Sean Williams (AutogenAI) — Winning With Words, Not Hype
What you'll learn in this episode:
What proposal pros can do today to be future-ready (skills, workflows, human-in-the-loop)
How to evaluate proposal AI (what actually moves win rate)
What leaders say behind closed doors (the business case anchored in revenue growth)
Christina Carter (00:08) Joining us is Sean Williams, CEO and co-founder of AutogenAI.
In this episode, you’ll learn three things: (1) what proposal people can do today to be ready for the future, (2) how to cut through the noise and find the best technology, and (3) what leaders are really saying behind closed doors—insights that haven’t hit LinkedIn or keynotes yet. Grab a notebook—Sean’s about to drop truth bombs to help you not just keep up, but get ahead.
Christina Carter (00:45) Welcome to The Stargazy Brief, Sean—how are you?
Sean (00:49) Fantastic to be here—thanks for having me.
Christina Carter (00:51) Many listeners know you as AutogenAI’s founder, but you’ve also got deep roots in proposals and the bid space. You come at this differently than many proposal-tech builders.
Sean (01:16) I’ve spent 20 years writing proposals—researcher → bid writer → bid manager—and I still write today. When we submit proposals, I review and write sections (using our software). I love the craft and the language.
The “broken workflow” moment: for years, vendors tried to sell me tools and I never bought because 90% of the proposal process is writing—crafting answers that score. The peripheral bits matter, but the hardest part is writing words that win. I used to say: “Computers can’t do that.”
Then in 2021, my best friend Andy Cowie (DeepMind/AlphaFold research) told me to look at large language models: “They can read and write.” I was skeptical, then blown away—LLMs could generate convincing P. G. Wodehouse-style one-liners; nine were rubbish, one was shockingly good. I realized proposals are an obvious use case.
So I raised capital and built a team half ML engineers/data scientists, half seasoned proposal writers—people who know what it takes to win with words.
Christina Carter (04:14) What misunderstandings are proposal and sales teams holding in 2025?
Sean (04:29) I talk about parrots vs prophets. Two years ago, saying “No one will write a proposal without AI” was prophetic—now it’s parroted.
Tech folks thought humans could be removed—click a button and you’re done—because they didn’t grasp the complexity of proposal writing.
Some proposal veterans underestimated how LLMs transform retrieval and synthesis. Today’s trap: not all AI is created equal. Many tools help you write losing proposals faster—they speed up losing. The point is winning more, and that’s hard to prove in a 30-minute demo—you need volumes of competitive evidence. Clear blue water is coming between vendors that truly lift scores/win rate and those that churn out 2,000 words of 5/10 text. Five out of ten doesn’t win; 9.5/10 does.
Christina Carter (06:55) So what separates 10/10 responses from 5/10?
Sean (07:14) We read as much Cicero as Turing. Great proposals have high propositional density—at least one idea per sentence. Low-density text (AI or human) feels repetitive and empty.
We measure density and, more importantly, whether dense propositions map one-to-one to scoring criteria, win themes, and what the customer wants to hear.
At AutogenAI we use ~100 proprietary benchmarks on proposal text. When proposals score high on these, they’re far more likely to win. It’s binary: win or lose—did you get first past the post?
Christina Carter (08:49) I was taught to write dry and boring early on—then learned to emulate great speeches. Persuasive, high-density writing got better feedback and more wins. Many AI tools still spit out low-density prose.
Sean (09:15) Exactly. Bombast (“we’re the best”) doesn’t win. Evidence wins. “We did this 10 times; here are outcomes.” You can state it dryly or compel with style—but back it with proof.
Sean (10:16) Humans evaluate proposals. We’re logical and emotional. Buyers need to feel understood—mirrored language, relevant proof, and a clear narrative. “Cold” evidence gets stronger when warmed with compelling language.
Christina Carter (10:51) Procurement may use AI to review, but champions still use executive summaries and key sections to sell internally. What are you seeing in evaluation patterns?
Sean (11:31) Think Google Maps. Executives want to zoom out on a 200,000-word proposal: high-level alignment with strategic objectives. LLMs are great for the zoom-out.
Domain owners (e.g., Head of IT Security) will deep-read the 10,000 words on security, line by line. The CEO won’t read those lines—but will rely on the security lead who must.
Christina Carter (12:48) Many worry AI will replace proposal roles. What do you say to them?
Sean (13:14) Proposal pros are storytellers, narrators, and guardians of the org story. AI is a tool, like a shovel or digger—it doesn’t dig by itself. The point of proposals is to tell your story about meeting the customer’s needs best. LLMs help us tell it better—richer evidence, faster synthesis—creating more human-level work, not less.
Christina Carter (14:25) Beyond storytelling, where do proposal pros shape best practice?
Sean (15:04) I hope this elevates proposal writers in the eyes of decision-makers. Leaders care about growth. If a “whiz-kid” tool promises to replace writers, you might “save” $220k—then watch new business collapse. You need a pilot to fly a sophisticated plane. Proposal AI is powerful, but people must fly it.
Christina Carter (16:27) There are 100+ tools. How should leaders choose?
Sean (17:05) Try them. Shortlist ~6 vendors with credible references, security posture, and real customer success. This is people change as much as tech change.
Then run half-day bake-offs on a real live proposal with your team and their trainers. Do that with all six and see which tool actually produces the best result.
Christina Carter (18:09) Some teams build assistants or rely on ChatGPT/Claude. What happens?
Sean (18:32) We often see IT say “we can build it,” then return 6–12 months later. They understand LLMs but not proposals. They oversimplify: “send a prompt, get 1,000 words.” AutogenAI’s 1,000 words are 9.5/10; their 1,000 words are 2/10. Ask: are you building your own spreadsheet, browser, or search engine? If you choose to build your own Google, you’re either a genius—or insane. Specialization matters.
Christina Carter (20:15) This space moves fast. If your tool is a year behind, you miss the gains. On hallucinations—how does AutogenAI tackle accuracy?
Sean (21:22) We launched our first version in Sept 2022 (six weeks before ChatGPT). Internally, we nicknamed the challenge the “Boris Johnson problem”: confident, fluent answers that aren’t grounded.
I asked our chief research scientist to “build truth.” He replied: “Give me a two-sentence definition of truth.” Philosophy’s debated that for 3,000 years—so we defined truth in proposals pragmatically:
A claim is “true” if we can show the sources, and
A human decides to accept those sources. This led us (early) to Retrieval-Augmented Generation (RAG). When we generate, we show sources: “from Doc A, Doc B, Site C.” Users can disable web, restrict to specific libraries, or set source preferences. We implement epistemic hierarchies: prefer recent over stale, consistent sources over outliers, etc.—borrowing from historiography to codify authenticity. Bottom line: we maximize traceability and put humans in the loop to choose. No system makes truth perfect; your docs can be wrong. Review is non-negotiable.
Christina Carter (25:40) Agreed—always review. What trends should proposal pros get ahead of?
Sean (25:45) Fundamentals don’t change—structure, evidence, ordering, language, redrafting. AI won’t make you tea or do your entire job, nor is it useless. It concertinas parts of the process—compressing time and boosting quality. Because the field evolves fast, you need a technology partner that ships frequent improvements and provides change-management support. Our customer success team pairs every release with hands-on enablement: what’s new, how to use it, and how it fits your org workflows. Powerful tools are pointless if teams don’t know how to use them.
Christina Carter (28:24) Let’s step back from AI. Buying behavior is changing—what are leaders saying privately that proposal pros should know?
Sean (28:51) Proposals exist to win business. Anchor the business case in revenue and win rate, not headcount reduction.
Example: you submit 40 proposals/year, win 50%, each worth $1M → $20M won. With AutogenAI, submit 80 proposals and lift win rate to 60% → $48M won.
Don’t celebrate saving $150k on headcount while missing $20M+ in revenue. The outcome we’re proud of: enterprises that double headcount and profit as growth compounds—our software helped enable that.
Christina Carter (30:30)
Where can people learn more about you and AutogenAI?
Sean (30:37)
Visit autogenai.com to book a demo and read our thought pieces from the last 3–4 years. We’re on LinkedIn; I’m Sean Williams. Email: sean@autogenai.com—reach out directly.
Christina Carter (31:19) Links are in the show notes—thanks for joining, Sean.
Sean (31:29) Thanks, Christina—really enjoyed it.
Q1: What separates winning proposal AI from “faster losing”? Tools that raise propositional density, map ideas to scoring criteria and win themes, and provide traceable evidence—consistently producing 9+/10 scoring text.
Q2: How does AutogenAI reduce hallucinations? Early adoption of RAG, explicit source display, user controls (disable web, restrict libraries), and epistemic hierarchies (recency/consistency). Always human-in-the-loop.
Q3: How should we evaluate vendors? Shortlist ~6 credible, secure vendors with real CS teams. Run half-day bake-offs on a live RFP with your people and their trainers. Pick the one that wins on outcomes.
Q4: Should we build our own? DIY often underestimates proposal complexity. Many teams return after 6–12 months. Specialization matters—use a product built by proposal experts + ML experts
Q5: Will AI replace proposal pros? No. Pros are storytellers and pilots. AI is the aircraft—you still need skilled people to fly it, apply evidence, and craft persuasive, on-brand narratives.
Q6: What skills should proposal pros build now? High-density persuasive writing, evidence curation, scoring alignment, workflow design, and tool literacy (RAG, retrieval controls, human-review loops).
AutogenAI: autogenai.com
Sean Williams (LinkedIn): https://www.linkedin.com/in/sean-williams-88336225/
stargazy community: https://stargazy.circle.so/join?invitation_token=0856b517503bca21eecbae1d058313543675481b-28d54c10-c886-4708-8b3b-ffd62cd3c935
stargazy website to find the perfect proposal tech for you: stargazy.io