Stop Learning AI. Start Doing AI. The 20+ Agents Running SaaStr

1. Key Themes

Training AI Agents is More Important Than Vendor Selection

The success of AI agents depends more on dedicated training than choosing the "perfect" vendor. Jason emphasizes: "Pick a leading vendor, for sure, it does matter, but you training is more important. Training is more important. There are very few agentic products...where you wanted to do something like a qualified, interact with customers, speak with authority, close deals...it wouldn't happen if we put zero minutes of training into it." [00:00:24]

The old enterprise software model is dead. "You think about how people bought enterprise software before AI. You would buy it from a sales rep, then you might hire an Accenture or a Blue Wolf to deploy it over like a year...No one will tolerate that in AI this year. Time to value actually often has to be even before you get a contract signed." [00:00:31]

The Stair-Step Approach: Start Simple, Build Confidence

Saastr went from zero agents to 20+ in 2025 by starting horizontal and going vertical. Jason explains: "We went from Delphi, a horizontal agent, which basically can self-ingest your content with no training...Start easy, start broad, get a few, get a win under your belt. Even if you started with say qualified...Start for the simplest use case. Because if you fail with your first agentic thing, you not only will it, is it frustrating, you won't know." [00:15:06]

They deployed agents in this order: generalist advice agent (Delphi) → outbound SDR → inbound qualification → support. Amelia notes the outbound was easier because "it was less risky to send a better email than a mediocre human than to trust an agent with a precious lead." [00:15:50]

Traditional SaaS Budgets Are Frozen, AI Budgets Are Exploding

There's a massive budget shift happening. Jason states: "Traditional budgets are frozen. CIOs are still going around the table at the end of the year and telling every functional head to cut some SaaS apps...half of our incremental budgets go into price increases. Salesforce raises prices 8% this year, 6%. That compounds. So if I'm growing my IT budget 6%, and my core vendors are all raising prices 7-8%, how much room is there for another business process workflow app?" [00:12:53]

But: "There's so much AI budget. And when you add it together, business software is growing faster than it has ever before. But if you don't tap into that AI budget, it's tough." [00:28:52]

2. Contrarian Perspectives

20% Agent Performance is Better Than Zero

Most people dismiss agents that can't match human performance, but Jason challenges this: "Here's the thing. And people get this wrong. 20% is a lot better than zero. Like if we never used qualified and only used this general agent, we wouldn't have the seven point because it can't set meetings. But even 20% is better than zero if it's good. Just the fact that instead of it being dope, like we used to use inner common, it would be a weak response time. So we switched to our agent and it could instantly tell you about 90% right your customer question." [00:12:42]

Don't Learn AI - DO AI

When a respected CRO reached out saying "I need to learn AI, I'll do anything, I'll intern for you," Jason's response was radical: "Like no, you need to do AI. Don't learn it. Buying a subscription to Claude does not count as learning AI. The way you're going to do it is pick an agent in an agentic product. Pick the simplest possible one use case...Deploy yourself, buy it, set it up...Own the deployment...If you deploy an agent yourself, train it, QA it, test it. You will be ahead of 90% of the world." [00:39:05]

He contrasts this with typical CMO approaches: "We encourage our team to use any AI tool they want. Like, we allow them to try perplexity if they want. And we encourage them to learn recipes online. I'm like, you will never learn a go-to motion in today's. You got to do it." [00:40:04]

Halloween 2025 is the Deadline for Startups

Jason sets a hard deadline that most would consider extreme: "For a start up, I said Halloween. Halloween is the very latest...Because you're going to get out competed if you're a start up...I tell you, maybe it's me, every VC I know, every investor I know, where the start up hasn't made the jump yet, they've given up. They've lost hope...If you're not in market with a disruptive AI agent as a startup on Halloween, you need a brand new team...Maybe even 80% of them got to go." [00:30:52]

His reasoning: "Everything's great now...even mediocre software just got better when you put the fours in...Cloud four, you take the same...I bet you it's a five times better product just changing the LLM." [00:31:21]

3. Companies Identified

Qualified

AI-powered sales qualification and meeting booking platform. Jason reports: "Amelia was telling me last week, qualified set up seven appointments for us...On its own." [00:01:41] Later: "I literally go agent to agent, like I'll go qualified...It's been selling tickets to our events. Like I can, it's done...probably done a hundred tickets to our London event...A hundred out of like 2000 is not nothing." [00:18:40]

Delphi

Horizontal AI clone platform. Amelia explains: "We started with a platform that's called Delphi and it's a clone of Jason. So it is all of your advice, your social media, everything on Saastr for the last 12, 13 years in the clone...20 million words of content, a thousand YouTube's, all the tweets, all the LinkedIn's." [00:10:11] This was their first agent deployment that gave them confidence to expand.

Replit

AI-powered coding platform. Jason notes its explosive growth: "Replit took 10 years to get to a million...They add Cloud four, now they're almost 200 million in one year, but it's the same IDE...The same software." [00:31:55] He personally invested "probably 200 hours expert now to make it great" [00:06:01] after 100 days of use.

Artisan

AI SDR platform used by Saastr. Amelia mentions: "I literally go agent to agent, like I'll go qualified, I'll go artisan, I'll go Delphi." [00:18:40] They use it for outbound sales agent work.

Momentum

Sales meeting recorder and go-to-market tool. Amelia describes: "Momentum, which is like a bit of a recorder and a complementary go-to-market tool to qualified...it will slack you a summary after each sales call and it'll also slack me anytime like a new opportunity is opened or meetings." [00:19:41] She notes it exposed non-performers: "This wrap was like, I have no activity. So like it literally exposes that." [00:20:01]

4. Operating Insights

Deploy Agents in Areas Where Nothing is Getting Done

Jason advises: "The layup are the things where there is a tool that exists that is trustworthy, where it's just not getting done in your work...There are companies that are very sales-centric organizations. There's companies that are very product-centric. If you like hate sales people and you like product, maybe you should start with agentic sales...Don't try to replace the stuff that's sort of working. Look at where there's literally no one doing the work." [00:22:31]

Agent Management Requires Daily Cognitive Investment

Amelia describes her routine: "I spend an hour with all of our agents in the morning...it used to be set up now. I check on all of them. I see if there's any follow ups we need to do...I literally go agent to agent." [00:18:40]

Jason clarifies the nature of the work: "It is a higher cognitive level. You have to think more. You don't spend time arguing with people. As I say with Amelia, the agents don't cry...that is exhausting, right? But it's exhausting in a different way...you don't have to use all your brain cells for an hour to sit with someone crying, but you do have to think. You might have to think about how to make that agent work better. Eight hours a day instead of three hours a day." [00:21:01]

Don't Make Hero Purchases - Solve Broken Problems First

Jason warns: "The failure that I see, especially CMOs make, is they're looking for a hero purchase. They want to go to their boss to the CEO or whoever and say, I bought AI...I guarantee it's probably going to fail because that hero purchase is probably going to be aligned with something you're decent at. The bar is so high and you're not going to know how to train it because it's your first agent." [00:25:25]

5. Overlooked Insights

AI Agents as International Time Zone Coverage

Amelia mentions almost casually but this is hugely significant: "Half the time right now our next events in London. So half the time I wake up and I spend this hour with our agents, mostly conversations for London have already happened because when we wake up in Pacific time, it's their afternoon, it's their evening." [00:35:24]

This represents a massive operational advantage that most companies haven't considered - agents don't just work 24/7, they provide native-quality engagement in every timezone simultaneously. For any company with international customers, this could be transformative for sales coverage and response times.

Agents Expose Bad Data and Bad Performers Immediately

Amelia reveals: "I thought we had okay data quality in Salesforce and I was like, no, it's terrible. But it exposes that." [00:19:30] More dramatically: "When we hooked up momentum...we had somebody leave our team that day because...this wrap was like, I have no activity. So like it literally exposes that like what are you going to do? You have to quit that day. You can't hide. You're going to be found out instantly." [00:19:46]

This is a critical insight barely touched on - AI agents don't just replace or augment work, they create unprecedented transparency that will force organizational reckoning with data quality and performance issues that were previously hidden or tolerated.