT - no coding is required. The process goes like this: You start a regular conversation by assigning the role to GPT and give instructions on its behavior and the topic. Does this sound familiar? Yes, just like the steps we created above for ChatGPT 3.5 to draft an answer for an RFP. The next step is to save that instance. Essentially, you turn the conversation into an app that others can run. This means you could create personalized instances of ChatGPT (GPTs) tailored specifically for certain types of RFXs.
Joining the global wave of excitement, we gave it a try and built the Proposal Generator GPT designed for generating RFx responses using historical files!
The excitement was high, and we quickly noticed improvements compared to using GPT-4 and its file upload option. The underlying model remains the same, as does the intelligence. Ultimately, the end result is still the answers to questions that you would need to copy and paste into your RFx response, but with specialized GPT you save significant time on prompting and you feel like you are having a discussion with the subject matter experts itself. You might need to fine tune the answers, but not necessarily. If you are responding to many RFx documents and you want to save time on prompting each time, then this kind of specialized GPT is a great fit.
Before deciding how to use ChatGPT or a GPT, we recommend getting yourself familiar with limitations and data security aspects in OpenAI’s File uploads FAQ. Up to 10 files per GPT for the lifetime of that GPT, as well as deletion of uploaded files to Advanced Data Analysis within 3 hours, is something to keep in mind. There are file size restrictions and usage caps per user/org as well, that you can read more about in the FAQ article.
In the same article you can find the answer to the commonly posed question: Will OpenAI use files uploaded to train its models? OpenAI states that they do not use content submitted by customers to their business offerings such as their API and ChatGPT Enterprise to improve model performance. Meaning, if you are concerned about your files being used for training AI models, you should either edit each file before uploading so the sensitive data is hidden, have an OpenAI Enterprise account or look for a solution that uses OpenAI’s APIs.
ChatGPT and GPT for Proposals: A Recap
In summary, the answer to "Can I draft my proposal with ChatGPT?" is a definite yes, but let's not ignore its quirks.
First and foremost, there's the issue of domain expertise. ChatGPT, while a brilliant tool, might fall short when it comes to generating content that requires the deep, technical knowledge of certain specialized fields. You might find it lacking the precision and depth needed in such cases. When uploading your files as a source for answers this risk is mitigated to a certain degree.
Next on the list is the challenge of limited context understanding. ChatGPT, as intelligent as it is, may not always grasp the unique context, nuances, and intricacies of your RFP. Specialized GPT, if fine tuned well, should remove this oddity.
Then there are the inaccuracies. While ChatGPT does its best to provide accurate information, it's not infallible. There's a chance it could produce incorrect or outdated information. So, always have your fact-checking hat on and verify any information it spits out.
If you are not using your files as the source, the ever-looming threat of plagiarism is there as well. ChatGPT crafts its responses based on publicly available information, which could inadvertently lead to similarities with existing documents or sources.
If you're considering building or using a specialized GPT model from the soon-to-be-available GPT store, be aware of the limitations and data security considerations we touched upon above.
In conclusion, utilizing ChatGPT, whether through the chat or a specialized GPT model, undoubtedly enhances the proposal-building process, offering significant assistance. While it may not replace the art of crafting proposals, it empowers us to work more efficiently. With the latest progress made by OpenAI we can surely state - It is an exciting time ahead of us!
At EstiWiz, we are looking forward to extending our offering on the OpenAI platform and we believe use cases we can incorporate with a no-code approach will increase and eventually some type of integration platform and code similar to Apex will emerge. We shall see.