Canva增速腰斩：AI时代创业者的10条判断

My cut on what Canva’s drop to 20% growth means if you’re running a pre-AI product, hiring against Anthropic, or holding a mark you set in 2021.

Canva cutting 2026 growth from 30% to 20% was the headline. The part I keep thinking about is smaller than the number and worse. We churned Canva at SaaStr. We churned Notion too. Neither did anything wrong, and neither of them lost a bake-off. We just stopped having a need for them, and then our agents built an entire ad creative operation without either ever coming up.

Ten things I’m taking from this week.

1. Run the agent test on your own product. Will your agents … pick it?

We built our own ad server and creative generation network at SaaStr. It runs on our agents. Not once did the agent think to use Canva for any of it, and Canva is a product I paid for and liked. If I were running any prosumer or self-serve product right now, I’d give an agent the job my product does, with no instruction to use me, and watch what it reaches for.

My learning → You can win back a customer who churned to a competitor. There’s nothing to win back when you were never in the running.

2. If your product existed to help people skip a specialist, that job is closing

Airtable was a no-code database dressed as a spreadsheet. Notion is one dressed as a word processor. Canva was a no-code way to design, and it was a genuinely disruptive product because I no longer needed a designer or any HTML. The entire category of “what humans can do without engineering resources” is winding down, because AI does that natively now.

My learning → Being the workaround for a missing skill was a great 15-year business, and it’s shrinking now.

3. The enterprise version of this is coming, just slower

The Gartner data suggests fewer than 10% of enterprises have successfully deployed an agentic application. I believe that, even standing at a big Salesforce event surrounded by the opposite message. The hot enterprise AI companies are still serving early adopters and outliers. On the prosumer side there’s no lag at all, because everyone is already ChatGPT-fluent.

My learning → Selling to the enterprise buys you a few quarters of cover, not a few years.

4. You can’t argue with a meme once it takes hold, you can only outgrow it

The dumbest AI meme of the past two years is that everyone will vibe code their own CRM. It makes no sense for 99.9% of the world. You can’t maintain it, you can’t build the integrations, and most people saying it have never used a CRM. It took hold anyway, the shorts jumped on it, and now it’s priced in. If I were Canva I’d be worried about the same thing happening with “you can make a poster in ChatGPT for free.”

My learning → Get in front of the story early. Once it’s a meme, only revenue growth moves it.

5. After Airtable and Canva, I’m marking valuations down harder

These are two events that quietly hit old marks and are very hard to hide behind. Honestly, most people should have marked both down last year. “We raised at $42B so we’re carrying $42B” is over. Start from what the company is doing and how fast it’s growing, and price that.

My learning → If you can’t defend a mark with a current growth rate, you’re carrying a memory, not a valuation.

6. On selling early, my own history says 50/50, and the fund math says hold

I ran the analysis across everything I’ve been involved with, angel and venture, on where I should have sold and where I should have held. It broke roughly 50/50, which surprised me. It still doesn’t change what I do, because I have three fund returners and my LPs are underwriting a 5x, 6x or 8x fund. Trim those and the math stops working. The blended exit argument sounds great on X, and it works far better at scale than it does for a smaller fund.

My learning → Trimming protects your average outcome and kills your outlier. Know which one your LPs are paying for.

7. Being invisible inside a great company is a real reason to leave

I was an SVP at Adobe running the number three business unit. We were doing $800M and it may as well have been zero. You sit with 50 other VPs and never get to talk about what you’re building. I’d read Jeff Dean leaving Google after 27 years the same way. If your work sits permanently third in line for compute behind Cloud and Gemini, and you can take your whole team and raise a billion dollars to do exactly what you want, you go.

My learning → Your best people rarely leave over money. They leave over being third on the agenda.

8. You need an S-tier for comp in the Age of AI, and it will be uncomfortable

There used to be two comp bands: normal people and AI people. Now there’s a third. The CEOs I work with informally at $100M to $200M+ ARR have all built one: four or so people on seven-figure packages with equity ten times what anyone else at that stage gets, and those people are the core of the next-generation product. At $200M in revenue that doesn’t break your model. It’s rough on everyone else, and it’s still the right call. A skunkworks with an S-tier is how the Palantir and Intercom style turnarounds happen.

My learning → The Anthropic comp signal isn’t going away. Build the tier on purpose or lose those people quietly.

9. The hiring test I’d use now

The old test was “did you get an offer from Google,” which told you someone was top 5%. The new one is “did you get an offer from Anthropic or OpenAI, and what was the job?” A lot of those seven-figure roles are watermarking, or getting the pulse animation right for 18 months. Interview enough people and you’ll find the ones who’d genuinely rather build LLMs for accounting.

My learning → You won’t outbid the labs. You can out-interest them for a certain kind of person.

10. Founder-run is close to the only thing I underwrite now

Most of the investments I’ve made that aren’t run by a founder are going to be zeros in this era, at $20M ARR or at $200M. If the price of avoiding that is a founder taking a much larger piece of the company, I’ll pay it without hesitating. Being a public company CEO today is miserable, and I say that having founded twice: no control, a board that doesn’t understand the product, plenty of money already made. I’d think about leaving the keys on the table too. Founder control has become a fair price for keeping people in the game.

My learning → In a platform shift you’re underwriting willingness to rebuild the company from scratch. Almost nobody who didn’t found it has that.

"Any investment I've made that is not run by a founder is going to be a zero in this age.

If the price of me not having a zero is getting Nick to 40%, I wish I was a shareholder. That's the price I'm gonna pay in a heartbeat.

I believe every non-founder has run my portfolio… https://t.co/GxPZG0obzS pic.twitter.com/O0EhAzISBt

— Harry Stebbings (@HarryStebbings) August 13, 2026

What I’m Doing Differently This Week

I’m running the agent test on everything we pay for at SaaStr AI and on the products at companies where I sit on the board. Most of them are good products, so that isn’t the question. The question is whether an agent would ever choose one. That already sorted Canva and Notion out of our stack before we consciously decided anything.

Jason’s Takes is the SaaStr AI companion to our weekly 20VC x SaaStr recap with Harry Stebbings and Rory O’Driscoll.