We spoke to grant consultant Hector C. Pagan about his thoughts on AI’s current and future impact on the industry. Mr. Pagan has raised over 75 million euros of international project funding from LIFE, Horizon, Interreg, Erasmus and other sources during the past 20 years.
How much concern exists within the industry about AI potentially replacing people in developing project proposals?
I would say that it’s a massive concern – like in most sectors of society. Part of that is maybe because of some exaggerations: you have many people advocating for AI who, of course, make it sound like it’s perfect and solves every single problem.
Right now, I think that the AI is not there yet. I know from different colleagues about a few occasions where researchers have said to them, “I have a proposal – I’m going to just use AI to do it.” In one of these cases, my colleague discovered a mistake that the researcher had made in the proposal, which ultimately screwed up the whole process and the proposal never got submitted. That same researcher is now coming back to my colleague and asking them, “Do you have time to help us with our next proposal?” Was the problem that the AI made the mistake or that the researcher didn’t use it properly? I have no idea.
To what extent can AI replace people in the development of proposals right now?
There are many different types of AI. For example, there’s the agentic AI, which automates tasks. But even with that, from what I have observed and understood, a lot of people aren’t able to manipulate it well. You need to know how to program to be able to use it. Even these so-called “off-the-shelf” products that claim to do amazing things don’t necessarily do all that.
One other way I have seen that AI is already benefiting the development of proposals is through ideation – it can help you to quickly shape your initial ideas about what you want to do in your project and how it should be structured. But these ideas still need to be examined closely by a human, to make sure that they match the call text and goals of the funder, the mission of the organization and that they can be implemented in the real world.
Tasks that are repetitive or follow standardized steps in proposal writing can be automated to some extent. Like emailing partners – you dictate the email (or the AI suggests text), and then it sends the email out to everyone. Or setting up meetings and inviting 15 partners. Or taking notes from proposal planning meetings. But this doesn’t completely replace a human being. There still has to be someone to trigger those actions at some point. And probably double-check that they are done correctly. I don’t think we’re at the stage of full automation yet. It’s more about having a human triggering those processes, which allows you to scale up and increase throughput.
For proposal writing, the most useful function of language models is the quick generation of text. For a non-native speaker, that’s probably a very valuable thing. But for a native speaker, it’s currently less of an advantage. You can have the AI quickly generate blurbs of text, but you then still have to go in, edit and adjust the text.
The idea that AI can currently write a whole proposal is unrealistic. Even if you say, “Write a work package on this,” it’s going to get things wrong. There are issues like AI hallucinations; the question of how much and what the source material is that the AI was trained on, and all of these things. The AI will get better and better, but right now it’s still a mixed bag.
Same goes with the programs for generating drawings and all of that – if you need a very basic, simple drawing, it’s great. Instead of me taking two hours to create this clean, simple drawing, I can use the AI and make it in five minutes. But if I were to show you my proposals, where we have drawings with interrelated pieces and all these different elements connected together – partners collaborating, one piece of research feeding into another, and feedback loops – the amount of effort you’d need to prompt it to draw something like that and make it look good… it seems to me you might as well just do the drawing yourself. Perhaps if you’re skilled at prompting AI, it might be possible to generate more complex illustrations, but for the general audience, it is currently likely to be a difficult and time-consuming task.
Learning how to properly make use of all of the different AI tools that are coming out also seems to be a significant barrier. Who knows which ones will still be with us in a few years?
I also worry about the effects that AI has on the quality of proposals. Sure, I can generate hundreds of pages of text, but unless I have absolute trust in AI – which I don’t think anyone can have at this point – you need to be able to effectively review the text and revise it. To ensure that it is accurate, can be implemented given the resources available, and is compelling to readers.
When do you think AI is going to make a more meaningful impact, if you can estimate it by any means?
In terms of proposal writing, AI has already made a meaningful impact as a tool that can help save time and scale-up output. For example, I think that an experienced person could increase their output by automating some tasks and quickly generating text and simple drawings. Now instead of doing one proposal, I could do two in the same amount of time. But I believe that there is a limit to this scaling up right now. Because it takes time to master the various AI tools that are out there and also, I would still need to spend time editing and reviewing the text. To fix mistakes made by the AI, to better clarify the description of your intended work and to make sure that the different pieces of the idea and proposal fit well together. Right now, I would not trust AI to tell me that an idea is “innovative enough” or that it fits the call requirements well enough.
But having a massively disruptive effect, like causing huge numbers of consultants to be out of business and everything? I mean, it’s hard to anticipate the speed of the technology, but I’d say we’re at least five, if not ten, years away from that.
This is because the proposal process involves far more than just writing. You have to have the right partners; you need to bring them in and facilitate communication between them. You got to form ideas and get partners to come up with the ideas, figure out how the different concepts connect to each other, manage conflict between the partners, identify where there are gaps in the consortium…
People who go out thinking, “Yeah, I’m going to develop a whole Horizon proposal and use AI to write it for me.” So far, what I’ve seen is that those people have failed, because most people overestimate how much work the AI will be able to do for them.
If professionals like yourself were to lose jobs to AI in 10 years, where would you go?
Where does anyone go when technology starts to make them obsolete? On one hand, I think that there will always be opportunities, right? For example, right now, you see that a popular term nowadays is “AI prompt engineer” – a person who’s really good at prompting the AI to be able to get the outputs they want. And that’s a skill that everyone needs to learn. That’s a job that didn’t exist a few years ago
I think that there will still be certain types of jobs that people will still be able to do – it’s hard for me to anticipate what they will be, but the innovative or entrepreneurial people will look and find the roles that they can still fill.
Of course, the effect on my job is just one of the concerns I have about AI in general. To a large degree it is being forced upon us with very few guardrails in place. The EU is one of the few political actors that seems to be taking these concerns seriously.
And in the long term, when we move to quantum computing where computers are a thousand times faster and more powerful than anything we have today, all aspects of our lives will be affected. How we work and live will need to be radically rethought. Maybe someone should write a proposal about that? 🙂