Everyone seems to be worried that the robots are going to take our jobs tomorrow. But the more I learn and think about the application of Artificial Intelligence tools to my work, the less afraid I get. At least for now. Let me explain why. 

Before I start, I should come clean about one aspect of my professional life. I have spent my entire career in the bespoke world of proposal writing. Think Rolls Royce cars, Savile Row tailors, artisan cheese. It is a very select world and it comes with the corresponding price tag. As the saying goes, if you have to ask how much it is, it isn’t for you. I know that’s an unusual perspective and I wanted to be upfront about it. 

My world is far from developing a proposal to deliver standard services, products, or a combination thereof. If you are thinking installation of electrical systems or plumbing in new construction, that’s not my area. That type of proposal development has been the focus of automation long before most of us heard of AI or large language models. There are libraries of component descriptions, automated labor estimates and so forth. All you need is to couple them with your prices, labor rates and your CAD system. If you are in the business of doing these by hand, your job has been under threat for a long time. And while companies tried to break into the very lucrative local and federal government market, they usually failed. There was simply not enough repeat business that warranted standardization or the upfront investments.

Today, I know that LLMs, with the right prompting, are very capable of generating text that reads smoothly and incorporates all the points you asked for. Given the right collection of past proposals, I am also confident that they are able to generate fully compliant proposal sections. There is just one problem. Well-structured, fully-compliant proposals do not win. I might even argue that such a “near perfect” first draft will slow down most proposal teams rather than help them. 

Why might that be? Isn’t it so much better to have a decent draft than an empty page? It is not, once you consider team psychology. As a proposal manager, I have often  been dazzled by a contributing author who, on deadline, provided a well-written proposal section that looked ready to be published. I thanked and congratulated the author and I didn’t seem them again for weeks, because they thought they were done. Only much later did I discover that what was delivered was material developed for another project and (most of the times, though not always) only the name of the project updated. Most of these sections, I ended up throwing away because they are either light on content and I couldn’t afford the space, or because they weren’t responsive to the current proposal call. All I had in the end was the illusion of progress that allowed the team to lose its sense of urgency. So do I want to use a tool that makes it even easier to create that illusion? Thank you, but no thank you.

There is an area in my work, though, where I think AI and LLMs hold tremendous promise, but it isn’t on the generative side. Where I believe they can really shine is on the compliance checking and horizontal integration. Horizontal integration, to put it in plain English, means that the price and weight of an article is the same in the introduction as in the middle, as on the last page. That is much harder to do than it sounds. Picture a 1,200 page document written by about 30 inexperienced writers. Fixing this in the final editing is a tremendous amount of work and very hard to get right. An algorithm that never gets tired and doesn’t miss anything seems to have a much better chance at getting this right than any human. Or why not have a writing assistant that makes all the data from other authors available as you write? Checking fonts, reformatting sections, and, and, and… There will be many more tools than I can imagine right now. Our tools will continue to change just as we don’t deliver reams of printed documents anymore. 

What I don’t see coming our way yet, is an automated system that can transcend the stacks of past proposal libraries and come up with a creative reinvention of a concept. Almost any successful proposal breaks the rules and constraints just in the right amount to be stunning yet still compliant enough. These concepts connect well-established methods and technologies in ways not previously considered. In my recent professional life, the Dragonfly project appears to have accomplished that brilliantly. And, by the way, left us competitors in the dust. 

As I haven’t seen AI produce something truly creative besides enthralling soap bubbles, I am not afraid of AI. At least not yet. But I am looking forward to using AI in new ways to make my work easier, lower the cost of doing business, and produce higher quality material.