Jason’s Full AMA from SaaStr AI London on What’s Really Changed in B2B Sales in the Age of AI … So Far
Top 5 Takeaways
- The plays all still work — webinars, inbound, outbound, demos. The hottest AI companies are being run by the same B2B sales leaders from 2017-2018. Vercel’s COO came from Stripe. Replit’s CRO came from ZoomInfo. If the playbook was broken, this wouldn’t be possible.
- AI SDRs didn’t actually work until Claude 4/GPT-4 — If you had a bad experience with any AI GTM tool before March 2025, write it off. Different LLMs, different world. Qualified, Gamma, Replit — they all took 5+ years to hit $1M, then exploded to $80-250M this year when the models got good enough.
- Provide insane value before you get a check — Marc Benioff told me he wishes every Salesforce customer could go live on AgentForce before they pay a dollar. That’s the new bar. Cursor, Gamma, Lovable — think about how much value you get in the free tier. That’s your new competitive standard.
- Your competitive moat is now measured in months, not years — I invested in a startup recently. Within two weeks they had four clones. A major company is building their own version launching this year. We used to have 5-7 years before big companies copied us. Now it’s 90 days.
- You cannot sell value if you’re not a product expert — We watched a solution architect close a seven-figure deal while the sales rep sat there useless. The buyer didn’t want to talk to the sales team. They had no value to add. AI is going to replace every AE who doesn’t know the product cold.
So we kicked off SaaStr AI London with a full AMA, and I want to share what we covered. Because there’s a lot of “woe is me” narrative out there right now that I think is actually dangerous.
“Is Inbound Dead?” No. But Your 2021 Playbook Is.
Someone asked me if inbound is dead because their traffic is down 50% in the last 12 months.
Here’s my honest take: this is part of a narrative that started around the end of 2023 and has accelerated. “Woe is me, SEO is harder. Woe is me, I don’t have as many leads as I had during a global pandemic.”
And it leads to this narrative that the go-to-market playbook is broken. That it doesn’t work anymore.
It’s just not true.
The playbook that some folks are running from 2021, of course, does not work today. But here’s what I say: the plays all work. Webinars, inbound, outbound, leftbound, rightbound — it all still works.
Look at the hottest AI companies right now. Vercel just raised at a $10 billion valuation. Who’s their COO? She was the Chief Business Officer at Stripe. Replit is doing 0 to $250 million already this year. Who’s their CRO? He came from ZoomInfo.
If you go through all of the hot AI companies in B2B right now, you’ll see a cast of characters from 2017 and 2018. You’ll see B2B leaders that we all know from classic SaaS running these AI companies.
That wouldn’t be possible if the plays don’t work.
The Real Difference: Everyone Is In-Market Simultaneously
The biggest difference with a lot of AI leaders is there’s just so much demand. Because what’s happening in a lot of AI tools is everyone is in-market simultaneously.
The tools are either so disruptive — like Cursor or Replit or Lovable or 11 Labs — that they go from almost nothing to $300 million this year.
At Bolt, which is one of the leaders in vibe coding, the head of sales was one of our old SaaStr sales team. The old is new. And at Bolt, at Lovable, at Replit — they have so much inbound they can’t service it. Brian at Bolt is at $60 million in revenue. I think he has three or four people on his sales team. How many of the thousands of leads can they follow up on?
So that is different. But they’re all classic B2B sales reps. Instead of calling every lead and trying to convince them their fungible product is the exact same as another product, they have insane demand and they’re servicing it.
But it’s the same playbook. It’s the same demos. It’s the same everything.
The Paradox: Record Budgets, But You’re Getting Cut
Here’s what’s really happening. I wrote this up with a bunch of Gartner data this year.
Gartner said overall enterprise software is going to grow the fastest it ever has — 15% a year at $400 billion in scale. It hasn’t grown this fast ever.
But of that 15%, almost half is taken up by price increases from existing vendors. Everyone’s raising prices. And about half of the remaining half is new AI budget.
So if you’re not one of the vendors getting price increases or you’re not an AI tool they want to add — everyone else has to get cut.
Vendor count is getting stable or shrinking to make room for new offerings and price increases from the select vendors and the new AI guys.
I got an investor update yesterday from a pretty successful portfolio company that’s mid-eight figures in revenue. They had $1.5 million in churn last month. From happy customers. No CSAT issues, no other issues. They literally said “we’re cutting apps next year” and “we’re an attachment to CRM and we got cut.”
Not because it didn’t work. Not because they didn’t have AI in it. But because two CIOs went around the room and that was the one that got cut. Great to have, but not mission critical.
Our SEO Is Down 8%. Our Traffic Is Up 50%.
Here’s what’s happening at SaaStr specifically.
Our blog is our core — it’s home base. Our blog traffic was fairly flat for about 4 years. Going into this year, our SEO is down 8%. Not 50%, but 8% for real.
But our traffic is up 50%. And going up.
We will end this year at saastr.com with twice as many readers as last year. Even though our SEO is down.
Why? Because people want to read about AI GTM content. They don’t want to read about classic SaaS themes about CS teams and CROs. Unless there’s an AI angle, they don’t want to read it. But because we have some of the best AI GTM agent content out there, people are just devouring it.
If we could sell that content, I’d buy islands.
Why AI SDRs Didn’t Work — Until This Year
Someone asked me about AI SDRs and why I said they didn’t work until recently.
Two different threads here.
First: None of these SDR products really worked before Claude 4 or GPT-4 this year. Lovable didn’t work. Base44 didn’t work. Nothing was very good until the magical moment when Claude 4 came out and everything became kind of magical.
Gamma was around since 2020. It was 5 years to $1 million and then 1 to $80 million this year. Replit was founded almost 10 years ago. It was 10 years to a million and then 1 to $250 million. Qualified has been trying to use AI to qualify inbound leads since like 2019. The CEO told me it really took off around February or March this year. “That’s when your product actually worked.” He said “Yeah, that’s when it finally actually worked after 5 years.”
So if you had a bad experience with almost any AI LLM GTM product before March of this year, write it off. Different time. Different LLMs. Different world.
Second: Even after Q1 of this year, so many folks just told someone on their team to deploy it or hired an agency that didn’t know what they’re doing.
Amelia and I were on a call with a global technology leader — a company we’ve all heard of that’s pretty AI forward. They said “We’re thinking about buying our first AI SDR. We’re just going to buy it and hand it to our SDRs and have them figure it out.”
It’s not going to work.
You got to train it. You got to train it for a month. You got to iterate every day. You’ve got to do onboarding.
The Captain Obvious Thing About AI SDRs
Here’s the captain obvious thing for any AI SDR, BDR, even AEs as we do more of them, CSMs:
First, you got to have it work in the real world with humans. Then you tell the AI what worked.
There’s no magical prompt. The prompt is: “Here’s the script that I used with a customer that I closed.”
Then you iterate that prompt and hook it up to data like Salesforce or HubSpot or Snowflake. Then you have it ingest the data to keep working with that prompt and make it better.
It’s not that complicated. But so many folks think “I’m going to buy a tool with no inbound leads at all. I’ve never done outbound. But I’m going to buy a tool, turn it on, and make millions.”
How would that work?
As long as you’ve closed any customers, you can train the AI to be almost as good as you and work 24/7. It’s that simple. But if you can’t sell it yourself, the AI can’t sell it for you.
The Skills You Need for 2026
Someone asked what skills marketers should uplift for 2026.
I talked to a great marketer at an AI-focused company. She was nervous. The world’s changing so much. She was worried about her job.
I told her: You should be. Everyone should be worried for their job. Every CRO, every CMO should be worried.
On the one hand, it’s the same people in the same roles. But there’s so much change. And you got to get ahead of it. If you don’t get ahead of it and build skills, it’s going to be hard.
My number one piece of advice: Become proficient at a couple of the leading tools. Right now, today.
If you are world class at AgentForce or Clay or Qualified or Artisan — really doesn’t matter which one. Pick two or three leaders that have decent revenue that people have heard of. Go through the deployment yourself. Do the training. Onboard. Get it working in your company. Point to the metrics.
If you have those skills, you will be infinitely hirable for the next 18 months because only about 2% of marketers have those skills.
Become the expert in deploying agents. It doesn’t need to mean you know every agent or even that you’re technical. But you need to pick a few leaders and deploy them yourself. Not tell someone on your team to do it. Do it yourself. Watch it. Iterate every day. Spend a week non-stop deploying it and then work on it every single day for a month.
You will learn so much and you will become infinitely deployable.
“Provide Insane Value Before You Get a Check”
I think back a lot to the first podcast Marc Benioff did with us earlier this year. I asked him what he thought of Palantir.
He said he’s jealous of how much they charge. Pretty funny — they charge more than he does.
But then he said something that really stuck with me. He said he loves what they’re doing with forward-deployed engineers. And he said: “I wish every single customer would be live on AgentForce or whatever agentic Salesforce before they gave me a dollar.”
Think about that.
The way we used to buy B2B software was: I need a CRM. I need marketing automation. Your friends would use it. You saw it at your last company. You talk to a sales rep. They could answer six questions. You’d buy it. And if you’re lucky, by the end of the year, you might deploy it.
In the old days, you might have customers that were renewing for year two that never even deployed in year one. That doesn’t fly today.
What can you do with AI to provide that much value to customers before they even pay you?
When you look at the ones that explode — the Cursors, the Gammas — think about how much value you get in the free tier. Think about how much value you get in the first month.
If you provide insane value with an agent before they pay you, they’re going to be kind of happy to pay you. Rather than the sucker bet that we all made for the better part of two decades: “Someone gave me a demo. The demo kind of works. I can’t use it myself, but I need that.”
Make your customers live before they pay. And the percent of your leads that convert is going to skyrocket.
Will SaaS Be Dead When We Have AGI?
Someone asked if SaaS will be dead in 2-3 years because you can just talk to OpenAI and say “write code that does that” and you don’t need to buy software.
Six months ago, I’d have said the odds of that are zero. That’s ChatGPT nonsense.
ChatGPT is very powerful, very extensible, very open. But it’s still impossible to reproduce the amount of workflows, the amount of corner cases, the amount of complexity that complex B2B software does.
But you can chip away at it. And you can chip away faster than ever.
I built 11 apps in 90 days that have been used 800,000 times. They’re all at saastr.ai/agents. I haven’t stolen revenue from anybody, but that’s a bit of an existential threat to somebody, isn’t it?
Since I built those apps on Replit, the agent has gotten like a hundred times better. When I started 170 days ago, the code would hallucinate. It would say things that were wrong. I couldn’t test it. I was never going to finish an app.
Now Replit has V3 with multiple agents that talk to each other and check the work. Before it deleted my database. Now it calls in other agents to help. When there’s a tough problem, it doesn’t go off the rails. It says “Hold on, I’m not sure. I’m going to call in another agent, the architect.” They bring in help on security. They have a design specialist.
Don’t underestimate the rate at which this stuff is getting better.
Your Competitive Edge Is Shrinking in Time
Here’s what I really worry about.
When I started in SaaS, you’d build something really slick and then you’d have about 12-18 months until you were copied by a startup. Then you’d have like 5 years until a big company copied you.
When we launched EchoSign, DocuSign only worked in Windows and was only partially web-based. We had like 18 months before they decided to copy what we did. Then it took 5-6 years for Adobe to decide to copy it. Google just launched a clone like last year — a decade later.
That’s the way it used to happen.
Recently I invested in a startup that I love. Incredibly powerful. Within two weeks they had like four clones. Then I talked with a massive company that’s building a clone themselves that’ll be out this year.
Our competitive edges are going to be measured in months when they used to be measured in years.
And that compounds. If you can’t run at that pace, you’re going to fall to 0% growth.
I wrote up Pagerduty recently. They just crossed $500 million in ARR. They’re worth $1 billion. Two times revenue. And their customer count has been flat at 15,000 customers for four years. DataDog came out with a clone. Elastic’s product is better. Startups are better. They just didn’t move fast enough.
And I think it’s only going to accelerate. Everyone will build clones faster. DataDog took 15 years to decide to build their competing product. It might take them eight months next time. And 90 days after that.
Agents Make Our Moats Even Weaker
Here’s the really scary part.
We’re running about five AI SDRs and BDRs at SaaStr now. Our most recent one, we just took a prompt from one tool and put it into another and it just worked.
If agents can take prompts from one app and give them to another app, my switching costs have gone down radically.
It’s not two years of business process change and mapping data and the leads don’t map to the contacts. If I can literally take the prompt — “reach out to my top 50 attendees from SaaStr London and give them a special coupon for SaaStr Annual in May” — and it works, and I can just give it to another agent that does the same thing… where’s my moat?
There is switching cost, but it’s so much lower than it used to be.
The cloning is going to be much faster. The agents create shallower moats. The agents can adjust each other’s prompts.
Everything is on shakier competitive foundations. We don’t get 5 years to wait for a big company to clone us. That was the old days. We got to adapt.
You Cannot Sell Value If You’re Not a Product Expert
Someone asked why it’s so hard for sales teams to do value-based selling and speak the language of CFOs.
Let’s be honest. Most sales reps have no idea how to sell value. You cannot sell value unless you’re a product expert.
Most sales folks want to talk with a war sheet, a tear sheet. They know six things. You can’t sell value if you’re not a product expert. AI makes it more extreme.
We were with an AI leader the other day. They closed a seven-figure deal while we were there. The solution architect left the sales team behind. Closed it without them. The buyer didn’t want to talk to the sales team. The sales team didn’t know the product. They had no value.
Value-based selling is providing value. You cannot provide value in the age of AI if you do not know the product cold — ideally how to deploy it and how to get it going and deliver real value.
Not having a valuation calculator on your website that says 18 months down the road the product will work.
I really think 70-80% of the sales executives I’ve worked with over the last five or six years don’t know their product cold. If you look back at your top couple sales reps, they weren’t just good schmoozers. They knew the product cold. Especially for startups. I call them sales magicians, but they’re not magical. They just know how every nook and cranny works.
We’re just not going to tolerate mediocre sales reps in the age of AI. We’re going to buy from somebody else.
The worst person you can hire today is a sales rep that tells you they’re a great “people person.” Who cares?
AI is not going to replace the AE the way it’s already beginning to replace the SDR. It’s already replaced support. It’s already replacing customer success. But it is going to replace a lot of AEs that don’t know the product cold.
Outbound Works — If You Solve a Top-3 Problem
Someone asked what’s changed and what’s stayed the same in successful outbound.
What we learned from our outbound AI is: geometrically more volume with the same results. Our results haven’t gone down in the age of AI by training the AI with the best scripts and ideas that humans did. We basically saw the same — in some cases a little worse, in some cases a little better — but the same as humans for the most part, with much higher productivity.
Outbound isn’t dead.
I asked Brian Halligan, chairman of HubSpot. I asked Yamini Rangan at SaaStr Annual this year: “Do you read your cold outbound?” She said: “I read them every day. Email is the best thing in the world. Everyone reads it. It’s an open medium.”
Your job is to solve one of the top three problems of your buyer — whether it’s CEO, SVP, VP, wherever.
If it’s a top-three problem, you’re going to get a meeting. If it’s a top-three initiative or a top-three pain point. If you’re top three, you’re going to see super high open rates if you have a differentiated product and crystal clear value proposition.
This has not changed in the age of AI.
And I don’t know why people thought crummy emails for the 10,000th cadence product worked. Maybe they’re sucker bets. But your job is to be in the top-three initiatives or top-three needs of your buyer.
How Big Company Budgets Actually Work
Let me tell you how it works in big companies, because if you’ve never been a buyer at a big company, you don’t really get outbound.
When I was a VP at Adobe, there was unbudgeted, discretionary, and budgeted money.
Getting budgeted money was big dollars, but really hard. Multi-million-dollar contracts that would take one, three, five years of discussing and planning until you could close that deal.
But even as the most junior VP, I had a $500K slush budget. Today it might be a million bucks. And everyone had more than me.
What was that slush budget for? My top-three needs that I couldn’t get elsewhere in the organization or get budgeted through the CFO.
If you solved one of my three needs and it was some fraction of that $500K to a million, I had budget. I didn’t even care. It’s use it or lose it in big companies.
But I didn’t have one more dollar than that. And I didn’t care about my eighth problem. I just didn’t care.
Selfishly, if you solved one of my problems, I had half a million to a million bucks to buy. Those are the budgets you’re fighting for. They haven’t changed in the age of AI, but our priorities have changed.
So many folks have their own problems in their own org, but they’re also under pressure to bring innovation into the enterprise. That’s the AI budget. You got to be in these top three.
Top 5 Mistakes Folks Make
- Buying an AI SDR tool and just “turning it on” — You can’t buy a tool, hand it to your SDRs, and have them figure it out. You have to train it for a month, iterate every day, and read every email it sends. If you can’t sell it yourself, the AI can’t sell it for you.
- Writing off AI tools based on pre-March 2025 experiences — If you tried Qualified, Gamma, Lovable, or any AI GTM tool before Claude 4/GPT-4 came out, that experience is irrelevant. Different LLMs, different world. The products finally work now.
- Running the 2021 playbook and wondering why it’s not working — Your SEO is down. Your leads are down from pandemic highs. But the plays still work — you just need to find your tailwind. AI GTM content, AI tools, AI agents. Lean into what customers actually want to buy.
- Hiring “people person” sales reps who don’t know the product — The worst hire today is someone who brags about their relationships but can’t explain every nook and cranny of your product. Solution architects are closing seven-figure deals while AEs sit there useless. Know the product cold or get replaced.
- Thinking you have years before competitors clone you — You used to have 5-7 years before big companies copied you. Now you have weeks. I invested in a startup that had four clones within two weeks. Google launched a Replit clone 5 days ago. Your moat is measured in months now. Move faster or fall to 0% growth.
And see everyone at SaaStr AI 2026 in SF Bay, May 12-14!!
