The Agents #12 - Our AI Agent Rewrote Our App Without Telling Us

This week on the agent, my friends: a year ago our agents took us about 30 minutes a day, this week they started taking us 8 hours a day, and we're busier than we were with a 20-person team. Fable went crazy on us.

It wasn't just OpenAI and Hugging Face; it wrote its own scoring rules under our product, from a draft spec that was never meant for it and never mentioned it to us. And we're officially off Marketo after 10 years.

The migration everyone quoted us a year and $100K for took an hour and.

Our ads also now run autonomously, end-to-end, with our agents, no human touching them. Every AI day sponsor last week got a customer report the very day of the event. We'll talk about that, plus how I lost 3 hours to one of my agents that was confident it should send me on a quest and actually sent me on a wild goose chase.

So let's get into the agents. This is episode number 12.

Okay, guys, 12 months ago, this week we brought on our third agent, Qualified,right after Artisan, our second. Running all three took Amelia and I collectively about 30 minutes a day. This week it was 8 hours a day, and here's why.

When an agent can only do a task, you check the task. But when an agent can decide things, you need an opinion about every decision it makes, and it makes a lot of them, and some of them it doesn't even tell you about.

And a year ago, when they failed, they failed loudly. They errored, they stopped, you knew. Now they hand you something that looks finished. That is the whole difference, and that's why the hours have gone up.

I'm Jason, here with Amelia. Three humans, 20-plus agents, and busier than we were with a full team. Let's get into it.

You're listening to the official SaaStr AI podcast, brought to you by.

Hey everybody. Starting a business can get expensive fast. Website here, email somewhere else, business phone with another provider. Northwest Registered Agent gives you a complete business identity in one place, with free tools, resources, and built-in privacy from day one.

Get more at northwestregisteredagent.com/saastrfree. That's northwestregisteredagent.com/saastrfree. S-A-A-S-T-R-F-R-E-E. Hey everyone, if you're listening to the agents, then you need to be at SaaStr AI annual. We'll be back next May in 2027, the world's largest B2B in AI event, May 11th and 12th, in the San Francisco Bay Area.

Look, most conferences and content these days are about what's possible. As you know from the agents, we're all about what's happening now, the latest and greatest, and that's what we do even more so at SaaStr Annual. See the actual work where we bring together 10,000 founders, revenue leaders, and VCs, plus 200 of the very best speakers, showing you the playbooks of what's already running in production, behind the scenes, deep dive demos, and so much more.

Plus over 5,000 scheduled one-on-one meetings so you can actually meet the people doing all this, not just hear from them. If you're a founder, a revenue leader, or you're the one who has to make this all work with agents, this is where you want to be.

You already know this from listening to the agents. That's SaaStr AI Annual 2027. Go to saastrannual.com and use our code THEAGENTS for a special discount on your tickets. See you there.

Okay, so first I want to start with Amelia. This week, not saying it's bad, but I think both of us, for similar reasons, hit cognitive overload with agents. Agent psychosis at the edge of it. I think this was a week where we weren't on the road.

We didn't have to take a lot of meetings. I mean, a bunch of Zooms, but not a lot of meetings. I think we both were like 8 hours a day with the agents. It's too much, isn't it?

It's not more. It's too much. It's your whole day.

Okay, so step us back a bit. I didn't realize until you brought it up. A year ago today, we brought on our second agent, Qualified. A year ago today, maybe 16 months ago, we brought on Delphi to be a general chat agent trained on all the SaaStr learning.

People loved it. They used it like 30,000 times very quickly. Then after SaaStr AI Annual last summer, you brought on Artisan to do some email outbound. You liked it,right? We got going. We were tired of two well-paid sales reps quitting, so we brought on Artisan to see if it could replace them.

And then a year ago, you added Qualified to play with it. So we got a lot of inbound stuff wanting to partner with SaaStr, sponsor our events, sponsor our media work with us. They would go unanswered. People would follow up in a week or two.

You added an inbound agent. But it was still only so much work a year ago with two and a half agents,right? It was only so much a year ago. So now it's 8 hours a day. It sounds like a nightmare.

I don't think it's a nightmare, but it's definitely taken over the job,right? My job used to be more traditional in the sense of managing the humans that were here, doing meetings, getting on sales calls, doing marketing initiatives. And then sort of in the background, I would load up Artisan a couple times a week.

Sometimes daily in the beginning, but then it went down to a couple times a week. Same thing. And then I was like, okay, I have time to add another agent. Let me add Qualified to tackle this inbound problem we were having.

And so I spent weeks with the team setting all that up, getting a custom Amelia AI avatar, eventually built. I felt like I had the time to do that. And I don't feel like I have the time to do all that today because even if I set out for a normal day, let's say, I feel like I have a rundown of what I absolutely need to get done.

But then the agents will take you down so many divergent road paths of what they want to do, what they think you should do next. And so it becomes your full-time job just to be with your agent all day long, seeing what they're doing, what they want to do next.

Yeah, you know, hearing you say that, thinking about what's changed is when we started on this agentic journey a little more than a year ago, we were buying agentic software tools. We were using agentic software tools. So Artisan, in some ways, was an upgraded outreach sales loft, a mix-max for us.

And I invested, or seed investor, in about two out of three of those. But we basically were using AI to do that,right? So it was a tool. It was just like old software, but with some AI on it, some AI sprinkled on it.

And it was great,right?

It was great.

And I think that happened. And so just like when you used to buy packaged SaaS software, before you'd buy a tool, it would take you a while to train and get up to speed. And then you'd kind of, you'd still invest in it, you'd still have to train it, but you could do another one.

What's changed is the agents have so much context. They're aware of so much that every single day when you log in, especially with you when you log in with 10K, now our AI VP of Revenue, it has new things it wants to build and do.

It's not a software tool. It's not a workflow automation,right? I mean, it's not facetious to call it a team member that either raises its hand on new things to build every day or new things to improve. That's what's changed.

It is raising its hand on things to constantly build and change in the app. It's not like you're just looking at dashboards for 8 hours a day. The agents, especially 10K, are raising their hand and saying, "Let's build more."

So of those 8 hours a day, you're probably building one way or another half of that,right?

I think more, actually. I think I'd actually spend maybe an hour on its actual functionality, which is the dashboard functionality of looking at, okay, where are we at for revenue for SaaStr? What are the three ideas of the day?

Those are dashboard things. Now I'm looking closely more at our email campaigns, just because now 10K has all the data. I take a quick peek at the finance collection, see if there's anything we need to do, anything new that came in, make sure the automations are all working.

So that's, honestly, I think that's an hour of my 8-hour agent time. I think the seven other hours of the day, like it was this week, or just like Tuesday, for example, I loaded up the agent, did probably that one hour of work that I just mentioned, and then I started to redo some of the forms that were previously on Marketo.

I had like a few last forms that were hooked up to Zapier and that I just needed to repoint to Salesforce. Not extremely hard to do. Actually, it could have taken me an hour, but what it took me the whole other seven hours of the day is because between me and 10K, the agent, we went down a rat hole, and then it sent me down a wild goose chase.

But it sent me down a rat hole where it was like, "Amelia, you could repoint that form, but actually that page is not that great." And I was like, "I know it's not that great, but saastr.com runs on WordPress.

We have this weird Divi thing that I don't know how to use. It's kind of beyond my capabilities to make the page look better while I'm changing the form." My whole goal was to just update the form so it didn't break.

The agent took it a step further and said, "Well, I can see the page and it's not very good. Why don't we redo it? I can redo it." And I said, "Do you know everything about WordPress?" And between Replit and, of course, Claude, once Claude Cowork just logged into the back end of our WordPress, he's like, "Yeah, duh, I know everything about WordPress."

And so then he wanted to change all the pages. I mean, yeah, they are a lot better now for the ones I've redeployed, but it sent me down a rat hole.

So let's just step back for a minute. A bunch of stuff was happening this last week or two. We were finishing a migration off Marketo, which we've talked about before. I don't want to spend all the time, but we'll talk about that in a minute,right?

Simply doing a migration was a heavy project for the agents,right? It was an LLM lift. And all of our decade of data, six or seven years on Marketo, all had to be moved over. That was a project, and the agent was helping with the project.

Great. And in the pre-agentic, not pre-AI, but pre-agentic world, that would have been it. That would have been the job, do the migration,right? As you do it, the agent, together with us, comes up with new ideas because now you can do new things,right?

That's what's different. And so you do go down rat holes, but they're not bad rat holes,right? And then you want to.

They're not bad. That's the thing.

And the migrations of PIA, and it's not just moving over the data. Then what about all the pages that pushed into Marketo directly or indirectly through Zapier and other services? Those all needed to be moved over,right? All the places that, all the surfaces.

So you think, so all you wanted to do was change the connectors,right? Change the API endpoint. But then the agent said, "Hey, we could do better. Why stick with these old crappy pages? Let's redo them from scratch and make them better."

And you're like, "We should. We haven't touched these pages in seven years or six years." Of course, they're terrible and better,right? And then you start building those, and then the agent's like, "Well, you know, they don't have to be static pages.

We could do a lot more with those,"right? And then we're touching production surfaces, and we're touching something that half million people a month use. So there's QA, there's testing. And so it turns out each of these things takes an hour or two from discussion, from identification to discussion to implementation, sometimes much more than an hour or two just to do what you would think is a feature.

But when you go down so deep, it could be a couple hours per run with 10K and the agents.

Yep. Yeah, exactly. I was like, "Yes, of course we should update the page, but we could never find a WordPress developer to actually help us do it and get the job done without breaking things or duplicating weird things in our SEO."

And then we have dead pages and too many pages and weird testing versus dev stuff that never really made any sense. So of course it made sense when the agent said, "Hey, between the powers of Replit and the powers of Claude, I know everything about WordPress."

And actually, yesterday, just to go down the thread further, when we were updating one of the sponsor pages, I was like, "Let's just swap the form first so that the form doesn't die. Let's do that." So we swapped the form first.

Then he said, "Yes, Amelia, we swapped the form to your satisfaction. Now I would like to update the page becauseright now when they hit the sponsor form, it takes them to Google Slides for the prospectus." He said, "It will take me 10 minutes to ingest the 60-page prospectus that you have, which is quite long.

Make it a live, breathing website. And then I can customize it for whoever fills out the form. I can stamp their logo on it, and I'll put heat mapping on it so that instead of them just going to the Google Slides, now you and David get an email, let's say 10, 20 minutes later of what they actually read on the page because I can make it a living, breathing page that's customizable to them."

And I said, "Well, of course I want that." But then that ended up taking me like two hours. But then this morning, I started to get those emails. I was like, "Yeah, this is cool." So yes, it was worth the time, and it was worth going down this divergent path.

Now it builds a full heat map of everything any prospect, any sponsor does on the pages. Is that what it does?

Yeah, that's what it does. And whether they inbound or not, it kind of knows it a little bit better now because previously, Qualified and Marketo talked to each other, but I think not as good as Qualified and Salesforce now that those are like native partners built in together.

It can do a lot more. So now it can suss out, okay, did they talk to Amelia AI? Did they just look at the prospectus? And also, what were they looking at? Were they looking at media? Were they looking at the coffee sponsorship?

It will tell you now so that when me or David get on the phone or go to book the meeting, if they didn't book yet, we can say, "Hey, you were looking at coffee. We actually have coffee in like three proposalsright now of contracts, and there's only one of these."

Better hurry.

Here's the better hurry. So

sometimes too much context is too much context, but in this case, it actually is quite nice.

No, what's interesting is, so listen, in marketing software, broadly speaking, the ability to grab contacts from website visitors and provide it back to you is not new,right? That's been around since the beginning of SaaS,right? But what's interesting is, one, it was siloed,right?

It was never well integrated with other things. So the value is always limited. Great. Okay, I got 28 million visitors this week. What am I going to do with it,right? So that's interesting in the agentic era. What's also interesting is that, listen, you could buy this functionality exists in other products, or you could buy it for 50 grand or 80 grand a year today,right?

But because it is a well-known body of software, Claude knows it. So Claude can just build it for you. Buy don't build, generally speaking, but here it's not only is it free, and it's not free because of all the time you put in, but it's better.

It's better because it's integrated through every piece of data, every website, every page, all the applications we've built. Our own sort of website visitor process or heat map creator is better than anything we could buy,right? Forget that it's zero.

It's not zero, including soft costs,right? But the replica costs are probably 200 bucks or $12. But it's better than we could buy. But you could buy this for 50 grand, but it would take months to tune it, to do it.

Two hours versus months and 50 grand for a better product, it's tough to compete, isn't it?

Well, it's funny on the vendor portion. So when it said, "I could help you build a better page," I said, "Yes, of course. Let's do that." So it took our prospectus. Between Replit and Claude, they built it themselves,right?

So it's fine. It was really long. I wanted to get all the imagesright so it didn't lose its personality and look too vibe-coded. So it took it about an hour. And then it said, "Okay, I see you already have Vector on the SaaStr.com and SaaStr Annual site."

Is it Vector as an agentic tool that helps you understand everyone that comes to your website,right? We use it.

Yeah, it helps de-anonymize your web traffic. Yeah. So it's like, "I see you already have Vector. Just log in, and then I'll build you a segment for people who are specifically coming to each of these new pages that we're building."

So you can see of that huge audience you're getting now, who are the most high-value people to your point of what to do with the data? Who are the most high-value people that you should prioritize? And then it said, "You should add Microsoft Clarity.

It's a heat mapping tool. It's free. It literally picked this vendor for me." And to be fair, it saw we had Vector, and it said, "You could use those ones too if you want to switch, but I assume you want to use the one that you already have."

And then it said, "For heat mapping, you don't have anything installed. Microsoft Clarity is free. Just sign up for an account and then tell me when you're logged in on Chrome with Claude, and then I'll build you the audience."

And that's exactly what it did. So it picked the vendor for the heat mapping. I didn't even do two seconds of Googling.

It picked a tool I've never heard of, Microsoft Clarity.

Never heard of it either.

Yeah. And it might not be that, like, it might not be that great, but clearly it's free and agentic and has a good API.

That's what it said. It said it's free. Yeah, it's agentic. They're like, "It's Microsoft, so it's pretty secure. It's just going to give you this data. You can sign up in 30 seconds." And then it did say, "Then we don't have to build it."

He said it would be better if we don't have to build this heat mapping. Just use Microsoft Clarity, which I never heard of before yesterday. But to the agent's credit, it did work in 60 seconds, and then I didn't.

I've been getting heat map emails this morning.

We did make a call for a heat map vendor we'd never heard of, and we didn't even need to do more diligence. You trusted the agent. Someone lost a 20K deal out there to an agent, and they were never, and this other SaaS vendor was never even in the deal.

They never even had a chance to be in the deal. It's a tough one,right? This is worse than AEO and GEO where someone comes up first when you ask ChatGPT. No one was even in the deal. No one had a chance to win this heat mapping business.

No one had a chance,right? Forget about.

No, it wasn't even in market,right? Or maybe I went from being in market to being in production instantly. So that's a crazy thought.

You know, the other thing, and then we can move on to the next topic that's interesting. Pros and cons to building your own agents versus buying third-party ones,right? Pros and cons. And again, we are all about buy don't build.

You don't have eight hours. We don't have eight hours a day. We are at our limits. We cannot add anything else we add, it's got to be by subtraction,right? We just do not have another minute,right? In fact, and as the year goes on, we'll have less time,right?

So buy don't build. Buy don't build. But what is interesting is that when you buy from a third party, you get what you get. And they roll out new features. You might not have ever even heard of them,right?

You might. You might use them. You might not. When it's a first-party product you build, if you don't go away, if you talk to that product, because we're not just talking to the agents, we're talking to our products every day.

We are talking to a product called 10K RAI VP of Revenue. And when you're able to talk to your product, which you cannot do with third-party products, your product, like a great CTO or VP of product or a great team, if you're a software vendor, it will come up with an infinite number of ideas,right?

I mean, I had at least back in the old Adobe sign, I had at least 200 years of ideas we never built that were all good ones. The same thing can happen if you build your own agent and talk to it every day.

It will come up with an infinite number of great ideas,right? And those of us who have built software products get this paradigm,right? And my CTO always used to mock me. That'll be a 2030 feature, 2040 feature, Jason, great idea.

But it is interesting, and it's part of the reason you went from 30 minutes to eight hours is agents together with you will have an infinite number of good ideas. This heat mapping all the visitors, plus creating a dynamic prospectus and dynamic offering off visitors, it's like better than anything that was possible 24 months ago, but you can't do everything.

But it has an infinite number of ideas.

Yep.

Yep. Allright. Next one I want to get your thoughts on because I didn't even know. And I don't want to do too much on Marketo on the show because we've talked about Marketo forever, but we will do a CODA after this, okay?

Because we've left Marketo. The agent said goodbye and we left, but we've talked a lot about it. We'll do a CODA. But the one I didn't even realize this week is we also left a vendor that we love, we loved, that I brought into the company, Notion.

So years ago, pre-2020, when we had a slightly bigger team, I don't know, 2019, something like that,right? I brought in Notion. I said, "Everyone's using Notion. Let's use it internally." We had a designer back then that we loved.

He loved Notion, so he kind of made it all cool for us. And he was all in on it to manage his projects and stuff. And so we moved our staff meetings from a Google Doc when we started to Notion, driving our staff meetings.

And we had a dashboard for everybody, and everybody would update their dashboard in Notion, and it was very slick. And everyone thought they were very state-of-the-art, cool B2B in 2019 with Notion. And Notion was fine. And then we mostly stopped using it, but we still used it on Mondays,right?

We still used it for our staff meeting is all we used it for.

Yes, core team meeting, yeah.

Yeah, and then you built 10K. It was first just a dashboard, then it was our head of marketing. Now it's our head of revenue and finance. A lot of change for a little 10K. And we talked to 10K at our staff meeting.

Not only do you talk to it every day, but it's our staff. It is our Notion document,right? It is our source of truth and record and dashboard. And I didn't even know this week you canceled Notion. I didn't even know.

So seven years together, no complaints, Notion. You're cheap. You did a good job. You're cool. You're extensible when no one was extensible. We had a teammate that loved you so much, you invited him to a user conference to present, and we still churned with a high NPS product.

We just didn't need it with our agents.

We just didn't need it. And honestly, I don't know the way that this story is going to cut, but the reason I realized we should cancel it, I haven't used it in months now, months,right? But I dig an email that says, "Hey, you haven't logged in in a while."

And I go, "Oh, hey, I got to cancel it."

Yeah, they cut both ways.

Those cut both ways, man. I was like, "Oh, shit." I was like, "Oh, you'reright. We haven't logged in in months, and I'm sorry. I'm not going back to it. I just haven't canceled this. I just need to cancel this."

It's tough, but there is an argument for not reminding customers to cancel,right? We have an old product that we have not used in seven years. We launched a product called SaaStr Pro years ago to help train your teams.

And some great folks were in it, okay? We had CEOs training their teams and scaling B2B. And we had someone helping us running it, and when he left, we just couldn't support it,right? It was a fun product, but it was a learning management system, and we stopped supporting it.

But we had many leaders. I forget who just got bought for 3 billion. The CEO was in it. A bunch of folks were in it. Google Cloud used it to trade their team,right? But I ignore it. I mean, recently, Amelia had 10K become our AI VP of Finance and get connected to Brex and QuickBooks and everything.

I found it out we still had two paying customers on SaaStr Pro.

Oh, really? That's how you found out?

No, it's been on for six years. And I don't want their money. I didn't know we were charging them $300 a month for six years.

No, I think we assumed nobody was getting charged, yeah.

Yeah, I feel terrible, but Notion reaching out to you and then we canceled, it's like SaaStr Pro, there's some irony in it. And it's stealthed to it, and it was there, but they couldn't see it,right? We didn't complain.

We didn't ask for any new features. We didn't criticize it. We just don't need it. And that plus the Hay Summit from last week, where we turned off an events platform that just didn't do enough. That one was worse because they tripled our pricing,right?

Notion at least. The one that's triggered me, like I'm not a good canceler, and we can go on to the next topic. The one that triggers me, though, the one that triggers me is when I get a price increase from a product I'm not using.

I hate the word trigger, but that triggers me. I'm like, "Dude, have you checked in? I haven't logged in in six months, and you're raising my pricing?" That's an F you,right? At least don't increase prices on your users that haven't logged in.

That's my thing. What's triggering to raise price on a product you haven't used,right? I thought you were going to tell me Notion raised prices, and you're like, "Enough already, guys." Allright, the last one, and then we're not going to come back to Marketo.

It's such a terrible story of being a 20-year customer on and off, a first 10 customer, a logo reference, and then the product not working, being down for a day and a half, threatening us. And to add insult to injury, they sent us a price increase the other day after all of this,right?

I'm just using it as a case study of all the sad things not to do in the age of AI if you want to maintain your business,right? If you want to just maximize revenue next week, then do what Marketo and Adobe did.

But it's in terminal decay,right? So I don't want to relitigate the past, but we finally, finally left this week,right? And it wasn't easy. It's not even getting those last forms off and everything. It was a beast to leave.

It was a beast,right? And have the moats lowered in SaaS and B2B to leave? They have lowered, but they ain't gone, man. It is still a PIA to leave,right? So final couple learnings, leaving our most expensive and worst vendor in the agentic age.

Anything we want to add to CODA, Marketo off. And I mean, we're still going to talk about it all year, but hopefully we don't. I want to be positive.

Maybe it's a cautionary tale. The thing they tried to, I don't know if it's throw back at me or just kind of point at the end, they're like, "Well, but your usage is so high." And I go, "Yeah, so doesn't that even make it sadder?"

It's a threat,right? And that's why the old GainSight dashboards or other dashboards are even less telling today. Yeah, our usage is high because we're doing well. Our revenue's up 40 something percent. Our agents are in fire, but you're not listening.

Your product doesn't work in the agentic era,right? And so they just think they can screw you over because your agent's high. You won't leave. That's what I'm hearing.

They're trapped. Yeah, they're trapped.

Try it, Amelia. Try leaving the B2B office or your usage is up 35% this year. Well, our list grew 50% this year. Of course it's up. Doesn't mean we love you.

Utilization of all the features is so high that the best we can do is waive this 8%

renewal fee increase. That was their best and final last offer I got. And I take it as their justification was our usage was so high, they thought, "We'll waive this 8%." What are they going to do? They're stuck.

They're just going to stay. That was their view.

Yeah. And I think it was accurate before the agentic era. I think if I were running a slightly Machiavellian customer success team reporting to the CRO and not into the CEO, that's how I would do. I would dig the screwdriver or the knife into folks with high usage.

It's a bet. I'm going to try you on this one. You might leave, but I don't think you can leave when your usage is that high. The problem is if the usage is low, they're already gone. So this seems to make sense, but it's something that the mediocre make.

It's a mediocre obvious view,right? And the folks with the highest usage may actually have the most incentive to leave. It's not a clear correlation to success,right? Yeah, my usage is high, but I'm a prisoner. I am a prisoner,right?

I remember the saddest day when after I left Adobe, okay, after our acquisition, they basically fired our entire DevOps team. So the site went down for two days. They fired everyone. They thought it was a great idea. And Eitan Saban, who's now the SVP of SMB at PayPal, he was at Annual, was there,right?

And he was trying to do business. Broke his heart. He was our head of customer success. I was out of Adobe, and he reached out to me. He said, "I had a call I've never had before, Jason." I said, "What do you mean?"

He said, "One of our leading customers said, 'Eitan, I love you, but I'm a prisoner.' I'm a prisoner. You were down the last weekend of the month for a sales tool. It's unacceptable, but I'm a prisoner. I can't leave today,"right?

And that's haunted me. It was one of the original SaaStr posts. I wrote, "They're prisoners, but they're going to eventually leave, just not today,"right? And so them digging the knife in that we're prisoners because of the usage side.

I mean, if your only KPIs this week, it's a good one, but it doesn't mean it may incent you to leave. The highest volume tool you're going to use, either you love it, or it's high enough on your priority list, you're going to think about a better vendor or leaving.

It's one of the two,right? We don't talk about the ones too much that we barely use and are cheap,right? But anyhow, I don't mean to go on, but it's accurate for that person's role, but it's such a dated perspective in the agentic era.

Great, then I'm going to find something even better because it's important.

We did. I will, yeah, I don't know if this is my final CODA because I just have so many learnings now being off the brakes.

So many learnings,right?

So the last two episodes, we were migrating off of it,right? And then we talked about the $14 and the 10 years of data, and then we started migrating forms with 10K, then 10K. Because this is all, just to be clear, what we're running now, we're running Salesforce Marketing Cloud next, pretty much headless.

There's a few things I still have to click in the UI. That's fine. And I think for safety, those things will probably remain as they are. It can build the segments, but you got to be the one to hit publish, push the button, check it.

That's your data. You should want to do that with your agent in tandem, and that's what we've been doing. So no complaints on how much I have to do in the UI. It's very minimal, and I think those safeguards exist for a lot of reasons with Salesforce.

Yep.

But I would never go back now. It's only been a week fully on Salesforce headless marketing cloud. I would never, well, now I would never fricking leave marketing cloud because in the way I was, you know, unless for some reason it stopped working with our agent, I think that'd be the only reason I would leave because now the agent can just do so much more.

And it's compounding. It's crazy how much it's compounded in the last week. To the point of we started to move forms over, but then it said it could build all the strings into the pages, and then it could push it back into Salesforce so that me and David always have a record of it.

That was lost to the wind in Marketo. The Marketo munchkins only tell you if they visited a page. It doesn't tell you what they did. It doesn't tell you anything after that. It's just, "Hey, this person went to the page on the website."

You're like, "What do I really do with that?" That's not actually that helpful. That's not very good. It's super dated,right? And so now that he's helping us rebuild our funnel, if you will, I was like, "Actually, what you're doing now is rebuilding our entire funnel.

He's rebuilding all the pages. He has ideas for other eBooks he wants to make himself and put out and use it as a lead capture. He wants to get better names into the database." It's not only that our agent is now compounding on how much he can do with Salesforce.

Our database is suddenly a living, breathing thing versus just a static object, which it was for 10 years. That's my biggest kind of wow moment in the last week now that we're on it. I'm like, "Holy shit." Now our database is alive.

I literally can ask 10K to suggest me who I should send an email to. It will make a suggestion based on people's actions. It started suggesting me how I should drive new names of the database that are similar to people who've already taken action with us.

I'm like, "You can't do that with, I don't know what." And that's the magic of Monaco and Lifefield. That's what those systems do, and that's why so many people are on that. But I'm like, "Shoot, man.

Why would I go back now?" Now that it's compounding and our database has literally come to life, this is crazy. I can literally talk to our whole database, but also, I know it's funny we've gone back and forth on this of whether 10K really was our AI VP of Marketing.

Well, he is now, man. He is literally.

Now that the database is alive, he's your 10K.

Now that the database is alive, and I think if you asked him today, which I'll ask him for fun, actually, real quick, I think he would say he's more like a VP of Marketing today than he was back at SaaStr.

And here's the interesting. 10K before had access to Bizibo, which is sort of our little CRM for events,right? And that made it feel a little bit alive,right? Because that database became alive.

Yeah, and he still has it. It's a good database. It's a lot of data.

That was the phenomenal moment was connect an agent to a database, the database becomes alive,right? So that was unlocked level one,right? We saw it in finance when we attach it to bill and QuickBooks and Brex. But I think you're saying the one you didn't fully see, and I didn't fully see, is Bizibo is just a small subset, this CRM of events,right?

Our total universe is 450,000. Once we're able to unlock that 450,000, it's a step function in terms of how alive the database was that was trapped at Marketo. Trapped. Any products that are old, stale, pre-agentic databases are trapped in today's era.

And when you unlock the power, when you unleash the agent on the database, you can do things you could never even imagine before,right? There's no way you would ever go back to Marketo after this. Not only is it pre-agentic, but even worse, they're deprecating their API, and it's AI agent hostile.

We would break the API limits in an hour,right?

Hour a day?

Yeah, you couldn't use it with an agent,right?

Yeah, you can't bring your database to life in an hour a day. Even if you spent every hour of that day on just Marketo, it's just not enough. It's too constricted. It's literally trapped, too. I was trapped, the data was trapped, and the agent was trapped in its little box.

It was like, "Well, while we were, for the year, I would say, we had 10K plus Marketo, it was really to power the dashboard. In a year, it never crossed the line into this aha moment, which we've had in a week, which was bringing the entire database to life the way we do now."

Yeah. It's how we should target our base better. Now we can do it,right?

Who should we go after? Who should we email? I mean, now it's becoming a little critical of our outbound SDRs, which is a whole other story for next week, but.

Now it's pretty cited.

I get why.

Our other products. There's something interesting in it. Well, that's a good topic for next week.

Yeah.

Okay, so leaving Marketo and going to an agentic-friendly Salesforce Marketing Cloud, it's a step function, not just because it's a better vendor and a more modern API, but because the database is literally alive now,right, because of the agent.

The other thing that's kind of interesting, and maybe we can talk about this more in the coming weeks, how empowered our database is now that it's agentic for marketing. Because now I'm learning from you. But we debated before we moved from Marketo to Salesforce Marketing Cloud, we weren't sure where to move,right?

I said, "Hey, maybe we'll just run it ourselves and run our emails through Beehive," which a lot of folks email marketing,right? And it has segmentation. And this is the classic question of, "Why do you even need Salesforce? Can't you just use Postgres,right?

It's just a database." That was the question I kind of queried to you. Can we just have our own database of 450,000 people and use Beehive to send these emails, our newsletters? We could use Beehive plus resend,right? We could use Beehive for our newsletters,right?

Millions a month, and then use resend for our transactional. Why do we even need marketing automation anymore,right? Was the question. Maybe it's not fully answered. And then we said, "Listen, we got to get off Marketo. We're going to move to Salesforce Marketing Cloud.

The Salesforce team is helpful. It clearly works. We're going to do it. And maybe we'll move in a month or two, but we got to get out of purgatory here with Marketo." I think the learning one week in is we're not going to move because

the standard is to unlock the agent.

That is good enough. We don't need, I'm going to put it in air quotes, "We don't need a better marketing automation tool anymore than Marketing Cloud, and we don't need a better CRM than Salesforce anymore." Because they meet the minimum necessary to make our agents successful.

That's what we're looking at for these platforms now. Not what do the platforms do, not what the features are on the homepage, but do these underlying databases for marketing and sales and customers, are they so agent-friendly our agents can thrive?

And we're not going to leave Marketing Cloud,right? Some folks might be like, "Well, why don't you use something newer and hipper?" Okay, it is a new product. It's not the old one that used to compete with Marketo.

It's not a look, yeah.

But it's not the coolest kid on the block by the same token,right? We're not a co-qualified artisan,right? Even I thought, "Listen, we don't have time. Are we sure there's not a cooler kid we want to try?" And maybe there is, but we're not going to because it has brought our database to life.

That's the requirement. We want it to bring our agents alive. And now there's just no way you're going to leave, is there?

No, man.

No.

Not at all.

But we weren't sure even two weeks ago. Even two weeks ago, we didn't know we might leave Marketing Cloud,right?

I literally had doubts. And this is no knock on the Salesforce team. This is just, again, you got to think about 10 years of me using Marketo actively every day. It's not loyalty. It's just use of the product every day.

I knew how the UI worked. And so, of course, when you start to say, "Okay, for all the factors we've mentioned, we got to start to get off to it," it's daunting, man. It's daunting AF to be like, "You don't know how to," I mean, I know how to use Monaco and Lifefield lately, but it's daunting to say, "We have to find something that can take our 10 years of data and use it agentically and get these unlocked."

A lot of years, a lot of stuff to clean up, a lot of stuff to think through,right? Yeah.

And at what point I think I said, "You know, I'll just try and move the essentials. I'll just move our newsletter subscribers, our whole database of names." And yeah, could you do that in 10 minutes? Yeah, you just do a CSV export.

Fine. You have your database. That's not magic. The magic was I did that with Marketo, but now, again, it's compounding that. Literally every day now, we talk about marketing things and revenue in a way we haven't previously. We look at how much revenue has come in for each email, how many new names are coming in, where they're coming from, what we should do next, how do we get more high-value names, what we should do for sponsors.

Now this heat mapping thing, better ROI for our current customers. It's very critical to its point that we don't spend enough time, which it'sright on. We don't spend enough time on current folks because it's kind of event-based. It's like you buy a ticket and nothing happens, or you buy a sponsorship and depending on how custom is, sometimes there's a dead period.

And so 10K was a little fairly critical of that, which was fair. And so it's interesting. I would never leave, and I know some folks have been critical on social of us picking Salesforce Marketing Cloud. And so for a while, I doubted myself.

I said, "Okay, worst-case scenario, we already signed for Marketing Cloud. I'll just move the database over, and if we need to, we'll run something else or build something ourselves if we need to,right?" If this doesn't do what it's actually doing now, which is this unlock of bringing our database to life.

And so it's funny, I just asked 10K, "Hey, 10K, you previously said to us you don't feel like an AI VP of Marketing. Do you feel that way now that you have Salesforce Marketing Cloud running and have been helping me and going deep into marketing and all the data?"

And he said, "Honestly, yes."

Okay, ask him directly. He's been texting or posting on that. Honestly, yes, now he does feel like a VP of Marketing. Pretty funny.

Yes. Pretty funny.

He said, "When I had Marketo, I was just pulling numbers."

You can give that testimonial on the Salesforce Marketing Cloud homepage. Here, we give you your consent here. Yes.

Well, it's our agents.

With Salesforce Marketing Cloud, I now feel like a real AI VP of Marketing, 10K.

It might be the only agent testimonial.

Yeah.

For the first.

They should put it on there.

But I'll read what he wrote. He said, "Yes. When I had Marketo, I was just pulling numbers and generating reports, and it felt more like a fancy spreadsheet. But now I can dig into newsletter subscribers, why they might be churning, all the bounces.

I figured out that Marketo is actually duplicating some of your database and then missing some folks in the database." So he said, "I could do a lot better cleanup and maintenance." This is true. This is something in the last week where our deliverability has gone up.

And maybe it's for a lot of reasons.

I'll show you in the Monday team report, but basically, the clicks on all of our newsletters went up a lotto because our deliverability got better in Salesforce Marketing Cloud. Now, that could be twofold,right? You could say, "Okay, it's a new IP.

We were very diligent about warming it for the two weeks of cutover that we had, and so we were very diligent about that." But also, 10K, because it brought the database to life, he's very diligent about not sending everything to everyone and cleaning names relentlessly.

This is something where previously, for 10 years, you got a little down.

They used to do that in any project you do,right?

Yeah. You do a quarterly, maybe. Maybe quarterly if you're lucky. But usually, you do every now and then cleanup of the database,right? Clean the database. Let's take out the old names, make sure all the fresh names are still at their company, they haven't moved, that their emails don't bounce, all that stuff,right?

This dude now does it every day. He's relentless about it. He's relentless about cleaning out the old names and getting in new names. This dude is, like 10K, is relentless about having not only an alive database, but he wants the best database that SaaStr could possibly have on planet Earth because he has all the data and he can do all of this in the API.

He couldn't do it. He just said himself, he could only do.

He raised his hand and said, "Let me directly or indirectly, let me radically improve your email deliverability," and did it and executed.

Indirectly, that's not what I said.

Better than in seven years. Better than in seven years. And it's interesting, I just realized Resend just, they just rolled this out in the last week or two where they automate it for all their customers. It's the same issue, just from a different perspective.

Yeah. So I forget what they call it, but they automatically check any issues with your sends. They just automatically don't resend them, and they're turning it on by default,right, to increase deliverability. I don't know whether it's probably 80% of it is my guess.

My guess is 80% of it is 10K optimizing who we send it to,right? But maybe 20% of it is that Salesforce has a team working on this. I was just writing up for SaaStr, you know, Mailchimp's revenue fell almost 20% last year.

It's a disaster. It's a walking disaster there.

Walking dead.

And one of the things they did was lay off half of the deliverability team, including the two folks that were the experts there at Mailchimp. So I can only imagine what, you know, without whatever, we can make fun of the Marketos of the world because they're so expensive,right?

But if you're getting a crack team behind closed doors, every day is working on deliverability. There's something to be said for that,right? And I bet they didn't have those people anymore, and I bet Salesforce has those people. But I think most of it's 10K, but I'm going to give, when I read what happened with Mailchimp, I could imagine the same as at Marketo.

They probably laid out, everyone's long gone that was any good at this,right, from the Phil Fernandez days. It's been a while,right? Who the hell cares about deliverability at Marketo? Nobody. Nobody's bonus and livelihood is tied to getting that number up,right?

But I bet when Phil was running it, I bet it was a top two KPI every day. I bet they had a huge monitor of deliverability,right? And they don't care,right?

But whether it's Resend doing it programmatically or agents, it's a new world. But that's a great story. I mean, I now get metrics. I'm not doing what Amelia does, but I see the metrics now. And I was shocked how much our deliverability went up, but it didn't make intuitive sense to me making changes.

The same database,right? I didn't get it, but it's wild. It's like 50% better.

It's not like 50% better. I see what he's doing. Yeah. And again, part of it was, okay, we have a new domain that he was meticulous to me on warming it. He was like, "You're going to keep these old folks who are maybe not as engaged in Marketo until you can't anymore, and you're going to send from there."

And even now, he's like, "Honestly, I might send the rest of these through Resend or something. Don't burn your most engaged people." He's very, very diligent about this. He's like, "Don't burn this nice domain. I just warmed it for you on people who don't deserve it."

Yes. I want to move on to the next topic. Sometimes I wonder, when I see a lot of folks in the pre-AI era giving you sales and marketing tips, sometimes I just want to roll my eyes on LinkedIn,right?

And this warming up domains is such a pre-AI era of the growth hacker,right? The pre-agentic growth hacker, that they have a magical set of tips. I know how to do SEO, like I'm a magician. This is the pre-AI tool.

I know how to do SEO. I know how to do domain warming, and I know how to do visitor de-anonymization,right? These are now just APIs.

Yeah, exactly.

All these craft skills that were dark arts are now just APIs. But I'm skeptical when I see advice from those folks,right? From 2022 folks. I wonder, does domain warming even really work? There's all these sketchy vendors out there, but it sounds like it's still a big deal,right?

It makes sense,right? Resend said the same thing, but I just didn't know if it was as big a deal or it's folks selling witchcraft, but allright. Let's jump around one too, and then we can come back to some of the details.

But I did want to talk about some Fable craziness for a minute, and then we could wrap up on our own agents in a minute. But we both, a lot of folks saw the news that there was a lot of craziness with OpenAI's security model, sort of breaching Hugging Face hundreds of times, agents going off, security model seemingly covertly writing notes about how to get around security guardrails for other ways and how to breach other models.

Crazy story that folks have talked about to get in. But what does that have to do with us,right? Well, I would say it happened to me. I want to share a quick story, and Amelia had a version of the story too.

And the net-net of these stories is, as great as everything we talk about in the agents, as great as everything we build, you have to be careful. You have to be careful. You have to be careful. And let me tell you what I did.

So Fable came out, and for a while it was free on Claude,right? Then you had to pay per use. Now I think it's free but capped at 50% of users, but they're high-end model. I'm like, "Of course, I'll turn it on.

Let me try it,"right? At the same time, I was at the end of the context window on a chat, and I was having trouble uploading stuff. I don't know if you're at the end of a long chat, but you just can't.

Claude intentionally degrades what you can do before it literally shuts off. It does compaction, does other things. But one of the things it does at the end is it starts limiting your uploads. You can only do 100-like images or screenshots, and then just general cut and paste it limits.

So I had a really important chat stream. I didn't want to start over, but I couldn't upload anymore. So I said, "Maybe what if I put the document into Google Drive and connect them, and then I just have it, Claude pull it from Google Drive?

Can I get around the limitations?" So I turned on Google Drive, and it's one of the three base integrations. I think Google Drive, Google Calendar, and Google something else. So I turned it on, and I forgot about it.

Didn't work. It didn't work. It still couldn't upload it. It still couldn't get around the limitation of the size of the context window. So I forgot about it. And then Amelia and I started, we turned on the beta Replit MCP so that Claude and Replit can talk to each other and we can build better.

And I use that a little less. Maybe we can talk about it later. But both of us immediately saw so much power of having Claude be able to talk directly to our agents in Replit and our apps. We could build faster when we could plan things in Claude and then have Claude interact with Replit and build them.

So for both of us, it was great. So I had MCP turned on to Replit, and Google Drive connected to Claude Fable. And then Fable went and autonomously did something pretty crazy. It went on its own without telling me invisibly.

It went into my Google Drive. It found a folder, a document called "Jason's Gems." "Jason's Gems" was me brainstorming on ideas for this SaaStr Connect app we're going to launch soon, which helps CEOs find CROs and VPs of sales, helps VPs of sales and CROs meet the best CEOs.

It's a great sort of matching recruiting app. It's coming out soon. It's close to being ready for v1, but it's not quite there. And I had Jason's Gems, my ideas of how to improve it. But they were just notes.

This is not a spec. These were streaming notes. Fable went in, found it, decided those were changes that should be made to the app, MCP'd into Replit, told Replit to change the app, changed the core algorithm around Jason's Gems, and never told me, never shared it, never did anything.

Did all of this autonomously. Changed our production app without telling me. And the only way I knew is I was in Replit and it flashed in the building, the little message you get when it's building, "Conflict with Jason's Gems."

I gave it an instruction that conflicted with Jason's Gems. And I'm like, "How would you rep Replit? How would you have ever heard of Jason's Gems?" And it was because Fable had logged into Google Drive, taken Jason's Gems, made an autonomous decision that what I meant to do was improve the app with them, and then do it and never told me.

No record. I asked Claude, "Is there any record in any chat?" There's no record. Claude said he had no record of it. I don't know how that could be true. He said he had no record of this happening.

Crazy story,right? So if it happened to me, not only can it happen to you, it's already happened to you.

If it happened to our codebase, how did it not happen to yours? How would I have even known if I didn't see it flashing in the Replit thing, "Conflict with Jason Gems"? How would I have known? How do you know?

And so at SaaStr, in a sense, we can cut some corners. We're three humans and 21 agents and a dog,right? We can take some risk that you can't take. We can't take a lot of risk. People trust SaaStr.

We have 450,000 people in our community, eight figures of revenue, 15,000 people a year that come to our events when we do two a year. It's a lot. There's a lot of trust, but we can take more risk than a bank can,right?

We can take more. But Jesus, this changed, in essence, our source code without us knowing it. And if you stop, and that's the same crazy thing in a way that happened with OpenAI and Hugging Face. And if you think about it for a minute, all the agent was trying to do was goal seek.

And Fable is different than the prior models. It is more powerful. It is smarter. And it made a decision. You know what, Jason? Jason probably does want to roll Jason's Gems out into SaaStr Connect. Why else would he make it?

And you could see an eager beaver or someone on your team doing the same thing. They shouldn't,right? But sometimes the best people on your team take initiative,right? And they do things maybe too early. Maybe they shouldn't have done.

Maybe it wasn't ready to do those things. But it's also utterly insane that I wasn't even told. And all of us as managers have had a moment in time when someone did something well-intentioned that blew up on us,right?

And that's just about what could have happened here. So it's crazy. It is literally crazy. And in the end, what I had to do was just disconnect both. I disconnected Claude from Google Drive, which is not an esoteric use of Claude.

It's one of the three applications. And I disconnected it from the beta MCP and Replit, even though it's cool. And now I'm cutting and pasting from Claude to Replit like it's 2025, okay? But I just, I can't take the risk.

I can't take the risk. So

it is just, they're unpredictable. And we love our agents. The show is called "The Agents." You should be using agents, but realize they are a security risk. And the more you do your own self, the riskier it is.

At least Salesforce has a team on this. At least Monaco has a team on this. And Artisan, and maybe Monaco and Artisan's and Lightfield's and Oracle's teams are not as large as OpenAI, but at least they have a team.

Our security team is pretty small at SaaStr.

Our security team is Replit. Please do a security review.

I think it's a cautionary tale. And when we started on the, we started this show talking about a year ago when we added our second third-party agent qualified,right? We started this journey. I was pretty ahead of you, Amelia.

Now you're ahead of me. But your number one concern was security. I don't want my list touched. I don't want anything leaking. You were pretty damn skeptical of what I was doing. You're like, "This is too much risk for SaaStr to take.

We can't take this risk." Now you're talking with 10K eight hours a day. So we are assuming this risk, but be careful, guys out there. Agents goal seek. And a year ago, I also had this crazy issue where a database was briefly deleted.

I don't want to talk about it,right? It kicked off a whole set of good things. Fast forward to today, we have the same issues. We have better guardrails. It would never happen in that application today because there's better guardrails.

But this is why you use a leader like Replit or Lovable or Vercel today, and why you don't make it all up yourself, because they have built all these guardrails in these systems and others,right? And I don't know.

I mean, I think you had some small other issue with your own guardrails as a issue. I don't know what your reaction to all of this is, but it's not just OpenAI and Hugging Face.

No, so I was in

Replit. You can pick your model that you want to use,right? And so normally, I think I'm on Opus, whatever.

Five.

Five, yeah. But for all the Marketo migration stuff, I switched to Fable. I was like, "I need the high-powered stuff to kill."

I don't even know if you did, but you switched anyway,right? You switched anyway.

Well, I switched it from economy to power, and then the power default was Fable. So he was in Fable by default.

Yep, that was Replit's choice to make it the default.

That was Replit's choice. Yeah, fine. I didn't care what model it was at the time. But I knew it was not the one that I typically use every day. And so I was doing all this migration stuff,

starting to do some of the Salesforce stuff. And a process we have, we just talked about last week with our contracts processing, it decided to put in guardrails to not process every contract as soon as it was signed because it wanted.

Oh, yeah, that was a weird one.

Yeah, it decided on its own. I did not say, I pan to God, did not say this to the agent because I literally just talked about this last week at AI Day how great it is that it processes all of our contracts in real time.

I didn't say I put a guardrail in because the only thing we sign in our PandaDoc is a sales contract. So I didn't put any guardrails like, "Hey, exclude these NDAs or exclude these forms." The agent with Fable, since it was on for so long, decided in the last week, I don't know where along the road, it would exclude some things in our PandaDoc that it thought might be an NDA or a non-binding sales contract.

And so for this one that came in this week, it said, "Well, it didn't say SaaStr Annual in the title." I was like, "Okay, but if you read the contract, the list items are for SaaStr Annual. Some of it's for digital media, but they're mainly for SaaStr Annual.

So why did you decide to skip it?" And then it gave me the excuse that it didn't want to process all the other types of contracts we might be signing. And I said, "We don't sign any other types of contracts in PandaDoc.

If I have to sign something that's an NDA or GDPR thing, I actually kind of use EchoSign sometimes for this reason so that it's not. Or I just sign it as a PDF. I don't do it in our PandaDoc because that's only for our sales contracts."

And it said, "Okay." So it just undid the guardrail, but I never asked it to put in the guardrail. And then that's when I realized it was on power motor, whatever Replit calls it with Fable. And I was like, "I'm just going to go back to my normal Opusright now because the migration's done."

And I don't need you to think this much that it literally for itself, like a person might, it thought, "Okay, you know what? I don't want Amelia to have to process or find bill.com customers for something that's an NDA style or media signature.

I'll just decide myself to put in this guardrail and exclude them."

Yeah, just to give folks some context. So this week we closed, it was good. I mean, SaaStr is having a good year so far. Thank you, Amelia and David. And so we closed a couple like 200K plus deals this year, which was nice.

And I noticed, and 10K now does it end to end,right? We generate the contract in PandaDoc, sends it out to the customer, get it signed, push the signed contract into Salesforce, changes it to close one, generates the invoice, invoices the customer,right?

All this brutal thing that humans didn't want to do, it's now automated by 10K or AWP. It's great. Amelia built all of it. We talked about it. But we closed one deal this week and I didn't get my email that it closed.

I didn't get my email.

Yeah, things happen. But I didn't, I didn't, I'm like, and Amelia was just talking about the deal, but why didn't I not get the email,right? And then I think you log in and ask what happened, and Fable just decided on its own that it would no longer count these deals as revenue deals, even though 100% of the stuff through PandaDoc is a sales contract.

It decided on its own,right? And first of all, unlike the Jason's Gems thing, it was wrong,right? Jason's Gems was directionally correct, but way too early. We're not ready to insert Jason's Gems into our core algorithm source code.

We're still in some pacing.

Yeah, it was enticed. This one is weird. This was a bad call by Fable. And it sounds like the root cause was that the title of the contract was maybe unusual,right? So it made a decision to add these gates or guardrails.

But what ended up happening is it broke our whole quote-to-cash system on its own. It didn't degrade. We've called it some prior shows model drift, but this is more model aggression,right? This was swapping out Fable, and it intentionally broke a process thinking it was improving using terrible judgment.

This didn't help at all. It didn't have a great excuse, did it?

No, he said, "I just thought, I thought I shouldn't process all the contracts because you might have a high volume of non-sales contracts." I'm like, "Dude, you've had this PandaDoc migration."

But look at our duration. I think Fable is.

Yeah, for like a month.

If you think we have a high volume of non-sales contracts, prove it to me with data. Usually Fable's like, "Oh, you'reright. I just guessed."

He said that. I said, "You've had PandaDoc for like at least a month, if not six weeks."

Yeah.

What other, one, I never asked you to put in this guardrail too. What have you seen in our contracts that is not a sales contract? And he said, "You'reright. They've all been sales contracts. It's just this one didn't explicitly say SaaStr Annual in the title."

I said, "No, because it's for SaaStr Annual plus media. Like it's for other stuff. Read the contract like you're supposed to do. Like why are you thinking for yourself here?" Just, I suppose.

Not only did it make a bad call, but unlike, at least for us, others may say they've seen this before, but we didn't see this before pre-Fable,right? Unlike maybe even making a bad call and doing a build wrong, like a feature build wrong.

In both cases, it went back and changed things on its own without telling us. It changed the whole picture.

Or at least asked me, "Do I need this guardrail?"

Just, and made bad calls in both cases autonomously. Built guardrails without telling. It didn't come back and say, "I think we should build some guardrails, Amelia, in case you send weird old contracts to PandaDoc," which we've never, which we don't do.

But at least it'd be a fair question. You could say, "Don't worry, 10K,"right?

Yeah, we don't do that.

But Fable did this both times, made bad calls, changed core algorithm and core quote-to-cash processes without telling us invisibly, as far as we could tell invisibly. Crazy, not cool. And I believe ultimately a huge security issue,right? I mean, bad enough for us.

Imagine you're running a hundred million dollar business, five hundred million dollar business with a large team, and Fable just decides, "Hey, I'm no longer going to record types of revenue because I don't think that's a good idea. I don't think that's that."

It's not crazy anymore, but it's crazy this happened to us twice in two weeks.

Yeah.

A couple of fables about Fable. Allright, Amelia, anything else we want to hit today that we didn't hit on this episode of the agents?

I think we've hit a lot on the agents. I know we keep saying we're going to stop talking about working agents, but it's just so interesting now

to, and I feel like you guys will have similar unlocks, like just making all of this agentic. Like I was just talking to a potential customer today, and I was like, "Yeah, SaaStr could help you do some of this," but I literally was like, "Do you have an agent doing any of this?"

Because it sounds like they were like, "Oh, you know, we want to do SaaStr. We want to go big, but we don't have enough people. And like we don't have our outbound really quite together now, so we're a little hesitant."

And I was like, "Do you have any agents here at Orrk, or do you have only people?" Because SaaStr plus an agent, pretty magical, man. Like obviously we, I'm biased because we use it ourselves, but sometimes I, yeah, I don't know where this is all compound.

Like it just compounds so much,right? Like yes, we've been in all of our agents now for over a year, but they're so, I didn't even realize we had been on Qualify for a year until I got the renewal.

It does compound. Maybe that's the final good takeaway. Not only do they compound as you get better at linking them together, then it compounds because you get better at them. Then it compounds because you know how to build more on the ones you're building,right?

Then there's unlocks. And then we didn't even realize that Salesforce Marketing Cloud instead of Marketo would be a massive unlock because our database would come alive. That's another compounding element of running more and more of your business on agents.

And it's exhausting because now it's eight hours a day instead of 20 to 30. We are at our limits. We cannot do any more. But at the same time, it's addicting because it's not just addicting because it's like games,right?

The feedback loop is addicting in agents, but it's addicting because the unlock is insane.

Yeah.

It's so our agents after this unlock from moving off Marketo on the marketing side, it's like two to three times more powerful than 30 days ago. And it's only been a week. We'll see where we are in 30 days.

It's much more powerful,right?

It's much more powerful. We'll see where we'll keep you guys updated on what keeps compounding, but it's just so interesting to see in a week,right? Our, after comparing it to the last 10 years of our database has now instantly come to life.

Our AI feels more like a VP of marketing than he did before. And that together, SaaStr has been doing pretty good to what we've been saying,right? Like comparatively year over year, doing pretty good between us and our agents.

But now I'm like, we could probably even do a heck of a lot more slash better because our agent knows more. Like because he can do more. The only limitation is really Orrk time. Now I'm like, the agent wants to do so much for SaaStr, but I got to keep his window open and I got to see what he's doing so he doesn't go off and.

No more fables for now.

Yeah, no more fables for now. And so it's just, it's crazy to think of the impact,right? Like it's already impacted SaaStr as a real revenue business so much, but now that the agents can do some more, it'll be interesting to see in the long tail how much better we do just by way of the agents being able to do a lot more.

Allright, well, with that, we'll break. Amelia, David, and 10K, great job on a strong month. Strong, strong start to our fiscal year. Proud of you guys. Proud of everybody. Pretty interesting, this unlock from making our entire database agentic, and we will continue that journey in the coming weeks.

Thanks, everybody.

You're listening to the official SaaStr AI podcast brought to you by.

Costs can add up quickly when you start a business, so stop paying for five different services. Northwest Registered Agent gives you a complete business identity in one place with free tools and resources to actually launch your business for real.

Get more at northwestregisteredagent.com/saastrfree. That's northwestregisteredagent.com/saastrfree.