HubSpot is one of the greatest classic SaaS and B2B companies of all time. The majority of the SaaStr AI community runs on HubSpot. We are big fans.
But … I just tried HubSpot’s new AEO (AI Engine Optimization) tool announced as part of its big new AI push. And this new tool? Yes sure it technically works, to some extent. But not as well as agentic competitors. That’s just not good enough in 2026. It’s another example of a pattern across many B2B folks at scale: shipping AI products and features that are maybe 60% as good as the best AI point solutions in the category.
And nobody is going to pay for a 60% AI solution in almost any category today.
And so if you ship 60% AI and agentic products to bolt on to what you have … folks may use it if it’s free and included … but you aren’t going to get back to growth.
What I Actually Got From The HubSpot “AEO” Tool
Let me walk you through the experience:
- A “70% Brand Visibility” score. OK, great. But what do I do with that number? Where are the action items?
- A “0% Sentiment Analysis” for our content and “64.29% for our events.” You sure? SaaStr has published 7,000+ posts over 10+ years. Zero percent sentiment score? That’s not a data insight. That’s a broken feature.
- Recommendations: None. Literally none. “No recommendations just yet.” Then why am I here? And why would I put in my credit card?
- Citations: Stuff I already know.
So I got a dashboard with some numbers that don’t obviously map to any action I should take, no recommendations, and citations that tell me things I’m already aware of. Then it instantly asked me to pay $50 / month for more prompts.
Which is more than almost all our somewhat similar agentic products charge.
If it were free, fine I guess. But it’s not enough as a standalone agentic product. The competition is just too good, too fast and too furious in the AI Age.
One of the execs said I did not understand the tool. Perhaps. But a 0% score for SaaStr? And no recommendations at all? Not gonna cut it for an AI-grade product in 2026.
This is Just One New Product That Will Improve. But The 60% Problem Is Everywhere Now
This isn’t just a HubSpot issue with one relatively minor product it just shipped, I’m just using it as one example from this week. It’s an industry-wide pattern. Hundreds of B2B vendors are racing to ship AI features that check the “we have AI” box but don’t actually solve the problem better than the dedicated tools already in market.
Six to nine months ago, maybe you could have gotten away with this. The bar was lower. Buyers were more forgiving because everyone was still figuring out what “AI features” meant. If you shipped something that sort of worked, you could get credit for being early. And the LLMs weren’t quite as good until late 2025.
That window is closed.
The AI native point solutions are in many cases great. Replit and Lovable for building, Reve for images, Higgsfield and Opus for short video, Gamma for presentations. The Big Guys can’t just ship solutions that are 60% as good. If they do, folks may use them. But they won’t really pay for them.
The dedicated AI-native tools in every category have had another 6-9 months of compounding improvement. They’ve ingested more data, shipped more iterations, and built deeper integrations. The gap between “60% good enough” and “actually best-in-class” hasn’t narrowed. It’s widened.
So I Built My Own. In 60 Minutes.
I do not recommend you vibe / build anything you can buy. I don’t. But as an experiment, I built my own AEO tool in Replit. It took 5 minutes for a quick clone of the basic concept, 60 minutes to get it to truly work, and about 2 intermittent hours to make it slick.
Try it here. It’s free. And in fact, I used its recommendations to get the AEO score for SaaStr.ai website up from a sad F to a C+ in a few minutes.
Is it better than HubSpot’s tool? Honestly … it might be. At least right now.
Here’s the #1 difference. My vibe-coded version actually tells you what to do to improve your AEO score. HubSpot’s tool … didn’t. It just said “no recommednations.” The SaaStr AEO tool I quickly built gives you a score (saastr.com: 67/100, Grade C+), then generates ready-to-use prompts you can drop into Replit, Lovable, Cursor, or any AI coding tool to actually fix the problems, as well as options to drop into WordPress, Shopify and other platforms. Category-specific breakdowns with real issues and real fixes. Structured Data scoring 20/100? Here’s exactly what’s wrong: no JSON-LD structured data, no Microdata markup, and here’s the prompt to fix it. Content Structure at 59? You’ve got 11 H1 tags and 479 H3 tags on the page, and here’s what to do about it.
It even generates a complete prompt like: “I need to improve my website’s AEO score. My site saastr.com currently scores 67/100. Here are the specific issues: JSON-LD Structured Data at 0/100, Microdata Markup at 0/100, Heading Hierarchy at 40/100, Table Content at 30/100, Content Freshness Signals at 20/100. Please update my website’s code to address all of these.”
That’s a prompt you can paste into any AI coding tool and get working code back. That’s actionable. That’s useful. HubSpot’s tool gave me a sentiment score of 0% and no recommendations.
This is a real threat now. Not a theoretical one. When a non-engineer can build a competitive alternative to your AI product in a day using Replit, your 60% solution has a shelf life measured in weeks, not years. And it’s not just me. There are thousands of builders doing this across every B2B category right now. Every 60% AI feature that ships without real utility is an invitation for someone to build a better version over a weekend.
Instead of putting in their credit card.
Having said that, to be clear, my AEO tool may be better today. But will I maintain it? Keep updating it? Etc? Maybe not. That’s still why I recommend buying over building 🙂
Figma Make: The $1B Case Study in Shipping Late at 60%
If you want to see this dynamic playing out in real-time with real revenue numbers, look at Figma Make.
I tried Figma Make recently. I used my classic vibe coding test for every new vibe platform: I asked its agent to redesign the SaaStr AI site. What I got back was generic gradients, placeholder content, and trite AI-startup aesthetics from 2025. Purple gradients. That was bad from the leader in design, but what was much worse was it couldn’t even grab the content from our existing site to build a new version. But even worse — it hallucinated what SaaStr even is. And thus the site it designed was far worse than what we have today.
Every other leading vibe coding platform can pass this test in 2026.
Figma Make that I used in March 2026 might have been competitive in the summer of 2025. Back then, Lovable was just crossing $100M ARR. Replit was still finding its AI agent footing. The vibe coding category was nascent enough that a design-tool-turned-app-builder from a company with Figma’s distribution could have carved out real territory.
That was nine months ago. The window didn’t just close. It slammed shut.
Today, Lovable is at $400M ARR. They added $100M in a single month in March. Replit is also around $400M+ ARR and targeting $1B by year-end, and just raised $400M at a $9B valuation. Combined, just these two platforms alone are at around $800M in ARR and accelerating. Add in v0 and the rest, and the vibe coding category is well past $1B in total ARR. This revenue materialized in roughly 12-18 months from nearly zero.
And Figma Make? Figma only started enforcing AI credit limits, their first attempt at monetizing AI features at all, on March 18, 2026. Less than a month ago. They’re selling credit add-ons at $120-240/month and $0.03/credit pay-as-you-go. Management is describing the ramp as “measured.” The AI credit revenue so far is nominal. It’s a rounding error next to what the dedicated vibe coding tools are generating.
This is what happens when you ship a 60% product into a category where the dedicated players have been compounding at triple-digit growth rates for a year. Figma had every right to win this. $1B+ in revenue. Millions of designers already in the product. The brand. The distribution. The enterprise relationships. And they built the wrong thing for the moment — a design tool that generates generic Claude artifact mockups when the market wanted a building tool that makes design a step you skip entirely.
The products that could have been competitive 6-9 months ago are pack-ins now. Figma Make is Exhibit A.
Why This Keeps Happening
I think there are three reasons so many vendors keep shipping 60% solutions:
- First, they’re building AI features by committee. The product team gets a mandate from the CEO to “add AI” to the platform, so they bolt on a feature that technically uses a model but doesn’t have a clear user workflow or outcome attached to it. A score without a recommendation is just a number. A dashboard without actions is just a screensaver.
- Second, they’re trying to boil the ocean. Instead of picking one AI use case and nailing it and making it 95% good, they spread across five use cases and deliver all of them at 60%. A dedicated AEO tool that does one thing brilliantly will beat a platform feature that does five things poorly every single time.
- Third, they’re underestimating the pace of improvement in point solutions. If you’re a platform company planning a 12 month AI roadmap, the point solutions you’re competing against are shipping weekly. By the time your 60% feature launches, the best point solution is at 90%-100% and has been iterating on real customer feedback for months. You’re not catching up. You’re falling further behind with every sprint.
- Fourth, and maybe most honestly: internal team resistance. Most people at most software companies just want to work at the pace of the prior decade. They want the 18-month roadmap. They want the quarterly planning cycle. They want the three-sprint discovery phase before anyone writes a line of code. They want to ship the way they shipped in 2022 because that’s what they know and it was comfortable and it worked fine then. And leadership lets them, because fighting your own org on pace is exhausting and politically expensive. So the AI feature ships on the old timeline, built with the old process, at the old level of ambition. And by the time it launches, the market has moved two generations past it. The 60% solution isn’t just a product problem. It’s a culture problem. The companies shipping 95% AI products right now have teams that are building and iterating at a fundamentally different speed. Not because they have better engineers. Because they don’t have an internal constituency fighting to slow everything down.
The Bar Has Permanently Moved
What made this moment different from, say, the early days of mobile or cloud is the speed of quality improvement. In mobile, if you shipped a mediocre app in 2009, you had years to iterate before the market consolidated. In AI, the improvement curve is so steep that a 60% solution from April is basically unusable by October.
The products that could have been competitive 6-9 months ago? They’re pack-ins now. The products shipping at 60% today? They won’t even be useful pack-ins by Q4.
If you’re a B2B vendor and you can’t ship an AI feature that is genuinely best-in-class — or at least within striking distance of best-in-class — in at least one specific use case, don’t ship it. A bad AI feature does more damage to your brand than no AI feature at all. Because your best customers, the ones who would actually pay for it, are going to try it once, see the 0% sentiment score and the empty recommendations tab, and decide your platform “doesn’t really do AI.”
And then they’ll buy the point solution.
Ship something great or don’t ship at all. The market has moved past grading on a curve.
