SaaStr Annual Conference 2026 - The Sound of Inevitability
What I Built
I've been an Evernote user since the late-2009 era. I was doing a bunch of ecommerce consulting and Evernote was the first app that felt like it was made for taking notes. Hundreds, maybe thousands of notes later, I've still got one problem with it: it's terrible for conferences. I take notes. I take pictures of slides with my phone. I'm not one of those hold-your-iPad-up-to-take-pictures kind of guys. Guy in the red shirt in front of me, you know who you are. Then I transcribe some of the slide content into my notes. Then I give up and email or LinkedIn the presenter begging for the original deck.
For those who don't know, SaaStr is an annual conference where SaaS companies, vendors, and operators gather in San Mateo for three days of presentations, exhibits, and networking. This is my second year. I attended as many of the large-stage sessions as I could, with Claude.ai and Gemini open the whole time, trying to figure out a way to task the LLMs with helping me take better notes. There's no perfect app. When I got home I started collating my notes with the dozens of slide photos I'd taken. Then I pulled up the livecast on YouTube because some of my photos were shit.
I figured there had to be a better way. I started experimenting with yt-dlp to grab the videos. That turned out to be imperfect. Then I hit the limit of my Python skills (which didn't take long) and phoned-a-friend to Claude Code. What we ended up with is pretty neat. I say "we" generously. After my initial efforts, Anthropic did most of the work.
The whole thing takes ~10–15 minutes per talk and costs ~$0.35 in Claude API calls. Everything caches to disk, so re-runs only redo what changed. I put the code on Github if you're interested.
For now Evernote survives. It did its job. My iPhone did its job. Evernote lives to fight another day. Until I build PeterNote. To be continued.
What I Thought
Last year my company had just launched its first "AI-enabled" product feature. I was watching every session through the eyes of a product manager. Where could we shove the LLM next? Should we monetize AI or include it in the standard feature set as a competitive differentiator? This year was different. Different because of me. This year I was looking through the lens of a business owner, a founder, an operator, an entrepreneur. I was looking at what AI is going to do to my current company. My future companies. My future...period.
1. Pre-2022 SaaS is cooked unless it keelhauls itself.
Rory O'Driscoll (Scale Venture Partners) drew the only cleavage on stage that mattered: pre-2022 vs. post-2022. Jason Lemkin (SaaStr) ran the math. SaaStr's own sponsor floor flipped 21% to 81% AI-native in one show cycle. Five of six watched B2B names are reaccelerating only by grabbing AI budget. Palantir went from 27% to 80% growth on 11 consecutive quarters. Lemkin on the "modest AI features can't fully monetize, hoping for a recovery" cohort: that recovery is not coming. Adam Guild (Owner) gave the only survival template in the document. Triple-triple-double metrics on the surface, existential AI threat underneath, rebuilt over the objections of every investor and customer in the room. Stephanie Cohen (Cloudflare) ran the macro version: three prior platform shifts (PC, mobile, social) left content economics intact. This one didn't. The "AI-first feature" your exec team is staking the company on is cope.
2. Your AI-native competitor has four employees. In two years, you'll have 25 of them, each taking 3%.
Jeanne DeWitt Grosser (Vercel): 10 SDRs to 1, pipeline up. Eleanor Dorfman (Anthropic): a 3-person GTM productivity team built the entire stack she demoed on stage. Owner's 7-person website team beats Squarespace's and Wix's ~1,000-person equivalents on outcomes that matter to restaurants. Stripe receipts: Lovable $0 to $400M ARR in 16 months. Cursor $1B to $2B in three. And it's not just one of them coming for you. AI dropped the barrier to entry so completely that you don't face one 4-person competitor. You face 25 of them inside 24 months, each chewing 3% of your market. Death by a thousand AI-natives, none of which show up as the named threat in your board deck. "We'll increase output instead of layoffs" is the public-company euphemism for not having the guts to rebuild. Lemkin said it cleanest: "we're going to give up on reskilling by end of year… layoffs themselves accomplish nothing." The 2026 layoff wave is about clearing deck for agent-native teams, not cost cuts. Your 50 people aren't 50× more productive than any of their 4. You're a different species fighting 25 category fights at once, losing each one on cost structure.
3. Distribution is the moat. Selling beats product.
The only companies growing logarithmically own a distribution surface, not a better widget. Owner's Grader (free AI wedge, ~$1/customer in compute) now starts 83% of new customer journeys. Was 0% two years ago. Anthropic, Cursor, Lovable, and Replit ran parallel PLG and SLG from day one. Maia Josebachvili (Stripe) called it "decades of enterprise revenue compressed into two years." Guild's free-wedge to $500/mo bundle isn't a pricing chart. It's a distribution architecture with the CAC profile of freemium and the LTV of enterprise. O'Driscoll's six "moat types" are all distribution at the root: sensors, marketplaces, FDEs, full-stack, network effects + data, data flywheel. The companies that look like product wins are mostly distribution wins with product attached.
4. Your next buyer is a machine and the canon doesn't apply to it.
Stripe: agent traffic to docs 10× in 2025, on track to pass humans by Q4. Cloudflare: more than 50% of HTML requests already non-human, two-thirds by year-end. SaaStr left Marketo for Resend on agent-friendliness alone. Josebachvili: decades of human-psychology pricing ($9.99, good/better/best, all of it) is irrelevant to agents. Per-interaction billing, which humans rejected for 25 years, suddenly works. There's a public API Agent Report Card grading vendors live. If your API can't be discovered, evaluated, and purchased by an LLM with no human in the loop, you have a 12-month problem you can't see on your current dashboards. This is distribution channel #1. Most pre-2022 SaaS hasn't even audited for it.
5. If your customers live in chat, build the product there.
Dorfman's whole pattern at Anthropic: "Slack in, ticket out, Claude triages." Vercel's Deal1 runs inside a Slack thread. No app, no screen. Denise Persson (Snowflake): "The dashboard is dead." Guild: "Every time somebody logs into our website builder to manually intervene, it means we have failed." DAU/WAU as a health signal is inverted. The real number is outcomes completed without the user opening anything. Every greenfield product decision now starts with: which of these screens does the user actually need, and which is vanity-branded context-switching? For most workflow tools the honest answer is a Slack bot, a Teams bot, or an MCP server. Not a React app. The UI is a tax the user pays to use your product. Stop charging it.
What it Changes
Agent Smith has Neo pinned to the subway tracks. The train is coming. Smith leans into his ear: Do you hear that, Mr. Anderson? That is the sound of inevitability. That is the sound of your death.
Rory and Jason kept hammering the same divide on stage: companies established before 2022 have a tough row to hoe. Adam from Owner hit on it personally too.
Several years ago my family spent a year traveling the country in an RV. We wanted to see a few spots before deciding where to settle after leaving Colorado. At one campground I got talking with our neighbor, who had just come down off his roof, hands streaked with white Dicor — the self-leveling sealant every RV owner learns to keep in the toolbox. He told me there are only three types of RV roofs: those that have leaked, those that are leaking, and those that will leak.
There are only three types of companies right now. Agent Smith is going to visit all of them.
Roofs that have already leaked are companies built AI-native after 2022, or that legitimately pivoted there. LLMs and agentic tooling are wired into every layer: product, sales, finance, marketing, dev. Every employee is an orchestra conductor. Using AI isn't a strategy. It's the operating system. They run lean, ship in days, and operate on the assumption that if they don't keep customers happy, someone else will at a lower price.
These companies will survive and prosper. They won't dominate the way Salesforce did. There's too much competition for any single one of them to own a category. But they're inevitable. They're the train.
Roofs that are leaking are companies that have, at face value, put together what they call an "AI Strategy." They added a chatbot to a pricing page. Their devs use Replit. Their marketing team uses Synthesia for videos. An LLM proofreads their email drafts. They added a few LLM trinkets to a few of their product functions. Maybe even a chat support feature. There's a slide in the board deck about AI productivity gains. The business model didn't change. The unit economics didn't change. The go-to-market didn't change. But there's a section in the all-hands now called "Our AI Strategy."
Here's the test.
Look in the mirror. You probably see Christie Brinkley. That's fine. Everybody sees Christie Brinkley. The question isn't what you see. The question is whether you're getting asked on a thousand dates. The dates are the receipts. Did your sales cycle shorten? Did your CAC drop? Did your gross margin expand? Did you start booking deals you couldn't book last year, against competitors you used to lose to? Did anything in your unit economics actually move? Did your growth rate accelerate? Meaningfully?
If yes, you pivoted. Welcome to camp one.
If no, you didn't. You bought AI tools. Your eyes are lying to you. And the four-person AI-native competitor you can't see on your radar right now is currently building the product that's going to take your customer base apart inside 24 months. With $1/customer in compute, a CAC profile you can't match, no politics over whether they should let go of their D-players "because they've been with us for so long," and zero cost structure to defend.
Most companies who think they've pivoted are in this camp. They will not be around in a decade. Or they'll survive but fail to deliver any meaningful return to shareholders, which from an investor's seat is the same thing.
Roofs that will leak are companies who don't think the first camp exists. Or if they do, they're not afraid of it. They'll tell you AI is a passing phase. They'll tell you it doesn't apply to their business. Their moat will protect them.
For a vanishingly small number of them, that's true. AI may not come for them.
For the vast majority: you're cooked. You just don't know it yet. You haven't even looked in the mirror.
Smith is already in your ear. You just don't recognize the voice.
The receipts on the broader picture: roughly a trillion dollars deployed into pre-2022 SaaS at 8–12x ARR multiples. Those same assets now trade at 4x. That's before repricing them against the AI-native alternatives competing for the same customers. In the public market early this year, SaaS shed nearly $1.6 trillion as the agent-software thesis hit the comps. PitchBook has nearly 13,000 companies across all sectors sitting in PE portfolios with average hold duration north of seven years. A lot of capital, locked in a lot of companies that can't be sold at what was paid.
And don't even mention Medallia.
A month before I sat down to write this, Thoma Bravo — the most respected software PE shop on earth — handed Medallia to its creditors. Five billion in equity, gone. Not impaired. Not marked down. Gone. The lenders took the keys. Medallia was customer experience software. Literally the same kind of company I described in takeaway #5. Thoma Bravo bought it in 2021 for $6.4B at 9x revenue. By early 2026 it was throwing off $200M of EBITDA against $300M of annual interest. The numbers stopped working before the AI excuse could even form.
Did AI kill Medallia? No. Execution killed it. Lack of growth killed it. But AI set the stage by making growth structurally harder. The debt did the rest. In Rory's words from SaaStr: it's not the public markets being a bitch. It's the public markets demanding more from these companies.
Medallia is the first major casualty of this cycle. It will not be the last.
On the other end of the spectrum there's Monaco.
Sam Blond came out of stealth with Monaco on February 11, 2026. AI-native sales platform. Founders Fund led the $25M Series A. From zero revenue in February, Monaco added seven figures of ARR every month — March, April, accelerating. They opened day one of SaaStr Annual on May 12 by announcing a $50M Series B led by Benchmark. Three months between A and B. Day one of the conference, Lemkin posted that afternoon: SaaStr Ai. Day One. 6-figure deal booked. No big deal. Just Monaco.
That's what a roof that's already leaked looks like.
So go look in the mirror.
The mirror will lie to you. Everybody sees Brinkley. Check the receipts. They'll tell you whether anyone's actually asking you to dance.
If they are, you're the next Monaco.
If they're not, you're the next Medallia.
Either way, Agent Smith is going to visit every company. The only question is whether The One knows what to do about it.