Should I get an Artificial Intelligence (AI) tool to do proposals?

*Guest Contributor

By David Grow, CEO, Defense Proposals Group, LLC

April 20, 2026

As the CEO of Defense Proposals Group, I often hear that question and have been asked to share my thoughts. My answers will likely surprise you. Yes, I believe you should get an Artificial Intelligence (AI) tool. I have reviewed and corrected many proposals written with various AI tools. I also know from years of Source Selection evaluations that bad proposals can win and good proposals can lose. Here are three good ways to employ AI in proposal writing:

1. Increase your bid rate. If this is your goal, then absolutely use AI. You will generate more proposals more affordably. I will discuss quality impacts and trends later, but there are situations when a contract simply must be awarded, and your AI generated proposal may be good enough to increase your probability of a win (P(Win)) across your bid and proposal enterprise.
2. Prevent the “blank page” fear that hits most writers. Humans dislike starting to write from scratch and facing a blank page. Having AI generate a draft will help lay down a draft that can be edited and give some good ideas on how to approach the response.
3. Replace your Pink Team initial authorship. Your Pink Team is likely a group of experts focused on a daily job and the proposal work is an additional task to perform beyond duty hours. Using AI to generate your Pink Team drafts can save you time and move you to Red Team reviews more rapidly.

While you see the potential value in using AI tools, there is also a bad side. I have reviewed documents written by several different AI tools and engaged in discussions with AI tool manufacturers. There are many risks and issues that I have seen across my client base. Here are some key things to consider as you contemplate using AI for proposals:

1. Document Believability/Confidence: You and your employees speak for your firm, but your AI tool does not. It simply processes prior data. Documents are thus less trusted and believed once identified as AI, and AI generated products are generally easy to identify.

Here are some of the common trends:

Punctuation: a human will tend to use a dash (-), commas (,), or parentheses to identify parenthetical type comments. AI tools tend to use em dash (—) punctuation extensively – and I have seen more than 250 of them employed in a ten page document. Rarely does a human use an em dash.

Style: AI tools tend to make declarative statements at the end of paragraphs to create emphasis, with a structure like “We don’t just tie our shoes — we make sure the shoestrings are glued.”

Compelling lists: AI tools frequently build lists of things that the client knows or does, and the lists read very well. Once you look at the lists, you realize that the AI tool did not know what was most important and did not know where to end the list – it just built a list. If a client seeks to detail the ability to facilitate overnight delivery of parts, the AI may build a list of truck reliability statistics, tolls paid annually, and traffic citation histories. The AI list reads well but may convey that the firm does not know the importance of the list items being proposed. The key points were likely on-time deliveries, turnaround times at loading docks, and schedule compliance – all missed by the AI in favor of other data.

Tables of Information: AI tools also are terrific at making informational tables, and I love the way they do it – often well structured. However, I also see that most of those tables have changed the language of the requirement or response approach so that it no longer matches the RFP content. I recently saw an AI tool take a list of past trouble ticket categories (5 qty), a list of work areas (7 qty), and turn them into a list of six “domains” without ever explaining what a domain was. It used the domain term throughout the proposal, and even the client had no idea what a domain was. I saw another client’s proposal’s AI list the five categories of a critical reference document by number. When I tracked the reference document, none of the categories bore a number, and AI had reworded all of the listed items, so the traceability and compliance was gone. The client believed that the requirements were addressed, but instead they were ignored by the AI’s modification of the table and use of non-existent numbering structures.

1. Culture: When AI tools come into your firm, you also face challenges of how your firm culturally adapts to the tools. I have seen firms believe that the AI generated text is the authoritative information and write around it, never checking it for accuracy. In this past year, massive changes have impacted proposals. Mandatory environmental and diversity content has been removed. New technologies abound. Many authoritative documents, regulations, instructions, and manuals (including the Federal Acquisition Regulation (FAR)) have been updated. If you train your AI tool on prior proposals or requirements, then it will give you a solid response based upon outdated information and practices. Remember, you and your staff get the emails and read the latest articles; but your AI tool does not. When the FAR changes, the AI tool is trained to the old document and will apply the wrong rules. When the regulation is updated four times, the AI tool may still be processing the version from four generations ago. The solution sounds easy – check the references. However, AI tools generally do not tell you their references. They simply process the information. You need to figure out the source and check it. Make sure the AI system is treated as a tool – not an author. Make sure the AI’s work is checked. Do not expect the AI to excel at current events unless you are carefully feeding it the latest data. It likely has no idea about new processes or what firm went out of business five years ago. Check its work.

1. Operations and Disclosures: As the prior examples show, AI tools change your operating environment. Be honest. Most clients give me AI based products and act as though they are expert written. I get well into the review before I feel confident enough to say that I identify it as an AI product. Absolutely let people inside and outside your organization know that you employed AI. The work needed to finalize an AI document is vastly different from the work needed to finalize an expert authored document. The expert’s words can be trusted and add detail and innovation to your proposal that should be leveraged. AI is reprocessing what is already established. If your editors do not know which one is the original author, then they may give your AI some undue credit and let errors enter as though they were innovative. AI can process data and write some quality text, but it does not innovate on your behalf.

In summary, AI can add value and can also do damage. How you apply it, identify it, and review its work makes all the difference to your end document and P(Win). Use it as a tool, identify that it was used, and ensure that your experts check everything that it says. Like any other tool, AI can move you ahead if properly used and can do damage if it is misused. AI is not a substitute for a human at present; any more than a hammer is a substitute for a screwdriver. Simply buying a tool does not add value or quality but using it effectively and properly can do both.

© 2026, David Grow, Defense Proposals Group, LLC