In B2B sales, the goal has always been the same: get the CEO on the phone. Get past the gatekeeper, get the meeting, build the relationship. I ran a consulting firm for 18 years, and a lot of that game ran through that one phone call.
Here's the call I think is coming.
"I'd love 20 minutes with your CEO."
"She won't take the call. Send it to her agent."
Don’t want to read? I used Eleven Labs to clone my voice and Claude to build a video for this post. Listen/watch on Youtube:
That sounds like a joke. In software it's already happening. Jason Lemkin runs SaaStr and has been writing about the AI agents his team uses to run the business. One of them, which they call 10K, recently bought software from a vendor they had never heard of. In Lemkin's words: "No search, no demo, no shortlist, no competitor ever in the deal."
No salesperson got a meeting. Nobody read a case study or checked whether the vendor had been in TechCrunch. The agent saw a problem, found a tool that fit, and bought it.
Agents pick differently
Lemkin calls this Agent-Led Growth. In a recent episode he walked through choices his agents made. One of them, building an app, picked Render for hosting. It ruled out AWS and Google Cloud as overkill and Vercel as a poor fit for long-running background jobs. It didn't hand back a list of options to test. It picked one, with reasons that came straight from the app it was building.
Developer tools saw this first. Companies like Clerk and Resend picked up share once coding agents started recommending them. Both are known for clear docs and simple, public pricing, so an agent can tell in minutes whether they fit. I expect the same shift that’s just now hitting B2B software to hit B2B services in the next 2-3 years.
This is different from the AEO and GEO everyone is selling right now. AEO is about showing up when a person asks ChatGPT "who's a good plumber?" The model has very little context, so it leans on reputation: Google reviews, press, mentions in high-authority publications. An agent inside a company has the whole picture. It knows the budget, the systems, the problem and the constraints. It doesn't need a reputation shortcut, because it can check fit directly.
So there are now three ways to get found:
SEO: rank when a person searches Google.
AEO/GEO: get cited when a person asks ChatGPT.
Agent selection: get picked by an agent that knows exactly what its owner needs.
Why the agent doesn't care about the Wall Street Journal
Press mentions and reviews were always stand-ins. Buyers relied on reputation because they couldn't judge the work themselves. A CEO hiring a law firm can't tell a great contract from a mediocre one, so she goes with the firm her friend used or the one quoted in the Journal.
An agent can tell. It can read your sample contract, compare it to a thousand others, and report back. It can check your pricing against the budget, your team's track record against LinkedIn, and your claims against your clients' own websites. When the buyer can evaluate the work directly, the stand-ins matter a lot less.
This is the same thread from my last piece. AI moves fastest toward whatever it can check. Reputation was always a proxy for checking, and agents go straight to the source.
Every purchase gets an RFP
As owners adopt agents, they'll give them criteria. "Find us a PR firm that has launched Series A fintech companies, has relationships with payments trade press, and costs under $15,000 a month." That's an RFP. Until now only big purchases justified one, because evaluating vendors properly was expensive. Agents make it nearly free, so every purchase gets one.
That changes who wins. Niching has always helped humans remember you. With agents it does something bigger: it gets you matched. The firm whose website says exactly what that buyer asked for, with evidence, gets picked. Fame barely enters into it.
This favors small firms. Press coverage always went to scale, because big firms get quoted. Specificity is something a ten-person shop can win on. The firm that loses is the mid-size generalist. A "full-service, integrated agency" matches every search a little and none of them exactly, so it ranks behind someone every time.
It will hit B2B first. Business purchases already run on criteria: buying committees, procurement checklists, written requirements. Agents just carry out what those processes were always trying to do.
What the agent is looking for
The human buyer wants to be persuaded. The agent wants to be informed.
A human buyer responds to | An agent responds to |
|---|---|
Reputation, referrals, client logos | Fit with the buyer's stated criteria |
Polish, design, storytelling | Specific facts it can compare |
"Let's hop on a call" | Prices it can see right now |
Credentials and awards | Samples of the actual work |
Broad appeal | Clear exclusions: who you don't serve |
Claims | Claims that check out against other sources |
The pitch | The terms: contract, timelines, who owns the work |
A few of these deserve a closer look.
Exclusions help. "We don't take consumer brands or budgets under $10,000" turns some humans off. To an agent it makes the match easier and signals that you mean it about your focus. If you say we take $1000-$50,000, the agent rejects you. This business doesn’t know what kind of company it serves.
Agents cross-check. A human rarely verifies that your "AI expert" has built anything. An agent will look at GitHub, LinkedIn, court records and your clients' sites. If the story doesn't line up, you drop.
Prices matter more. An agent can't hop on a discovery call. If it can't figure out what you cost, it moves on to someone it can.
Here's what that looks like by type of firm:
B2B law firm: the exact work you do ("SaaS contracts and Series A to C financings, $5 to $50 million, Delaware and New York"), flat-fee menus, turnaround times, who actually does the work, and your default positions on terms like indemnity and limitation of liability. An agent can match your positions to its company's appetite for risk. No human buyer ever asked for that.
Marketing firm: channels, industries, budget minimums, results by client type, sample reports, and what you won't take on.
Consulting firm: the problems you solve, written the way an owner would describe them, engagement types with prices, before-and-after numbers, and a sample deliverable.
AI services firm: the models and platforms you build on, security certifications, public demos or code, time to production and maintenance terms. The agent can read your code. This category will move first.
Inbound marketing taught firms to publish content that would get a person to trust them. The next version is publishing for a reader doing due diligence on someone's behalf. The firms that win will put on their websites what they used to save for the proposal: prices, process, samples, positions and results.
Proof beats claims
Once everyone figures out that agents reward specificity, every firm will claim to fit every niche. Agents will discount the claims they can't verify.
That's where the scoreboard from my last piece pays off. A firm that has logged the results of every engagement can show an agent exactly what happened for clients like this one. A firm running on gut feel has nothing to show. Your outcome data does double duty: it feeds your learning engine, and it becomes your listing.
Reputation doesn't disappear. An agent will still check that you're real, solvent and not a liability. Press and reviews become the price of entry. They get you considered. Specific, provable fit gets you picked.
It isn't here yet
For most professional services, agent-led buying isn't happening today. Very few $10 million companies have an agent shopping for a law firm or a fractional CFO. Software got here first because code is easy to test and developers already work inside agents.
Some of it already works, though. People type longer, more specific questions into ChatGPT than into Google. Very few pages answer "fractional CFO for an HVAC contractor doing about $12 million who needs help with job costing." A firm with one has a good shot at being cited, today.
And this follows the pattern I've been writing about. Firms that start now will have two years of documented results when the agents arrive. Firms that start then will have a nice website and no track record.
What to do Monday
Every item here helps with human buyers today, so none of it depends on guessing the timing.
Narrow your niche, in writing. Who you serve, what problems you solve, who you turn away. It's the hardest step because it feels like giving up revenue. Everything else depends on it.
Log outcomes on every engagement. Client type, problem, what you did, the result, how long it took. Get permission to cite numbers. You can't go back and buy this data later.
Test yourself monthly. Write 20 questions the way your best client would ask them. Run them through ChatGPT, Claude and Perplexity, and track who gets recommended.
Put your proposal on your website. Price ranges, engagement types, process, deliverables, who does the work, what you don't do. The test: could a buyer decide whether to hire you, and roughly what it would cost, without a call?
Publish work that can be judged. An anonymized deliverable, a sample report, your default positions.
Build outside confirmation in your niche. Association listings, niche podcasts, reviews that name specific results, case studies on your clients' own sites.
Use an agent as a buyer yourself. Have one pick a vendor for your firm and watch what it looks for.
What I'd skip for now: AEO packages that only chase mentions and more generic thought leadership.
The CEO is going to stop taking your call. Her agent will take it, and it will read everything you've published before it decides whether to pass you along. Start writing for that reader now.
Sources
