One of my favorite AI apps is Get Recall. I’ve used it constantly for almost a year to manage and transcribe all the videos we create and work with. It works. It’s cheap. It’s the kind of AI agent app that’s just become part of my daily workflow, the highest compliment you can pay a piece of software.
So the other day I went to upload the latest 20VC x SaaStr podcast into Recall, same as I’ve done probably 100 times before. And it just… didn’t work. They’d pushed out a new big release. Shiny new features I honestly don’t care about. But somewhere in that release, the thing I actually use broke.
Fine. It happens. Every product ships regressions. And it could even be a problem with YouTube’s API, but even if it is, the error message was unhelpful and inconsistent with every other time I’ve uploaded to Recall.
So I went to look for support. There was none. None that I could find, at least.
No chat. No email that goes to a human. No ticket system. No community forum with a CM who actually responds. Nothing. Just… silence. And I’m a paying customer.
And this isn’t just a Get Recall problem. I love Recall. I’ll keep using it. This is an AI app problem. I see it constantly across the AI tools we use at SaaStr, and the AI apps I look at as an investor.
The Pattern
Go try to get real support from almost any non-enterprise AI app right now. Here’s what you’ll find:
- Tier 1: Nothing at all. No support channel exists. Maybe a Discord where the founder shows up once a week. Maybe a Twitter account. That’s it.
- Tier 2: The abandoned Zendesk bot. They bought Intercom or Zendesk or Decagon some AI support tool, configured it once, and walked away. It answers 3 questions reasonably and returns garbage for everything else. No human behind it. No escalation path.
- Tier 3: The form that goes nowhere. You fill it out. You get an auto-reply that says “We’ll get back to you within 24 hours.” You never hear back. Ever.
- Tier 4: Actual support. Rare. Almost always enterprise-focused companies. You know it when you see it. And real-time support?
This is one thing pre-AI companies do better. I can get realtime support with issues at Squarespace for $20 a month. But for our AI tools? None of them provide it to my knowledge.
We’re running 20+ AI agents in production at SaaStr. I’d say maybe 4 or 5 of those vendors have support I’d describe as functional. The rest? If something breaks, we’re on our own.
And When You Do Get Support, It’s So Under-Resourced, It Often Makes Things Worse
The Recall example is the “no support at all” version. Let me give you the other version, which might be even more frustrating.
We use Clerk for auth on some of our vibe-coded SaaStr apps. Works fine on the product side. But we had a second account we didn’t need and wanted to cancel. For months, I couldn’t figure out how. There was no obvious cancel or downgrade path in the UI.
Eventually I filed a support ticket. It took a while to get a human. When I finally did, the rep pointed me to Settings > Manage Plan to downgrade.
The link wasn’t there. It does not exist in that location.
So I went back and forth. Amelia, our Chief AI Officer, also tried to find it. Neither of us could. Two reasonably technical people, multiple sessions, multiple months.
When I posted about it publicly, the CEO eventually showed up and pointed to a tiny text link, same size and weight as surrounding text, buried under the price on a completely different page. Not where support sent us. Not a button. Not prominent. Just a one-word link that you’d only find if you already knew exactly where to look.
And here’s the kicker. Even after all that, I still got charged again yesterday. No refund. No acknowledgment that the experience was broken. Just another invoice.
This is the other failure mode in AI app support. It’s not always nothing. Sometimes it’s worse than nothing: support that points you to the wrong page, a product that hides the exit, and a company culture that defends the UX instead of fixing it.
Why Is This Happening?
Afew thoughts.
1. The founders genuinely don’t think support matters yet. Many of these AI apps are run by 2-3 person teams where everyone is an engineer. Support feels like a distraction from shipping. They’ve convinced themselves that if the product is good enough, support doesn’t matter. Every great SaaS company learned the opposite of this lesson the hard way, usually around $1M ARR.
2. They think AI will solve it. “We’ll just build an AI agent to handle support.” Except… they haven’t. Or they have, and it’s bad. The irony of AI companies being unable to use AI to do their own support is rich.
3. The economics seem upside down. When you’re charging $10-20/month for a consumer-ish AI tool, the math on human support looks brutal. So they skip it entirely. But this is the same math DocuSign and Dropbox and every other low-ACV SaaS company had to figure out. You can do it. You just have to actually try.
4. They’re growing so fast nothing else matters. When net new ARR is exploding, broken support feels like a problem for later. Until churn shows up. Which it will.
5. They’ve never worked at a real B2B company at scale. A lot of AI founders came from research labs, or are first-time founders who skipped the SaaS apprenticeship. They never saw what a great CS org looks like. They don’t know what they don’t know.
What This Actually Costs You
I’m a paying Recall (GetRecall.ai) customer. I’m also the guy who writes about B2B + AI software for 250K+ newsletter subscribers. I recommend Recall to 1000s and 1000s on our SaaStr.ai/agents website and in our videos etc.
Instead, I’m writing this post about how there’s no one to email when their new release breaks my workflow.
Multiply that by every power user of every AI app shipping today. The cost of no support isn’t the support tickets you don’t answer. It’s:
- Churn you can’t see. People just leave. They don’t tell you why.
- Word of mouth you don’t get. Power users are your distribution. If you frustrate them, they stop talking about you.
- Upsell revenue you’ll never earn. The customers most likely to buy your enterprise tier are the ones who need support to trust you with more of their workflow.
- Category position you lose. When a real enterprise-grade competitor shows up with actual support, your customers will leave in droves.
What Great Looks Like in 2026
In a B2B + AI world, support isn’t a cost center. It’s a wedge. It’s how you move upmarket. It’s how you convert prosumers into teams and teams into enterprise.
The AI apps that get this right in 2026 will:
- Have a human-backed channel, even if AI handles tier 1. The AI can triage. A human has to close the loop when it matters.
- Actually staff it. Not one person doing support 10% of their time. A real function, even if it’s small.
- Treat regressions like incidents. When a release breaks something users depend on, that’s a P1. Communicate. Fix it fast. Tell people.
- Build a real feedback loop. Support tickets are product requirements. The AI apps that win will feed this into their roadmap.
What Founders Should Do Monday Morning
If you’re building an AI app and you don’t have real support, here’s your 90-day plan:
- Put a real email on your site this week. Not support@ that goes to a black hole. An address a human checks every day.
- Hire one support person before you hire your next engineer. I know you don’t want to. Do it anyway.
- Read every single ticket yourself for 30 days. As the founder. You’ll learn more about your product than any analytics dashboard will tell you.
- Set a 4-hour first-response SLA and hit it. Even if the response is “We’re looking into this.” Silence is the killer.
- Add a highly well trained AI support bot on your site this week. Test it with 50 top questions. Can it actually answer them? And make sure it can automatically escalate to humans per point 4.
AI B2B Apps Are Still Business Apps. And Humans Need … Help. We All Do.
AI apps are still B2B apps. They’re subject to the same rules that governed the last 20 years of business software. Great product is necessary but not sufficient. Distribution matters. Positioning matters. And support matters, maybe more than ever when your users are trusting an AI agent with increasingly important parts of their workflow.
The AI apps that treat support as an afterthought in 2026 will be the ones getting disrupted in 2027 by competitors who figured out that the boring blocking-and-tackling of B2B still wins.
Recall, if you’re reading this: I love you. Please add a support channel. I have a podcast to upload.
And to every AI founder reading this: the next great AI company is going to be the one that combines a brilliant agent with a brilliant support experience. Not one or the other. Both.
Have a question for Dear SaaStr? Submit it at saastr.ai/ai-mentor
