How I Team Up with GenAI to Craft Conference Talk Proposals
TL;DR: GenAI can help you write talk proposals, but if you use it without injecting your own curiosity and humanness, reviewers can tell. Here’s the process I use to shape real, human abstracts that reflect my curiosity and experience. I developed this process for technical talks, but it adapts well to other formats too.
I’ve spoken at a lot of conferences, big and small. I’ve given keynotes. Breakouts. Workshops. I’ve been around the block and written many conference talk proposals. I’ve also been on program committees and reviewed others’ proposals. There’s always room for improvement, but I like to think I know what good is.
In my view, here are some elements of a good talk proposal:
- A clear, engaging idea (or a run-of-the-mill idea with a compelling angle)
- Some clear details about the talk, like concrete examples or stories, or talk format and tools covered
- Clear audience
- Clear takeaways
As you can imagine, the widespread use of GenAI to generate talk proposals has changed the game for both submitters and reviewers. As a submitter, you can type just one line into ChatGPT, for example:
Write me a KubeCon talk abstract on Validating Admission Policy
And it produces something that, on the surface, looks pretty good!*
Abstract: As Kubernetes adoption continues to grow across organizations, ensuring workloads meet security, compliance, and operational standards before hitting the cluster is more critical than ever. Enter Validating Admission Policies—a powerful, built-in alternative to third-party admission controllers and dynamic webhook solutions.
In this talk, we’ll dive into the evolution and capabilities of Validating Admission Policies in Kubernetes, exploring how they provide a declarative, in-cluster mechanism for enforcing rules at admission time. We’ll cover practical use cases like blocking privileged containers, restricting image registries, and enforcing labeling conventions—all without the overhead of managing external webhooks.
Attendees will learn: – How Validating Admission Policies work under the hood – The differences between validating policies and mutating policies – How to write and deploy policies using the Common Expression Language (CEL) – Best practices for policy testing and gradual rollout – Real-world scenarios and lessons from production use
Whether you’re a platform engineer aiming to harden your clusters, or a developer advocating for better guardrails, this session will help you move from theory to practice and adopt a more secure and manageable Kubernetes environment—using only what the API server gives you out of the box.
But as a submitter, you should be asking yourself: If literally everyone can make a good-looking abstract this easily, why would a reviewer choose mine?
And the truthful answer is: They wouldn’t. And they won’t.
“But Whitney!” you might be saying, “This strongly hits all 4 of your 4 criteria of a good conference talk proposal! A clear idea, clear details, clear audience, and clear takeaways!”
You’re right. You’re so right. Which is why you want to submit this so badly, don’t you? But don’t do it!!!
In the age of GenAI, let’s add one more criteria to our list of what makes a good talk proposal:
- A clear, engaging idea (or a run-of-the-mill idea with a compelling angle)
- Some clear details about the talk, like concrete examples or stories, or talk format and tools covered
- Clear audience
- Clear takeaways
- Not obviously written by GenAI
As someone who has looked at hundreds of talk proposals, I can spot one that is written by GenAI from a mile away. And you know what it tells me? This person did not put any effort into their proposal. And if they didn’t put any effort into their proposal, why should I believe that they’ll put effort into making their talk if they get accepted?
But what are these AI tells that experienced reviewers see? Let’s go over some of them.
An Incomplete List of GenAI Tells
How many of these do you see in the abstract above?
- Heavy use of em dashes
- Repeated “from ___ to ___” constructions
- Paired adjectives (humans usually use one)
- “In this talk, we’ll…”
- Generic three-part lists
- Long, dense sentences
- Bullet lists
- “Whether you’re ___ or ___, this session will ___”
But more than these tells, GenAI abstracts give off a feeling. Like, this abstract is too smooth, and it has no indication that the author has any actual experience in tech. Or, come to think of it, there’s no indication that the author has any actual experience at being a human.
So let’s amend our “what makes a good talk proposal” criteria one more time and add:
- A clear, engaging idea (or a run-of-the-mill idea with a compelling angle)
- Some clear details about the talk, like concrete examples or stories, or talk format and tools covered
- Clear audience
- Clear takeaways
- Not obviously written by GenAI
- Human emotion and personality
The last two are admittedly similar. But you can manually write a proposal that is devoid of emotion and personality, you don’t need GenAI for that. And humanness is EVERYTHING, especially now with GenAI in the mix.
Now I’m not saying to abandon GenAI altogether when working on that proposal. That’s not reasonable. GenAI is truly helpful sometimes, and we’ve all got important lives to live. I’m all for outsourcing what we can!
To make a conference talk abstract, here’s the process I work through. I use GenAI interactively and use it more like a teammate who can help tease a talk/story idea out of my lived experience and interests.
My Process
Using this process, I transformed my ‘Kubernetes Validating Admission Policies’ talk idea from a perfectly fine (but soulless) GenAI draft into something that is closer to my voice and interests. See the prompts below!
Step One: Brainstorm Buddy
Ask your GenAI tool to ask you questions to explore possible talk subjects, choose one, and discuss what makes it interesting.
Step Two: Shape the Story
Take your chosen talk idea and make it into a talk outline. Use this process to verify that you would feel happy to give this talk, and to give feedback and examples. Also - save this outline! If your talk gets accepted, here you have a great starting point from which to build your talk.
Step Three: Write the Abstract
Generate the abstract! But first give GenAI context, guidance, and guardrails.
Step Four: Iterate
No matter how awesome this process is, and how awesome you are, GenAI probably won’t get it completely right on the first try. Be prepared to do some critiquing and iteration.
Step Five: Review Like a Reviewer
Before you submit, ask GenAI to role-play as a program committee member. Ask it to read your abstract and give honest feedback. Is it clear? Does it stand out? Would it get selected? Why or why not?
Okay, Whitney, stop talking and give me the prompts already!
I hear you, friend. Let’s do this.
Prompt 1: Brainstorm Buddy
Use this prompt to explore raw, personal ideas through conversation
Ask me one question at a time to help surface something human, specific, and worth exploring in public.
This might come from: – Something I’ve been working on – A moment I learned something the hard way – A pattern I’ve noticed – A question I haven’t answered yet – A decision I’m wrestling with
Don’t just ask what I know. Ask what I’m curious about, what I’m drawn to, or what I’d like to learn more about. Follow up when something seems alive. Stay conversational.
Ask me if a live demo might help tell the story. Would it clarify something? Make it more engaging? Help the audience connect? If I say yes, explore what that demo might look like and how it would fit into the flow.
We’re not building a talk outline yet. When we’ve landed on a promising angle or thread, write up a short, natural-language summary of what we surfaced. Then ask if I’m ready to move on to shaping the talk, or whether I want to add more details.
Prompt 2: Shape the Story
Use this prompt to turn an idea into a talk outline without losing voice or vulnerability. (Feel free to adjust the target talk length if needed)
That doesn’t mean keeping everything. Focus on what supports the core idea. Clarity is more important than completeness.
Help me build a structure that: – Reflects my point of view and lived experience – Highlights emotional or technical tension – Builds toward a moment of insight, reflection, or takeaway – Stays focused enough for a 30-minute talk
Keep the voice personal, thoughtful, and exploratory. Don’t sand off the edges to sound overly polished or generic.
Prompt 3: Write the Abstract
Use this to generate a conference-ready abstract that doesn’t sound AI-written. Add the name of the conference that you’re submitting to. Bonus points if you feed it a link to a past schedule too.
This is for «Conference Name». Please look up the conference online to understand: – The tone and style of accepted talks from recent years – The CFP format, character limits, and content expectations – The kind of audience this conference attracts
Write in third person, using the speaker’s point of view (e.g., “This session explores…” is fine—but avoid first person like “I” or “we”).
Aim for a tone that is grounded, curious, and human. It should feel like a thoughtful engineer exploring a real question—not like a marketing blurb or a how-to guide.
Avoid these common GenAI tells: – Overused em dashes—use varied punctuation – “From ___ to ___” constructions – Paired adjectives like “clear, actionable insights” – Generic three-part lists – Robotic transitions like “we’ll also look at…”
Instead: – Start with a relatable tension, question, or moment – Include specific, technical grounding—not just abstract vibes – Convey a strong sense of what actionable takeaway(s) the audience will explore, gain, or reflect on
Rhetorical questions are welcome if they spark curiosity—but tie them to something concrete.
Enjoy the Result!
This process transformed my ‘Validating Admission Policy’ talk idea and abstract from this:
Abstract: As Kubernetes adoption continues to grow across organizations, ensuring workloads meet security, compliance, and operational standards before hitting the cluster is more critical than ever. Enter Validating Admission Policies—a powerful, built-in alternative to third-party admission controllers and dynamic webhook solutions. In this talk, we’ll dive into the evolution and capabilities of Validating Admission Policies in Kubernetes, exploring how they provide a declarative, in-cluster mechanism for enforcing rules at admission time. We’ll cover practical use cases like blocking privileged containers, restricting image registries, and enforcing labeling conventions—all without the overhead of managing external webhooks.
Attendees will learn: – How Validating Admission Policies work under the hood – The differences between validating policies and mutating policies – How to write and deploy policies using the Common Expression Language (CEL) – Best practices for policy testing and gradual rollout – Real-world scenarios and lessons from production use
Whether you’re a platform engineer aiming to harden your clusters, or a developer advocating for better guardrails, this session will help you move from theory to practice and adopt a more secure and manageable Kubernetes environment—using only what the API server gives you out of the box.
to this**:
Abstract: Enforcing policies in Kubernetes often begins with a simple question: What do we need to enforce, and why? For many, the answer starts with Kubernetes' native Validating Admission Policies (VAP)—a lightweight, declarative approach that feels intuitive and clean. But as platform teams scale, the question evolves: Is this enough?
This session explores that very question through a real-world lens. The speaker, currently building a proof-of-concept platform, began by implementing basic policies using VAP. However, as the platform’s needs grew, so did the complexity of the policies. This led to an exploration of Kyverno, a policy engine designed specifically for Kubernetes, to understand its capabilities and whether it offered advantages that VAP did not.
Through live demonstrations, the audience will see the same policy implemented first with VAP and then with Kyverno. The session will highlight differences in developer experience, observability, and the tools' approaches to policy enforcement. By comparing these tools side by side, the speaker aims to provide insights into when each is appropriate, without advocating for one over the other.
Attendees will leave with a clearer understanding of the trade-offs involved in choosing between VAP and Kyverno, equipped to make informed decisions about policy enforcement in their own Kubernetes environments.
Before you Submit
Iterate on the abstract and make it your own. You’ll likely be messing with character count anyway since GenAI is terrible at that. As you iterate, don’t forget to ‘review like a reviewer.’ Ask GenAI to role-play as a member of the program committee and tell you whether they’d accept your talk and why.
LMKWYT!
If you try out my process, let me know what you think! And if you have any suggestions for improvements, I’m all ears! Let’s use GenAI to make us all more human, instead of more robotic. I firmly believe we can do it.
*I pinky swear that I only asked ChatGPT to Write me a KubeCon talk abstract on Validating Admission Policy one time and copy-pasted the first and only result.
**This is the first output of Prompt 3, without any additional smoothing or iteration from me. I added the title later, after its own round of GenAI-assisted iteration.