Proposals: From AI Draft to Human Win

When cleaning companies think about artificial intelligence (AI), they often picture faster answers, auto-filled forms, and proposals that almost write themselves.

But for proposal expert Chris Arlen, president of revenue-IQ, AI is not a magic proposal machine. It is a powerful tool-but only in the hands of people who still do the hard work of thinking, researching, and persuading.

"In our world, if AI is the tool, you are not the tool," Arlen acknowledged. "To produce a winning proposal, you still have to do the basics. You've got to write a persuasive story. You've got to make the customer the hero."

That sounds simple. But as Arlen reminded, doing those "basics" requires something AI cannot provide: Deep, specific knowledge about the customer and their real issues.

Proposals are not commodities

Many companies worry that proposals are sliding into commodity territory-everyone answering the same request for proposal (RFP) questions and letting price decide the winner. Arlen pushed back on that assumption.

"My clients are not necessarily happy about that, that it would be a commodity, where everything's exactly equal, and price is the only determinant," he said. "Luckily, it's not about that. It is about the value you get. The customers are buying the value for what they pay. And that's always been the challenge for them to figure out."

In Arlen's view, AI does not change that fundamental equation. It simply changes how quickly and efficiently contractors can get to their starting point. The real differentiation still comes from "inside baseball"-the confidential, context-rich client information not in an AI model's training data.

"You always want to base it, as much as possible, on inside baseball, on customer confidential information," Arlen said. "AI can't get that. So, you've got to get that."

That "inside baseball" becomes the proposal theme-what you lean on to write a story that resonates with evaluators without ever calling out their past failures directly. When you write to that theme, he said, "Now all of a sudden, you're speaking to something that AI doesn't have access to. And when the evaluators read that, it's like touching an emotional nerve."

The hype cycle and AI proposals

Arlen views current proposal tools through the lens of the classic technology Hype Cycle^™, popularized by Gartner^®.

"New technology comes out…it goes up to the peak of inflated expectations," he observed. "And then as people spend billions, it falls down into the trough of disillusionment. And then over time, they start to figure out, okay, this is what it actually does."

"Right now, the proposal AI tools, I think, are kind of at the peak of inflated expectations that are on the downward trend towards that trough of disillusionment," he added.

For now, Arlen is leaning heavily on generative AI-tools that create content-while watching the emerging wave of "agentic" AI, which can take actions toward a goal. "I'm still waiting. I'm a little cautious on agentic just at the moment," he said. "I will be hitting it hard, I'm sure. Right now, I'm looking at the generative."

Raising the floor, not the ceiling

Because AI is so widely available, Arlen believes it has raised the minimum standard of proposal writing-but not the top end.

"If you're new, if you're not good at what you're writing, man, just go to AI and say, 'Give me a proposal.' You're going to get something," he said. "The bad news is everybody else has it too. So, all your competitors can do the exact same thing. And what that means is now the floor of proposal writing has just…risen."

What still separates a winning proposal from a generic one is not the tool, but the human who uses it, and the depth of insight behind the story.

How AI transforms research and analysis

Long before Arlen writes a solution or drafts an answer, he focuses on analysis. That has not changed with AI; it has only accelerated.

"In the past, I'd have to go on site, talk to and pull teeth from salespeople to figure out what's going on with this client," he noted. A few basic web searches might add some surface-level detail, but that was it.

Now, he works from a detailed, reusable prompt. "I have like a 500- or 1,000-word prompt that I use that gets me very specific things," Arlen said. "I want to know about the business. I want to know about the evaluators. I want to know about the department. I want to know about potential competitors. I can get all that stuff, boom. And that becomes the basis of designing a solution."

He does not retype that thousand-word prompt each time. "I do it once," he said. "I even have placeholders in that prompt. A placeholder for the company name, the project name, the department name, etc. This is my end user, the people running that. So, I can put that in; it's a one-time load, and it's there. Good to go."

But AI research only gives him the "above the waterline" view. He still goes back to sales and operations to ask: What is really going on? What is below the waterline? That human intelligence is what fuels the proposal theme and the story.

What to automate and the 60-10-30 Arlen rule

So, where does AI truly shine in proposal work? Arlen pointed to several areas:

• Shredding the RFP to summarize requirements and risks.
• Opportunity analysis-whether you should even be bidding.
• First drafts of answers, using your existing content as a base.
• Checking basic RFP compliance: Did you answer every question?

"That's something that's another hard thing for my clients," he said. "They'll have a ton of documents, but they don't have the information at their fingertips. They'll say, I remember writing an answer to a question like that. Now, we can load that up into a big doc."

From there, AI can assemble solid first drafts-but it cannot finish the job alone. Arlen uses what he calls his "60-10-30 rule."

"With really good prompts and uploaded project files, I can get about 60% of what I want to that particular answer for that RFP question," he said. "Any further iterations really only get 10% more improvement on that."

The final 30% is human-and critical. "You've got to be checking that stuff," he said. Whether it is factual qualification questions ("How many years in business? How many people? What certifications?") or applied "How will you do this?" questions, generative AI can easily hallucinate answers or promise the impossible.

There's a popular AI saying that "Generative AI is 97% accurate, but 100% confident," Arlen said. "You're always going to get an answer. The problem is you've got to be checking that stuff."

That means no proposal should ever go out the door as an AI-only document.

"We want to secure contracts. We want our proposals to win," he asserted. "And that takes being effective, which means being persuasive. And that's a human thing. Customers make decisions on emotion, and then they look for the facts to justify it."

Getting started with AI on your next proposal

For cleaning companies that have not yet used AI in proposal work, Arlen's advice is straightforward: Start small but start smart.

In the rare instance that you are the only proposer on a small, one-off opportunity, "use AI to shoot it out straight," he said. "Why? Because the customer doesn't have to compare it to somebody else."

But when there is real competition, you need the whole process. "You'd better do the full nine yards," Arlen suggested. That includes shredding the RFP, doing serious analysis on the company, its evaluators, and competitors, and pulling your best existing content into the mix.

Then, use AI for what it does best-speed, structure, and first drafts-while keeping humans firmly in charge of the story.

"Use it, spend your creative time on something else, use the tool to help you with the mundane," Arlen said. "But remember, write like a human because your customers are."

BONUS VIDEO CONTENT:

issa.com/proposals