Generative AI is getting better every day. We now have ChatGPT4, Claude 3, Gemini, Perplexity, and whole gamut of specialized GenAI tools that await your discovery. And people are using these tools in creative ways.

For example, UNICEF now explicitly states that they will not accept job applications that use Generative AI. I wonder how they would know when 70% of job applicants use GenAI today. Maybe if  the application was an obvious copy/paste or careless applicants included “as of my last knowledge update” in their documentation. In that case, the application should be trashed because it’s a bad application, not because the applicant used GenAI.

Which brings me to today’s question: “Should it matter if a proposal is written using Generative AI?”

How Is Generative AI Different?

We are already using different types of AI to write everyday correspondence. I know you used AI in your proposal process, even if you didn’t think of it as artificial intelligence. For example:

• Autocomplete is a simple AI automation program.
• Spellcheck and Grammar Check are pattern recognition AI.
• Google Translate is AI interpretation.

Generative AI is using probabilistic algorithms – a fancy way of saying it guesses at the next most common associated word or phrase, versus “knowing” anything – to create plausible text that may or may not be based on a known reality. It still takes human cognition to understand if the text is factually correct and edit it to make sense in the context of the full document.

3 Ways GenAI Can Improve Proposal Writing

I liken Generative AI in proposal writing as using a good assistant. It can help, but it’s not going to generate a perfect proposal for you on the first try. At least, not yet. I see three ways that we can use GenAI in the proposal process today. I’m sure you can think of more – add them in the comments!

1. Understanding Requests for Proposals

I already use the Ask Your PDF Research Assistant custom ChatGPT tool to summarize long documents and highlight key themes. This is especially useful when reports don’t have a strong executive summary.

Most donor RFPs have many pages of rules and regulations that crowd around a few select pages of technical explanations of what the donor really wants. I remember a USAID RFP that was 175 pages long, with only 12 pages detailing what was to be done. GenAI tools can cut through the clutter and pull out specific action items.

2. Developing Strong Win Themes

Once you know what you want to do, GenAI can build on proven responses – both from within an organization’s past experiences and from the greater corpus of development knowledge.

For example, DevelopMetrics has created DELLM, a GenAi tools that makes the Development Clearinghouse Exchange useful. It uses DEC entries, tagged and coded by development experts, to find which interventions were the most effective in different sectors and countries. That’s beyond the ability of any single human, but GenAI can do it easy-peasy.

3. Writing Compliant Submissions

I do feel sorry for local organizations trying to compete with traditional implementing partners. The 15 pages of certifications and representations required in many RFPs can be overwhelming even for me – and I’m a US citizen, Native American English speaker, and have 20+ years experience reading bureaucratic forms.

There is no way a 20-person NGO with 10 years of deep local connections that USAID wants, can ever hope to submit a compliant proposal when English is the 2nd or 3rd language learned – unless they use GenAI. That’s one reason I created the ADAIR and ADS Bots, tested GenAI on the FAR, and I am building a FAM/FAH Bot. We need more and better tools for local firms if we’re ever going to increase localization.

If you think GenAI isn’t ready to write proposals yet, then check out Grant Assistant – an LLM designed specifically to co-create proposals with human editors that include strong objectives, complaint activities, and buzzword usage that will be indistinguishable from (or better than) proposals solely written by humans.

What Are Generative AI Downsides?

Fake Quality: A good proposal put through Generative AI tools can become a great proposal, even a winning proposal. Hence, there is a risk that a subpar proposer could use GenAI tools to win a contract and then be unable to implement a successful project in compliance with the donors regulations.

How would that be any different from a subpar organization hiring a great proposal writer to create a winning proposal? Or donors not doing enough due diligence before awarding funding?

Overwhelming Volume: GenAI tools also make it easier to create proposals, allowing more organizations to submit more proposals using the same resources. Hence there is a risk that donors could be overwhelmed by too many quality proposals to judge properly under existing grading schema.

Might that really say that donors should have better RFPs? Or have different submission and grading criteria? Or just use GenAI themsevles to make proposal assessment easier?

Of course this all portends a day when machines will write proposals for other machines to review and reward – kinda like product review sites and Google Search today.

This is a 100% handcrafted post. Though how would you know if I used GenAI in its production – or not?

 

ICTworks