Jason M. Lemkin's Recent LinkedIn Posts

Our SaaStr AI startup valuation calculator has been used 919,000+ times. It's probably a 2 out of 10 technically. But in terms of utility? It's a 10 — because nobody else offers it free. You don't need the most elegant solution. You need the most useful one in the market that actually exists.

What we need in 2026: AI AEs and AI VPs of Marketing AI SDR products are now pretty darn good. For us, from Monaco to Artisan to Qualified to Agentforce, they are pretty darn good. And every month in 2026 they will get better and better. What doesn’t work yet: - We don’t have an AI Agent that can close most deals on its own, at least as an option for prospects. Why not? It’s just harder than sending texts and emails. A lot more variables, more training, harder. But very do-able. Looking forward to 2026 here. - We don’t have a strong third party AI VP of Marketing that can run 90%+ of your marketing activities for you. Autonomously. The AI Marketing tools here in general do a small piece. An AI landing page. An AI GEO. Maybe some ads — but never all of them, not really, and never truly autonomously. None can even seem to do highly dynamic, segmented marketing campaigns autonomously for you. Not yet. We need AI VPs of Marketing that can just do 80% of what you need, out of the box, autonomously. It’s very doable. It’s just a lot more work than an AI landing page. Looking forward to AI GTM in 2026!

Not every sponsor should man a booth for three days. If your goal is relationships — not badge scans — the lounge is the better play. The best conversations at SaaStr AI 2026 happen between sessions, not inside the booth footprint. We tripled 1-on-1 meeting space this year because that was the #1 complaint from sponsors. They wanted to close deals, not collect business cards.

There's a moment in vibe coding when you go from "hoping it works" to knowing whether a project will work before you write the first prompt. 12 apps and 1M uses in, now we know. It's not about prompting better. It's about understanding the system's outputs and limits before you start.

"I built an AI agent to talk to our AI VP Marketing. An AI Chat product. And it works great. Probably better than 98% of the chatbot or "AI assistant" products I could go buy from a vendor right now. It crushes the answers. It works as well as any third party chat I've ever used. It cost $0.20 and took less than a minute, custom to us. It just works. And ... I didn't have to hop on a call with a sales rep. I didn't have to sit through a demo. I didn't have to negotiate a contract, wait for onboarding, or file a support ticket when the integration broke."

"If you have a functioning, high-performing human sales team that’s growing, don’t blow it up. Layer the agents in carefully. Let them handle what humans don’t want to do — the 24/7 coverage, the lead reactivation, the instant response."

Buy 90% of your AI agent stack. Build 10% — only what's genuinely proprietary and can't be bought. Folks getting this backward (e.g. by over vibe-coding) are spending 6 months building infrastructure any vendor would sell them for $500/month. Your edge is in the training data and the integration, not the plumbing.

Our AI agent stack generated $5.8M in pipeline and a 2x improvement in win rate in last 6 months. The pipeline number is easy to measure. The win rate improvement is actually more important — it means the agents aren't just generating more opportunities, they're generating better-qualified ones.

When B2B growth falls to < 20% YoY, you are in the Danger Zone. It’s likely fueled growth mainly coming from retaining existing customers, raising prices, and selling your base more and more products. You aren’t acquiring enough net new customers. When B2B growth falls to < 10% YoY, you are in the Dead Zone. You aren’t dead yet, but if you don’t make radical change, your business is likely terminal. You have truly fallen out of product market fit. You need to re-found the company. For real.

AI PR is terrible for your start-up. Let me explain why. For the last 12-15 months or so, SaaStr and I am sure everyone at our scale or bigger is innundated with PR and podcast and media requests from PR firms and internal comms. They all are using AI PR tools. The pitches are always decent, well written. Never great. But never terrible. In fact, they are usually just good enough to get opened. But they are so unhelpful to us, it’s also clear they are wasting my time. So now I block them. I never used to block mediocre pitches from humans. Instead, I’d tell them not to pitch anything similar again, but I would tell them what we did want in terms of A-tier speakers, etc. But no point in doing that with an AI. And we also get 100x the pitches we used to. So I just block them. That means when you do have a great pitch finally … I’ll never hear it. It’s lazy PR. It’s not what you want.

I have a fresh $100M+ to deploy in SaaStr Fund. I also run an eight-figure media business. I also vibe code 1.5-2 hours a day. That's my whole budget. And with that 1.5-2 hours, we've shipped 12 production apps used 1,000,000+ times. The constraint isn't skill anymore. It's time allocation.

Your very, very best people are even better than you think they Dig in and you’ll see they are doing even more great things than you realize * Let them run * Help them hire * Let them make the mistake * Wait to top them * Thank them * Pay them

Lately humans are much ruder to our AI Agents. Most aren't of course but the number of abusers has gone up. The one that really stands out is that folks endlessly “yell” at one of our AI Agents for free tickets to SaaStr AI Annual 2026 (May 12-14). They ask it to overwrite its code, or create a nonexistant 100% off coupon, etc. When the AI Agent won’t, they get extremely abusive. Not occasionally like a few months ago, but now we see this constantly. Maybe because AI Agents are still fragile, and it works often enough as prompt injection, etc. Reason enough to buy AI Agents from truly hardened providers.

There’s a “doomsday” scenario for a lot of B2B apps that I think folks are subtly missing. The threat isn’t that we vibe code away the need for complex apps. The threat isn’t less seats over time, though that matters. Or even that new AI Agents do a better job than you. That’s big risk, too though. But the Doomsday Scenario is already happening and it’s this — we’ll do everything inside of Claude that we can. It’s cheaper, simpler, faster, slicker, and better integrated. And the more and more Claude can just take over a function — and it will take more and move over this month and this year — the more your app might quietly fall away. If not entirely, in part. You have to fight this aggressively. The “Fortnight Effect”. Claude is coming for almost everything in B2B. Sometimes just a little tiny bit. Sometimes … a lot.

A few months ago, someone reached out to me for help getting into investing in venture capital. He had experience at the edges, and helping at a VC firm, but almost no investing track record. I took the Zoom as a favor. But after about 20 minutes in, I realized I could help him. I knew a potential connection. And then he ... ended the Zoom first. He said he "had an important call to take." I guess with someone better. But he never did find that job in venture. The simple lesson: give your calls and Zooms time to breathe. Quietly schedule an extra 10-15 minutes in case any good call goes long. Because if you don't, you may run out of time to get to ... Yes.

It is no longer enough to being “using AI”. Hooray. You must be shipping with AI or your job is at risk This applies to every role

We made a mistake that we tell everyone not to make with AI agents. We stopped training and monitoring one of our original AI Agents. We now have almost 30 AI Agents. There's so much to do every day now, to keep up with them. y So we stopped checking in with all of them, every day. The less mission critical ones, we stopped doing ... 1-on-1's with, in essence. And then we found out yesterday, one hadn't been sync'ing with new data ... in over 4 months. Yes, it still functioned, but on increasing stale data. And it never threw an error, despite being a very well funded startup with a top dev team behind it. AI Agents are many things. But they aren't set-and-forget. Not yet.

"You do not need to be technical anymore to win with AI agents in Q2 of '26. You do not need to be even 1% technical. You just need to do the real work. This is your time. Or you’re going to get laid off because you’re not going to matter.” with Harry Stebbings Rory O'Driscoll + me

If you're just starting with AI agents: one for outbound, one for inbound qualification, one for customer support. Connect them through your CRM. That's it. Don't orchestrate 20 agents in month one. The compounding effect only kicks in after each agent is actually working well. That will take many months.

Does it really make sense the SaaS markets have crashed so far in 2026? AI was around before then. Can Claude repackaging itself to be easier to use for law, finance, banking, code reviews etc. in ‘26 really justify a -25% sell off of essentially every public software company? Probably not. But … But … it’s become clear all the economic value that’s going to be created in B2B is going to be created at the AI Agent level. And almost no public B2B onmpanies have proven they can ship an agent that will drive an increase in revenue. That’s the definition of success. Palantir has proven it. Salesforce has a good early start. But overall, where is the acceleration? Where is the revenue growth? There has been plenty of time. The models were good enough at least for all of 2H’25 to ship great products. The models have been strong enough to end any excuses since November/December 2025. The worry that many public leaders won’t be the leaders of the agentic era in their space, for the moment … seems well founded. Check-the-box features and half-hearted beta’s don’t count. Time will tell.

We were paying an agency $200-300K/year to review 4,000+ speaker applications. They burned out and stopped doing the work. I built a replacement on Replit in one day. It uses Claude to score every submission 0-100 across 5 criteria, tells applicants their odds, and runs on autopilot. It's much, much better than the agency ever did. The best automation is the one you build because you have no other choice.

Over the weekend, we hit our revenue goal for Q1'26: 140% of 2025. But this year it was done with ... 1.25 humans in sales, and 20+ AI Agents. A bit of the future, already here today. Was it all better? Not entirely. But the bottom line is we were able to touch and schedule qualified meetings with 2x the folks with 5 core AI GTM agents vs 12 months ago with 4+ humans in sales and no AI GTM Agents. The AI Agents are not perfect. Many of the emails and texts are good, but not really epic. But: - They do work 24x7 - And they don’t mind chasing all the leads. Not just the hot ones. - And they instantly respond with honest answers to any question - And they instantly take a meeting - And they are pretty good at reactivating stale leads, or ones that humans decided to … never really work. Not really. It’s not all better. We’d love to hire another cracked human sales exec that loves to work with AI SDRs and AI Agents. But still, our 1.25 humans + 20 AI Agents did outperform our all-human sales team last year. By +40%.

So proud to be partnered with Salesforce and #agentforce on the agentic journey! And come learn all about it at SaaStr AI Annual, May 12-14 in SF Bay!!

"The core CSM problem has always been the tension between personalization and coverage. You want every customer to feel like they’re your only customer. But you have 200 customers and a team of 4 CSMs. Something has to give, and usually it’s personalization. AI agents break that tension."

"And here’s the brutal truth: incremental change will not fix this. Hiring a new VP of Sales won’t fix this. A new demand gen strategy won’t fix this. A pricing refresh won’t fix this. You need to re-found the company. Not reorganize it. Not optimize it. Re-found it."