The New Top of the Funnel
There's a buyer out there right now deciding whether your company makes their shortlist. You are not in the room. Neither is your SDR, your website, or your carefully built nurture track. The thing in the room is an AI, and it's reading everything it can find about you while your buyer watches.
This is the new top of the funnel. A machine doing the research, comparing you against every alternative, and handing your buyer a verdict before a single human on your side gets a chance to say a word.
We have spent years optimising for a buyer who browses, clicks, downloads a PDF, and slowly raises their hand. That buyer is being replaced by one who delegates all of it to software. And most marketing teams don't even know what that does to every metric they care about.
This is also the reason a few of you have asked me directly why I've gone all in on agentic GTM, and why every conversation I have these days seems to circle back to one phrase: the context layer.
The Short Version
The buyer started using machines first, so we don't get to stay human-only on our side of the table. It is time to understand that the machines themselves were never the hard part, the hard part is what you feed them.
The Signals
In March, LinkedIn, working with Bain and NewtonX, published what I think is the most important B2B research of the year, and the headline number is worth reading twice.
94% | 20x |
|---|---|
Buyers Use LLMs First Of buying groups now use LLMs like ChatGPT or Gemini before they ever engage sales. Source: LinkedIn, Bain & NewtonX | Familiarity Premium Brands were over 20 times more likely to be bought when the entire buying group |
68% | 90% |
|---|---|
Zero-Click Searches Of US Google searches now end without a single click, up from 60% just two years ago. Source: SparkToro analysis | AI-Intermediated by 2028 Gartner expects 90% of B2B buying to be intermediated by AI agents, pushing $15 trillion through agent-to-agent exchanges |
CMO Readiness Gap
Gartner's 2026 Spend Survey shows budgets flat at 7.8% of revenue, with a fast-growing AI line carved out inside that flat number.
70% | 30% |
|---|---|
Call AI Leadership Critical | But only 30% say they’re actually ready |
Read those five signals on their own and they look like five separate news items. They aren't. Buyers moved into machines, the machines reward trust you built in advance, the open web is closing behind them, the whole thing has a 2028 deadline, and most teams are pouring money into the technology without the foundation it has to run on. That's one story told five ways.
What the Signals Actually Say
Here's what those five things add up to. The buy side of the market has rebuilt itself around machines that pull context together, while the sell side is still organised the way it always was, around channels that pump out content. That mismatch is the whole problem.
The Agent Has No Friction
You can feel why if you picture what an AI research agent actually does. It doesn't get tired and it doesn't reward volume. It reads everything, weighs the evidence, and ranks you against every alternative in the time it takes you to refill your coffee. Our funnels were built to manage human friction, to nudge a distracted person through stages. The agent has no friction to manage, and so the metrics we built around that friction stop meaning anything.
The Disease Underneath the Symptoms
The funnels that punish buyers for behaving like actual humans, the data that quietly lies to you, the five conflicting versions of revenue truth that show up in the same room. Every one of those pieces was circling the same thing.
Your GTM Has No Memory
Each tool sees one slice of the buyer and forgets the rest the moment the handoff happens. We got away with that for years because a sharp SDR could paper over the cracks. You can't paper over them when the thing evaluating you reads everything, and the thing selling for you needs everything.
Why I Went Full Agentic
There is a widespread argument today: now that buyers get everything through AI, the company wants to track and gate and hoard, the customer just wants answers fast, and the customer always wins. So the only move left is to do good marketing and HOPE the right people find their way back.
But hope isn't a strategy, and it definitely isn't one you can defend to a CFO.
There are two camps of marketers emerging from here: One camp wants to measure everything and assign credit to every touch, the company's old instinct.
The other camp, panicking because AI broke the old metrics, wants to measure nothing and trust the universe. Both are wrong, and they're wrong for the same reason. AI didn't kill measurement. It killed the layer you were measuring.
Why I Made the Bet
I didn't go all in on agentic GTM because it's fashionable. I did it because the other side of the table moved first, and you can't sit across your prospect who brought a machine and you decide to bring only your charm. You have to make a larger bet and bring your own agents on the table.
All said and done, the fact is that the agents themselves will not solve every problem. They're turning into commodity inference, and every vendor will be selling you one by Christmas.
"What decides whether an agent makes you smarter or just makes your existing chaos move faster is the memory you hand it."
The Context Layer: Both Sides of the Table
Think about what the two machines in this new market each need from you.
The Buyer's Agent Needs Your Evidence
Reviews, case studies, real customer proof, the trail of signals it can actually verify. When that evidence is thin or scattered across the internet, the agent doesn't come back with a bad answer about you. It comes back with nothing.
Your own Selling Agents Need Your Truth
Which accounts are heating up, what the buying group has already touched, what worked on a deal like this last time, why the last one stalled at procurement. Every one of those questions dies inside a stack where the call recording, the CRM, the campaign data and the community thread have never once spoken to each other.
It's the same wall on both sides, and the only thing that gets you through it is one connected memory of every entity, event, interaction and decision across your GTM motion, in a form that humans and machines can both act on. That is what I mean by the context layer.
The Scale Equation
A 10-person GTM team is going to operate at the scale of a 500-person. You can have an army of agents, but they can only work when you give them complete memory.
What To Do Next
Three concrete moves you can make this week, before the market builds its own version of you without you.
1. Interrogate the Machine
Take the twenty questions your buyers actually ask and run them through ChatGPT and Gemini yourself. Whatever comes back is your real top of funnel now, whether you like it or not.
If you're missing from the answer or showing up wrong in it, more clever copy won't save you. What moves the needle is harder evidence: reviews, named customer proof, numbers a machine can check.
2. Run the Memory Test
Pick one account that closed last quarter and try to reconstruct its full journey across every system you own. Calls, campaigns, community, CRM, web.
If your team needs more than a day to do it, you can't brief an agent properly. And the buyer's agent can't find your truth either. That's the same gap showing up from the inside.
3. Count the Committee, Not the Channel
That 20x familiarity number is the metric that's going to survive all of this.
So stop asking which channel deserves the credit and start asking how much of the buying committee would recognise you before the first call ever happens, especially the people in finance and legal who kill deals without ever telling you.
End Notes
For years we measured what buyers did out in the open, on our sites and inside our funnels. That stage is gone. The decisions now happen in a room we can't enter, between a buyer and a machine, and the machine decides using whatever it can piece together about us.
That cuts both ways, which is the whole point. The buyer's machine judges you on the evidence it can find. Your own machines are only as good as the memory you can give them. Same gap, and one thing closes it: a connected memory of your buyers that humans and agents can both draw on. Build that, and you're legible to the machine on the outside and sharp with your own agents on the inside. Skip it, and you'll spend the next two years watching your dashboards predict nothing while you wonder what changed.
P.S. The context layer isn't a metaphor for us. It's the thing we've spent four years actually building at RevSure, one context graph of your whole GTM motion that an agent or a CMO can act on. If the memory test in this issue made you wince, write to me at [email protected]