- The Product-Led Geek
- Posts
- đź‘“ When the PLG Funnel Shifts Into LLMs
đź‘“ When the PLG Funnel Shifts Into LLMs
Welcome folks! đź‘‹
This edition of The Product-Led Geek is authored by Matt Henderson (Director of Growth at Sentry) and it’s an absolute banger.
It will take 8 minutes to read and you’ll learn:
About the shift coming for users to convert on your PLG apps in LLMs rather than your web funnel
How chat is now a platform - a colossal shift for how we view websites
Why most web transactions are bots but sites haven’t yet adapted
The steps we can all take to unlock growth by improving our app’s accessibility to LLM/agent workflows.
As usual with guest posts here it’s injected with my commentary throughout.
Let’s go!

GEEK OUT
When the PLG Funnel Shifts Into LLMs
OpenAI recently announced that they are partnering with ecommerce sites to offer native chat checkout, meaning users can buy something straight from their chatGPT conversation.
They followed that up with announcing an apps platform, which involves them suggesting apps when they’re relevant to the conversation and then running an app from your chat.
Picture this conversation:

You can feel this coming right?
B2B always comes with compliance, security, governance, enterprise-concerns and can be harder than buying some shoes or a hat online, but for PLG, this seems almost inevitable.
It would be even easier for companies that offer usage-based or per credit pricing.
“G” at Hypergrowth Partners puts it clearly in his post
Imagine you’re a marketer at Cursor who needs contact data for 50,000 CTOs at the top engineering companies in the US.
Instead of logging into ZoomInfo, navigating their interface, building lists, and managing credits, you just tell ChatGPT: “Pull me 50,000 COT contacts from companies with the most engineers.”
You can now already have ChatGPT open Zillow in-chat (after you’ve signed up and integrated it) and have it find you a specific home.
Right now it’s just for consumer apps like Spotify, Alltrails, Open Table, Target, and others coming, but with Microsoft pulling strings behind the scenes of OpenAI, we know the focus will eventually include the enterprise.
Geek note: LLMs don’t just replace the top of the PLG funnel - they instrument it. Every chat interaction becomes a measurable query, a synthetic “intent event.”
Growth teams need to treat these as new search signals, trackable the same way we once measured impressions or SERP positions. The challenge: no public API for this data yet, so you should start logging and classifying inbound LLM traffic now.
Here’s the thing -> we are already halfway to this exact workflow applying to PLG.
We have the Model Context Protocol.
MCP is just the rails or the bridge between agent and system, but how close are we really to an agent or LLM doing the provisioning of a user with their identity and the user skipping a website entirely?
Your web funnel would be drastically different or might completely evaporate.
We’re watching the early formation of searchless discovery.
MCP is about control surfaces.
Once agents can authenticate, run APIs, and track identity, they also control which tools are suggested.
That’s the new ranking algorithm.
Think of it as SEO collapsing into “LLMO”:
— Ben Williams (@theplgeek)
9:21 AM • Oct 23, 2025
This all begs the questions:
When will it be standard for agents and LLMs to be doing the conversion piece for B2B (especially PLG)?
How will your website and app capitalize on Chat as a new growth platform beyond referral traffic?
Chat as a platform will lead to agentic conversion
To understand what this might look like for B2B, we can simply look at what they are doing with ecom.
OpenAI introduced “Agentic Commerce Protocol (ACP)” which is just a fancy way to say a way for apps to natively integrate into ChatGPT for payments, product feeds, etc.
On their site it says ACP “enables a conversation between buyers, their AI agents, and businesses to complete a purchase“, so why wouldn’t that eventually apply for Product-led growth applications?
Geek note: This is the “API-ification” of sales. ACP is more than a checkout layer - it’s conversion as protocol. Once your product is represented as structured endpoints instead of web flows, discoverability becomes a data problem: how well your endpoints are documented, described, and semantically tagged for agents to invoke. That’s why schema and developer docs now are your marketing.
PS: Also not to be confused with “Agent Client Protocol” - the other ACP - yes AI has a bad naming problem!
What’s really shifting is that chat is becoming a platform that is and will likely be embedded into all OS’s (see Meta’s glasses, Gemini in Pixel phones, Apple AI, etc.).
This is a tacit shift away from a search on a browser as the starting place for digital transactions.
Look no further than Atlas, OpenAI’s new browser as evidence.
It’s a step towards an agent/chat as the foundation for everything. Just as the browser with web apps was a shift away from opening your desktop and having everything you needed in the 2000s/2010s.
Microsoft no doubt has learned from Chrome’s dominance and seems to be behind doubling down on chat being a platform.
It looks to be a blue ocean opportunity for growth teams.
Geek note: I encourage everyone to think beyond the framing of Atlas as “yet another browser”
Atlas is OpenAI’s move to capture the missing telemetry layer.
ChatGPT already sees what users ask, but not what they do next.
Atlas closes that loop.
By owning the browser, OpenAI can watch how LLM-driven intent flows into real-world action - which docs are opened, which links are clicked, which products are boughts, which tasks fail.
It’s the same playbook Google ran with Chrome: control the interface, own the data, shape the standards.
For developers and growth teams, that’s a black box.
If OpenAI gets full visibility into “what users tried to solve but didn’t succeed at,” they’ll own the new keyword-gap data of the AI era.
What I’m trying to do with DevTune is to build the outside telemetry - measuring how LLMs perceive, recommend, and interact with your product when you can’t see inside the OpenAI (and other LLM providers) data firehose.

This is the shift: A b2b buyer wouldn’t buy directly from the SERP in Google, but they would buy after an aided conversation with chat.
1 out of every 3 consumers already would allow agents to make purchases and 48% of buyers last year used AI tools to research software.
This is without a doubt heading towards 100% adoption.
Overall this means that rather than your SaaS website looking like, let’s say, a shoe storefront, it will likely look more like a shoe fulfillment center.

Geek note: When half your visitors are bots, your “conversion rate” becomes a machine-read metric.
Sign-ups will still be an important KPI but not the only one.
I can see a world where Agent-Completable Coverage (something like the % of core actions that an LLM can perform end-to-end via your public API or docs) increasingly matters.
As Matt highlights, the PLG funnel is shifting from humans clicking buttons to machines executing playbooks.
Growth lever: turning your website into a fulfillment center
And we are JUST on the precipice of crawlers and agents hockeysticking that number to the moon in the next few years, especially with AI embedded browsers.
So the question becomes: how accessible is your website and digital properties for AI?

Shows % of bot traffic over time (Imperva bad bot report)
Geek note: Once half your traffic is bots, your UX stops being about pixels and starts being about parsing.
LLMs and crawlers don’t care how your homepage looks - they care how your docs, schemas, and APIs describe themselves.
The next generation of SEO is structural: clarity, context, and consistent metadata.
And the conversion funnel is shifting toward successful function calls.
If your website is like a storefront made for humans, and the majority of interactions with your store are bots or agents, you have a misalignment.
Here are a few ways of thinking about the shift using the analogy:
Web as a storefront | Web as a fulfillment center |
|---|---|
Major time/effort investment into a welcome experience | Less fancy store front, more of a well structured backend |
Only way to buy is via demo or through site/app | Agents provision accounts in your app on behalf of users. Custom demos/experiences. |
Experience is focused on the most popular use case of your product. | Everything is custom to each buyer’s needs |
Gated behind multiple steps/clicks | Deeply integrated into chat as a platform |
Everything packaged together in plans | Usage based and made-to-order |
Hundreds of marketing pages convincing a human they have a need for their service. | Great documentation on how agents and less-so people can get what they need. |
Potential ways to optimize your channels for conversion:
Here are a few things I will certainly be looking out for to put on the roadmap:
Be known for what you do:
Explain specific solutions tailored to what your product does for different spaces, industries, etc. This could look like solutions pages, template galleries, recipes, integrations pages, etc.
Make your website accessible to bots/agents first, then humans. Focus on great documentation of how your product works technically.
ex: Sentry Docs
Make it easy to be adopted by this new workflow
Integrate your app with chat (at least ChatGPT as they open up apps platform out “later this year”) - they launched with only a handful of apps, but this is a complete greenfield and they will almost certainly recommend your product over anyone else’s if it’s in their platform.
Consider the future of optimizing your devtool for AI-first discovery and adoption with tools like DevTune.
Deliver a bomb experience once buyers land in your product
Revisit your onboarding and first 30 day experience because it might be the first place a user lands rather than your website where they can see/learn about other features and products.
One clarification on all of this is that I’m not saying that brand feel and taste doesn’t matter and that everything should be built for bots.
I actually think this matters more than ever, because you have less touches with a prospect before they choose a product, and you need to make an impression (not only this but brand matters increasingly so for optimizing your website for discoverability).
People will still come to your website, experience your advertising, etc.
And those emotional interactions matter so much more now.
Evidenced by impressions until the first website visit increasing from 723 to 894 (23% increase) last year i.e. less heading to a website, more brand exposure off site.
Geek note: As chat becomes the new discovery layer, the concept of “impression” fragments. One model might mention you 10,000 times, another might recommend your competitor 100 times - and you’ll never see either in traditional analytics. This points to the importance of measuring representation. How often do LLMs name, describe, or recommend your brand? How does that compare to your competitors?
But what will this all look like in a few years?
My wager would be that PLG is condensing to artful and authentic brand touches + a product experience that lines up the new ecosystem.
Enjoying this content? Subscribe to get every post direct to your inbox!

THAT’S A WRAP
Before you go, here are 3 ways I can help:
Take the FREE Learning Velocity Index assessment - Discover how your team's ability to learn and leverage learnings stacks up in the product-led world. Takes 2 minutes and you get free advice.
Book a free 1:1 consultation call with me - I keep a handful of slots open each week for founders and product growth leaders to explore working together and get some free advice along the way. Book a call.
Sponsor this newsletter - Reach over 7600 founders, leaders and operators working in product and growth at some of the world’s best tech companies including Paypal, Adobe, Canva, Miro, Amplitude, Google, Meta, Tailscale, Twilio and Salesforce.
That’s all for today,
If there are any product, growth or leadership topics that you’d like me to write about, just hit reply to this email or leave a comment and let me know!
And if you enjoyed this post, consider upgrading to a VIG Membership to get the full Product-Led Geek experience and access to every post in the archive including all guides.
Until next time!
— Ben
RATE THIS POST (1 CLICK - DON'T BE SHY!)Your feedback helps me improve my content |
Reply