All right, so in this video we're going to be going over how to fine tune your agent settings. And when we say agent settings, we're not talking about the entire agent, like the brain and how to train it on your business and all these different things. Uh, we're simply talking about the settings within each flow. So as you see, there's a settings icon, which dials in the settings only for this WhatsApp flow, right? These settings will never carry over into SMS or email because they should have different settings and different styles of engagement and conversation. But if we're going to dial in the settings for this WhatsApp flow, we're going to hit this settings button here. And here we're going to pop things up.

So as you see, we have a few things. We're going to start from the top. So the first thing is the conversation settings with the first question being, why the conversation is happening. So if we hover again, this provides context about what triggered this conversation and its main objective. Examples, what action the user took, what page they were on, what they're trying to accomplish. This helps the AI understand the conversation's purpose and respond accordingly. So for this one, we could say something like the lead opted in to a form to learn more about Agent Kong and book a demo. Short, simple, sweet, anything else you want to maybe add about the conversation specifically feel free to, but sometimes shorter is better, shorter is sweeter, especially with AI. When you give it too much instruction, it just can become confused. Like sometimes less is best with these large language models.

Next, we're going to go to advanced response rules. This is optional, guys. I would recommend leaving it empty to start and then adding later to correct behavior. But here basically you're going to, again, leave empty in most cases, use agent voice for toner personality instead, which we'll get to in a second. Use only when you have specific rules or restrictions, things like never mention you're an AI or a bot because we will have people asking, you know, are you an AI or is this an AI, especially now. Uh funny side note, I have an appointment setter buddy who was calling leads and he got stopped in mid conversation, the lead was like, are you an AI? And he was like taken back for a second. He was like, no, are you an AI, which is a great response.

But people are becoming more skeptical than ever. Are you AI? Are you AI? Are you like, we see it all over the headlines, all over the news, all these video models like especially boomers, older people, you know, people are getting more skeptical. They ask the question. It's a common thing now in 2025 moving forward to ask, are you an AI? So if you never want the AI to mention being a bot, then you'll tell it here, never mention you're a bot or an AI or an agent. Uh, obviously you can take things like no exclamation marks, match the user's conversational style, uh, don't use technical jargon. Uh, maybe another one you could also say is like critical. This is actually a very important one. One thing to mention, the word critical when it comes to AI is like a dog whistle to a dog that just stops perks up the ears, it's like instant immediate attention and like obeying this command. So the word critical again is a dog whistle to a dog. You say critical, never tell the lead the price if they continue to push and persist, always mention how we work with companies that do over 30 million per year who spend over 100 K per month on ads and we also work with companies who make under one mil per year and they barely spend anything on ads. Always defer pricing to the call stating how we can provide a custom proposal and plan based on their needs and their business.

Uh so that is maybe one response, right? Like never ever say the price, always defer back to this, keep pushing that. Uh and yeah, as you slowly start to test the agents more, you might see things there where you're like, yo, I'm not happy with how it responded or maybe you realize I forgot to train it on this. You can always come back here in the advanced response rules and here is where you can dial in some more of the training on how the AI needs to respond. I'm not going to keep talking about that. I think you guys understand what this is. Again, I would recommend leaving it empty to start unless you want something like this, you know, never tell the lead this price or, uh, never say you're an AI or bot, if you get that question, maybe say no, LOL, are you an AI? just to kind of lighten the mood, right? Like you can say things like that.

Uh so these are just again, more advanced response rules. I would highly recommend though, like don't come in here and try to train it and be like, our AI is trained from the ground up on how to perform really well. So you don't need to come in here and start doing a bunch of nonsense. This is only specific rules where you don't want the AI to do certain things. So moving forward, we have the agent voice right out the gate, we give you guys some prepackaged and pre-standardized ones, right? Like mildly flirty, confident, persuasive, charismatic, conversational. I have seen people do things like chill, uh which by the way, if there's nothing in here, you can just go in there, you can start typing chill. You can either hit comma or you can hit enter. I've seen like chill, cool, uh laid back. Right. So you can also type stuff in there as well. And this actually does a big, big difference, like these words, chill, cool, laid back, mild, flirty, like they actually do have a difference in how the AI communicates in its voice.

So feel free to actually play with this until you find it like sweet at the right spot. Uh often times we just leave it empty because AI is pretty good right out the gate, but if you really are looking for a specific voice tone style like this right here would be great for IG DMs. If you're selling, you know, some sort of offer to like mass market, you know, 17, 18, 19, 20 year old kids and you want to just sound chill in the DMs, right? Like chill, cool, laid back. If you do something like this, it'll say like, yeah, bro, totally got you LOL. It'll use words like FR uh so for real FR, maybe in here you can also say things like, use acronyms like FR for real, got you, not much, how about you?

uh things like that, right? Like you can tell it to use acronyms and it'll start using acronyms. So that is pretty much how you can start to dial things like that in. Then we have the reading level. Most of the time guys, you're going to be looking for middle school. The average reading level for people, I think is like fifth grade or something like that. Uh so trying to do, you know, like masters might be off putting, but if you're also talking to doctors or lawyers, maybe you'll want to do like university, right? Like kind of match their tone. Uh most of the time though, you're going to be best just leaving it at middle school. Then we have AI typos, so you can actually add typos to your AI.

Uh so as it generates a response, you'll see typos in there, which is really funny. Uh now what it's at again is 5%. Um so you'll see 5% out of 100 means for, yeah, AI type of frequency per 100 words. So five out of every 100 words here will be a typo. Uh here will be 20 out of every 100 words. Uh 3 to 5 is kind of what we recommend if you want to keep it feeling a bit more human-like that way again, people see a typo and they're like, oh, AI doesn't do typos and you might, you know, flip them off guard real quick.

Uh then we have the message controls. So message controls is very, very important. Uh here the AI will, um, wait. So we actually spoke about this in another video, feel free to go watch it. I think that is the how your agent works just so you can see a live flow. But most of the time guys, I would say like 25, maybe 45, maybe even 60 seconds. And why this is important is let's say someone sends a message and they hit enter, then they start typing another message and it takes them 45 seconds to type the second message and hit enter. The AI is going to wait 60 seconds.

So if we got message number one immediately and then message number two 45 seconds later, the AI will receive both message both messages and then generate its response. But if you only had this at 3 seconds and the person said one message, they started typing out another one and they say another one, the AI will start generating reply to the first one because the second message came after 3 seconds, it will respond to the first one and then when it gets the second one, it will respond to the second one separately, which can create a jarring experience. And it's pretty much like a human, right? Like if if you and me were texting and I texted you one message and you immediately started typing and hit send, but you didn't know I was typing another message and then I typed you another message, you'd be like, oh shit, I didn't see that and you respond to that one. And then I would be like confused a little bit. It creates a jarring experience, which is why we recommend you know, 45, you can be a bit shorter, snappier, 25 but like 45, 60 seconds is pretty much where we I would recommend that you keep it there.

That way you're giving the human enough time to send as many messages as they need and then once they do, the AI will get all the messages within 45 seconds and then respond to anything that came in within 45 seconds. Next, we have the character limit. So this is fairly simple, right? It's the maximum number of characters the AI can use in a single message. 255 characters is recommended for most natural text conversations as it feels more human-like. I believe I imported this out from another guy who was doing IGDM setting. Uh so obviously on IG DMs, you probably don't want to be as verbose and use as many words and characters. So he dropped that down to 175.

If you find that, you know, actually a really, really good point. If we come here and you know, the person says something like, uh, hey Alicia here just to confirm this is Matei, right? And I'm like, Alicia who question mark, uh, I'm married or why are you texting me? And then you tell the AI over here, say something like my boss Matei wanted me to personally reach out and offer you a free AI sales audit on this call. We'll help you audit your sales process and see if you benefit from an AI agent. Can I send you some availability? This is probably cutting it pretty close to 255 characters. So if the AI also has to respond to what I said here, like, no, who is this? I'm married. How did you get my number?

Um, and it has to incorporate this based off your instructions, it's going to cut away some words from here. It might not respond as, uh, as well as it should because it's trying to conserve its character limits. Uh so again, pay attention to that. Like if you find sometimes that the AI just doesn't seem to be responding well and following your instructions well, you can always come up here and you can maybe do like a 455 or like a 500. Important to note that this is not, as right there, you see, note, AI can write shorter when needed. This is a maximum, not a target. So you don't have to worry, it's not like it's going to always generate 500 character responses. Uh just something to keep in mind is the character limit does have an effect on how well the AI responds and follows instructions if you're telling it to say certain things here and also it's trying to respond here.

Uh so that you can always come back and fine tune over the longer run. If you're not too, too happy with it, you can always increase the character limits there. Uh up next, we are going to have the message bubbles. So message bubbles, you can turn them off if you so desire. If you turn them on, then the AI is going to break longer text into smaller human-like text bubbles. So quite literally like you are a human. All right, so here is an example of the message bubbles, right? Got it. Whatever, he personally reach out, there's no setup fees, retainers or ad spend. Would you be open for a quick call? Right. So it's sending it in multiple messages, which is exactly what message bubbles does. So right out the gate, it's going to come to 60 characters per bubble.

Uh that is kind of what we recommend. Minimum size is obviously 30. That way, if you tell it, you know, like five, which I hope, uh, you shouldn't, but I think we actually limited it to 30. If you tell it five characters, you're like, it's going to say one word and it's going to be really, really weird. Uh right out the gate though, most of the time 60 is good. And then you have bubble delay. So you obviously don't want to send four bubbles all in one shot. So this is where the bubble delay comes in. So it'll send the first message, then it'll wait a minimum of two seconds and a maximum of six seconds and then send the next bubble.

This kind of varies, so it's never going to be 3 3 3, it'll be like, you know, 2, 4, 5, 6, 3, 2 and it kind of just jumps in between it. Um and then it will send each bubble. That way it mimics real life human style and feel where they're going to get a bubble, they're going to wait a few seconds, get another bubble. This is again, standard. I believe every time you open a flow, this is turned on and it's set up with these settings. If you do want to change them, feel free to. I would say this is probably going to be the one that you change the most, like maybe you want shorter and or longer bubbles, maybe you'll go like 90 characters limit per bubble or 100. But that is the message bubbles.

Up next, we have the work schedule. You can turn this on and you can tell the AI it's only allowed to work during these hours. Maybe something like 7:00 a.m. to maybe like an 11 30 pm and then the time zone, let's just say it's um Eastern. So Eastern time. So now your AI will only work. So if it gets a message at 12 or 1:00 a.m., it's not going to respond. It's only going to respond again at 7:00 in the morning. So that is the working schedule.

Up next, we have the follow up cadence. So this one is actually pretty cool. Uh as you see here, there's four follow ups. I'll just go ahead and delete all these so you can see. Uh it pretty much I think will come empty out the gate. So if you want the AI to follow up, you can add follow ups. So maybe we're going to do one day, one day, one day, then maybe one day, we'll probably do uh two hours the first one, then we'll go 4 hours, then maybe one day and then one day, cool. And then maybe we'll go two days, two days, two days, two days.

So what do we have? We have 2, 4, 6. And then one more follow up eight days and then maybe we'll do, you know, a few more and we'll kind of leave this one week apart. So he hasn't responded once a week, we'll then hit them up. So now when the AI, basically the way this works is we count it from the last message the agent replied or sent. So that becomes like a flag in the timeline. So if the AI sends a message now and it doesn't hear back from the lead within two hours, it will then generate a follow-up. Again, we've trained it on how to follow up. Uh it's never going to say, hey, just following up, it'll say things like, hey, just between meetings had a quick minute or hey, I had a quick minute, I wanted to see how you're doing.

Um so we teach it how to do the follow ups properly and how to continue pursuing the goal that it's currently trying to pursue, whether that's booking the appointment, asking for availability, um whatever your goal is. So that's how that works. So wait two hours. If it still doesn't receive a message after this one within 4 hours, it will send another one. If after one day from sending this message, it doesn't, it will send this one, and so on and so forth. You can also add prompts in here. So again, the AI will automatically generate a response based on the conversation so far, optionally you can provide specific instructions.

Honestly, we don't see too, too many people using this. We've added it just as an additional Lego piece. Like if you want to go balls to the wall and you want to really fine tune the way this thing responds and feel free to 9.9 out of 10 times is going to be perfect just the way it is. So I wouldn't really play with that. And that is the follow-up sequences. Then moving forward, we have the different follow-up modes, which is very important right out the gate. It's going to be continue and end. So it runs through your follow-up sequence once and then stops.

So it will run through your follow-up sequence, which this is the sequence, basically all these cards, uh it'll run through it once and then it will stop the second one is infinite mode. So it will go through this entire sequence and then it will, if still hasn't responded back, it will then come back and do 2 hours, 4 hours, 1 day, one day, one day, so it's like an infinite follow-up mode. And then restart on ghost allows your agent to basically always come back after it gets ghosted. So let's say it sends a message, it doesn't respond. It waits two hours, sends a message and 4, then on the third one after one day, it sends a message.

Um then it kind of resets, it comes back here, it'll hold the conversation. If the lead ghosts again and it doesn't respond, it'll wait two hours, sends a message and 4, 1 day, one day. So that's kind of the difference between continue and end versus restart. Restart only restarts it um once the person reaches back out and then ghosts, whereas continue or sorry, infinite mode, uh it'll just go through the whole thing all over and all over again. So that's follow-up mode. Then we have the smart features that are mainly always on.

Uh so smart milestone detection is this, it will automatically detect when users have provided info for upcoming questions. So for example, if we are here and we want to, I I can't think of a flow on this one, but let's just say we have here, let's go ahead and let's edit, um and we have this one where it's phone, if we already have their phone number and we want to skip it by the time it gets here and if this says like ask the user for their phone number, but we already have their phone number in the CRM, then it will actually skip this step so you don't have to worry about that.

So that is again like the smart milestone. If they've already given us like an answer to this one ahead of time, it'll skip it. Um then we go into the smart follow-up. So if someone says, hey, I'm busy, can you follow up with me tomorrow at 7? The AI will create its own follow up and it will schedule that follow up and follow up to them tomorrow at 7, like the human requested. So it has a nice smart follow up detection where it will create its own follow ups and it will schedule the follow up accordingly. Uh and then we have off topic redirection always on because you don't want your agent to get into, you know, Trump versus Biden politics or you know, like Ukraine politics or, you know, Illuminati stuff or like whatever weird conversations people might be throwing at it like you don't really want it to do that. So we can always turn this well we leave it on standard, but when this is on, if the conversation goes off topic compared to its goal and what it's supposed to do, uh it will just generate a nice response and it will gently steer the users back towards the relevant discussion. But that is going to be it for the conversation settings. I'm not going to continue to beat this one. I think we've gotten everything down. If you guys do have any questions, just jump in on Discord, ask our support team. We're more than happy to be here and assist.