SaaStr's 10+ Top AI Agents: AI SDR, AI BDR, AI RevOps + More: The How, The Who, The Why

Jason M. Lemkin walks through how SaaStr built 11 core AI agents for sales and outreach. The setup takes 2-3 weeks to warm up, ingests historical data, auto-generates messages to a pending outbox for manual review, and maps everything back to Salesforce. They cover build vs buy tradeoffs and what actually works in production.
Chapters
Welcome everyone to granted hope you're yourself.
We'Re streaming we're here with us to.
Everybody what we're gonna do with the slides but we're gonna do something a lot of folks ask for which is on what we're doing how we're using AI catalog diet we've got as Adam will show you we've got about 11 core agencies 20 total we're going to dive into like the six please those and the six are probably most interesting you guys for AI just these every other show you the actual data give a little bit of an overview that can dive in on most of them and I'll help most of them we buy and we'll show you who buys and then we actually put some ourselves on and we'll be diving that but for your questions especially at BERS the biggest takeaway I can tell you I don't put in the deck is there customers in the mood boarding we're number one because the swap but most of them things point real ready but the agents there's no set forget about the nose and forget and just for. None of these names are as putting in their apps in the work and show you the actual game for sure but you could never do those training deeply and doing it every day every day so many AI fillers from you know late 2024 early 25 that you see all. And they're still monitoring this session and fixing things as often as wrong and they don't commit that you can now extremely be some of the failures do you really feel like just someone that he did and there's a proper train or there's no data to just knowledge to that formal.
I am really quickly playing a link to this full in the chat there's also word here in chat. Okay the first one we'll go through is our AI we did a full deep dive on how to do AI specifically. Most recently just now we had a full employee for you know 100 so how we use it in our view as a form so I have trained it on in different ways so the way it's visible is because it can do so in the last hundred days it's done it's always set out do a prop here to take it to business from our market specifically any secret is talking about what they do and support training with AI agent is what have we set up initially to install let's say we want every email to run out value to their audience we want to always straight into it and say you know how learnerville Data is faster but we've ingested the most relative data. It also looks like we're doing a website it looks very personally we basically customize more on them versus. One more you can get out to them the better results you see in.
A how we this board of sponsors and tend how do we pull historical data put how does that agree with.
Yeah so there's a couple ways you can do it. You can pull in some automatic signals so you can do directing the SAS company I do a grass pull from iron or hill combine mainly on and you just throw up two and what we'll do for America take some time the first time you set up anything late so any AI 2 to 3 weeks to warm up that's common so keep that in mind I don't have time but all need time to warm up and to work your mailboxes. So all these you don't want you want to use you want to use so every AI. It will take another if you're already probably freeing liberty you to do that. You know we have what is it all that they acting at once to make sure the right person and then it will start to generate messages like you see here on this slide. So it starts generating messages it will put them into your pending out box.
How many messages did you manual before you let the AI do it up.
I was saying the beginning of read old messages so it was hundreds and thousands like more people you start to just start a huge. Does that and people and then again I read because you can set the agent to say how many messages you want to do so for us historically we always feel like we want to email after an email from E or a human after you are taking about three actual human touches so to look at it and our z said okay three touches in email and one so 604 your 800 I would read all of them literally now I just. How do we want to respond to them that is a bit more manual they can turn it on so it's like automated responses But I still like the.
For we've had probably 50,000 people over day come to our in front of and have hundreds of where how you and add to text of the AI star are you you manually creating files and manually how did that whole workflow used for laughter reactivation?
Yeah it's a good question. So there's a few things it does I will actually show you guys the back and forth yeah full screen. I'll pull my guy over one second. So let's see can you guys see this tab okay audiences are now I'm just checking honestly those weeks sometimes because sometimes they'll pull Seekers or secret manuals that didn't include any secrets for London. But yeah part of human trained they are biggest friend has been and yeah I was my time now in the inbox so this is where I can see who's engaged with us. I can see what sent back to and then you can go through the ping message here to all of our social messages to My goodness. We got more so yeah I mean I do think we have it's pretty high. I think this one's 7% that one's high but you know.
If it's adding value helping them.
Yeah. It's just that you're getting those maybe two, three times a month and then now also getting sequence where you're a high for me also to your point like how we sending data and anyone feel that I want to say something specific to this person connecting to them and works a little bit more you can find it and then the other thing you can do is you can upload a do not send list. So how we use it you want to drag customers for this. When you get customers you can say copy you stop emailing everybody there it's good. So then you have a manual to that like who wants but.
Our salesforce is our website, right?
Correct. Not today AP's training. It's all going to one of our campaigns here. Okay. And this is why it's also again responses than I have for our sponsor for code not really any flatlines. It's fairly low so I a little bit more hard on this. But here's a. You see, you can see it's like you could say those names now don't contact them. We might do one the pitch where you can do that. Right. So this is where I spend a lot of time saying okay this is the offer and here the pay points is very like the seekers and you can just keep additives so it won't show automatic today. You know ask for versions that's coming by the year or now you put it in yourself. You know pick the pick the ass paypoint of brew points I can use. Just because it was better. But you can. You can do that. Cool. Before I go on to the next one there's a question about you guys like to use arts solution to like I delay on build versus buy right. Like do you consider I have a lot of use and the overcoming sessions like I use for different use cases like I haven't used it in this way for specifically an outbound aicr I just find even the cost of saying they bought all mint boxes that we have in our they. Just did not have like the patience to say okay I can't read up. They were at a loss. For data. We have all of our sponsors in them. Yeah.
I would say one thing is to just go we're not saying these are the work for us. That's what we're saying. There is no maps that you could on your own right. If you and there are we don't have other mindsets. We have lots of engineers we tools and we have a lot of data. Okay we have more data. We have three by three where we tools can be honest and they fall with us or they wouldn't help us. They said couldn't use or didn't want to do with our app we want to do so there's just There's a classic so source that the first try to figure out who's using what. But since we talk to them do they work the way you want to work? Right. For example Clay there's a massive ecosystem of agency play right now for us. You don't need $100,000 agency that that helps we do but that might be a huge ban. So who wants to highest down in the world is HD. So if you're super air you more than we have right. We smart with log where vendor would Help us and 1 Tools need a engineer a solution agency. Okay I know why your own team can do this with engineering almost everyone needs help and devices type can vendor provide you a type and I almost pick that vendor the one that you have the most confident you combo on not the smooth smoothest deck processor they all make.
But I think the way they done it so far is that like we want our AI to be successful. We want work we want to see the responses and outlets that we've seen in use case for our right so we use our.










