AI Reply Agents: Handle conversations, and move deals forward automatically

15 min With Alex Gopshtein Section 6 Lesson 36 Next

Every reply sitting unanswered in your outreach inbox is a backlink deal going cold. This lesson shows how Pitchbox AI Agents read each inbound email, send a reply within a natural 5 to 20 minute window, negotiate placement pricing against limits you set, and chase up until the goal is met. When the agent is unsure, it stops and flags the email for your team instead of guessing.

In this video, I'll show you how Pitchbox's AI agent handles every inbound email, responds to them promptly and drives every opportunity in your campaign to its goals. Before we set up AI agent in Pitchbox, we need to understand the two main components of the agent system in Pitchbox. It's comprised of AI agent, think of it as the brain. It's the employee that's reading your email and decides how to respond to an opportunity. The AI playbook is the delivery system. It's the SOP of what are we going to do after we've read the email and generated a reply?

Do we just save it as a draft and leave for your team to review before they send? Or do we want the entire system to be handled on its own? Generate the email, schedule the email, get the email out. Now let's see how to set one up. First we want to identify which campaign we want to set up AI reply agent on. So once that's marked as not configured, that's a good start. Here I'm presented with the two step process of setting up our reply agent. One is AI reply agent, second one is the playbook. We've just seen it and that's our next step. So you have number of AI reply agent preconfigured in your system.

You can create new ones or configure the ones that is already given to you. So I'm going to choose guest post and also for the playbook, since it's the delivery system I'd like Pitchbox to handle it all for me, so I'm going to say send and chase. Click continue. I can dive a little bit deeper in understanding what is going to happen. So AI agent automatically selected my AI reply for guest posting because that's the goal of my campaign. And then I have my playbook. Remember AI playbook says how, when and whether replies go out or not. So the condition drives that decision.

So I don't want AI agent to handle any of my communications if there are these milestones. But you can configure and add more conditions as you desire. I'd like Pitchbox to handle my replies automatically. I'm going to give it five minutes to 20 minutes wait time, so it looks more natural. I have my adaptive AI chase up already configured, so click on save and activate. But you might say, okay, the agent's configured, but how will it respond? What will it say? Right. So we could do that by previewing the guest post AI agent. This one we've used since Pitchbox has already given it to us.

It's already predefined for me, but I have the ability to edit it or create an absolutely new AI agent based on my settings. But just to give you an example, AI campaign goal for this one since it's a guest post, is instructing the AI to get us to the goal of getting a guest post accepted, basically saying have an agreement with the pricing that we've defined, an acceptable pricing to us. Under instructions, we can tell AI its tone, topics, writing samples in case those are requested, guidelines, payment, so on and so forth. Whatever you would expect or typically have in your daily outreach activities when you run your guest post, whatever they are, you tell AI how you want them handled.

I'll give you an example under the payment. So if the payment is asked, then I can basically say calculate the range between 30% to 45% of an asking price and present a counter offer. I could also say that I don't want to pay any more than a specific dollar amount. You can pass some custom variables in here as well so that you can actually drive this negotiation based on the information from the particular opportunity. As we scroll down, we have an exit criteria, which is extremely important because we want to let AI know that you're going to reach your goal.

And at that point, I'd like you to stop the conversation. At this point, I will consider this opportunity as a win. Same applies to negative. If somebody said, I'm not interested or you're not able to negotiate a particular price, you could also tell AI agent to say, look, at this point, you can stop and mark that opportunity as lost. So on the positive side, I could basically define what I see as a positive outcome, like for example, they've agreed on a price, either accepted or negotiated or we've accepted theirs. And then what I want to happen is set the milestone to almost there, and I can add some tags on a negative side.

I want to mark it as lost and I can add some tags. Tags could be to let you know that it was managed by AI. You could add a tag AI lost or AI won or AI negotiated. It is completely up to you. If I have emails in my inbox already, which I do, and I have configured the AI agent and AI playbook on my campaign, I would want AI agent to handle them, but they're already here, they've been already responded to, but AI agent wasn't configured at that time, because AI agent will start working for everything moving forward. So any emails you have in the inbox, they haven't been handled by AI.

But that's easy. I can select all of them and I can click on run AI reply agent. And reply agent will handle those emails. Now, we want to wait a minute to give AI agent a little bit of time to process and decide whether it can or cannot send. So you could see three of them just disappeared. That means they went over into the agent flow. If I go over to the agent flow, I can see that, hey, we've just responded to those emails and it says milestone is Got a Reply. Now it's in the chasing. These are just now. This has just happened. I'd like to see what happened actually, right?

How did they respond? So I'm going to head over to the opportunity detail and I can see that my agent actually responded and it's automatically turned on chasing, which means that if Melissa does not respond to me, then Pitchbox will automatically send follow-up emails until we reach our goal. Now let's go back to the agent flow and take a look at one that actually has reached the goal. So I can see that this particular one, agent was executed, draft was created, email was sent, we were in chasing, and the goal was met, which means that this is considered to be a win.

From the AI agent's perspective, it has reached. It reached the goal, meaning that the exit criteria was met. Let's take a look. So we see that AI agent reply says goal was reached and it tells us why the exit criteria was met. It says that our target confirmed that she wants to move forward and will publish and it actually gives us a blurb from the email how AI has decided. And it says, yeah, sounds great, let's move forward. So our target has accepted it. Our agent actually was instructed to send a sign-off email. So it did send that last email says fantastic, really great to hear that.

And our opportunity was marked as almost there. At this point, it is up to your team to pick this up. And since it's a guest post, a content needs to be written and submitted. In the case of link inserts or niche edits, you can actually use AI agent to drive it even further to a point where the actual link has been placed and mark that opportunity as a win. Now let's go back to our inbox because when I instructed AI agent to run the reply, it actually came back and said there was one failed. Let's take a look why that happened. When I go to opportunity detail, it tells me that it's skipped.

AI reply was skipped under the guest post agent. Why is that? And it tells us the skip reason. It says that it was asked about pricing and content topics. The user interactions require referencing specific topic ideas from the thread opportunity details. But the personalization fields such as the name and the blog post were empty. And therefore it was impossible for AI agent to respond. I wanted to point this out because everyone's afraid of AI responding for an email and making mistakes. Pitchbox is no different. It is still possible. But we've taken extreme measures to not send an email, not to respond to opportunities if the agent is unsure.

And this is a great example of this where the email will end up in the inbox with a tag that the AI reply has failed for you to handle it. In some cases, it might be that the instructions didn't have an answer to a specific question that the target asked. And it actually will provide you with suggestions of how to edit your agent to respond to a message next time around. Up to this point, we've been utilizing Pitchbox's AI agents. But how do we create our own? That's easy. We click on AI agents in the top right corner. And we have a list of agents that already come preconfigured in your account.

Now I'm going to click on new AI agent. And of course I can start from scratch. But as all other AI functionality in Pitchbox, we give you a library for you to choose from. So AI prompt library, we have chase ups and we have AI replies. We're going to go back to the ones that we've done before, which is guest post negotiation. Kind of similar to what we've worked on so far. So again, we can tune this, the campaign goal. But what I wanted to show you mostly is the instructions. Now the ones, prompts that are live in the library, they try to help you make every single agent personalized to your goals.

So here's an example. If they express interest and ask about topics, we can write about the following topics. And it basically says, go ahead and list those topics. Now I'm going to delete this and move this over and then I'll say list the topics. So let's say I am a specialist in fitness equipment. I could also say weight loss. I can also add any other topics that I specialize in. And therefore AI will actually generate responses based on the topics. Now in this case, obviously I would not want to call it just a guest post negotiation in general.

I'm probably going to call this guest post and I will call it fitness. Because now my instructions are very tuned to that specific goal. Payment, the same thing here. I can say I'm not going to pay any more than $50. Or if your policy is no payment for links whatsoever, you can adjust the payment and say, if they ask for a payment, we just let them know we don't pay. Now we're going to decide whether we want to have milestone set to negotiation, whether we want to add tags or not. And then how do we know this is going to work? So I click on preview and it's going to pull an email in my inbox.

And it's going to try to respond to that email. And this can be an example. There you go. Happy to share the details on topics. We write about this, this, and this. This is actually the email that was left behind in the inbox since it didn't have any information around the price and topics. And now we've given it, it pulled that particular email up. And now showing us, hey, I can actually respond. Now, obviously every single run of the AI might generate a slightly different email. I'm going to click on create. I can put it in specific categories.

You can manage that as well. And now I have my agent set up, which is guest post fitness. And now from here, I can apply this to other campaigns in my project. So I select a playbook because I need that. And remember, there's a two-step process to that. Select the project, link building multi-niche. And now I can select the campaigns that I have and apply this particular new AI agent to those campaigns. Now that I have my guest post fitness AI agent set up, I'd like to set up a custom AI playbook for it. So I'm gonna head up new AI playbook. I'm gonna give this a name and I'll call it Alex Fitness.

Under condition as we've seen previously, I like to make sure that AI agent doesn't touch my opportunities. If they're almost there, won or lost, that is my preference. You can add more conditions if you'd like. For the sending emails, yes, I would like to send it, not just draft so I can handle it myself. No, I'd like AI to send it. I want to wait between five minutes and I'm gonna tighten the window up a little bit. From five to 15 minutes, I'll wait, so as soon as my email is received, I'll wait five to 15 minutes and then I'll send the email.

And then for chase-ups, I'll schedule the chase-up as well. Meaning that if I responded or AI agent responded on my behalf, and there is no response from my target, I want to make sure that the agent keeps pinging them until I do get a response and we can continue the conversation forward. It's pretty simple. I like my adaptive AI chase-up. You can configure chase-up agents as well. Time of day, I'll let AI decide what is the best time of day to send that chase-up. Save playbook and Alex Fitness playbook is done. Now that you've seen how easy it is to set up AI agent in Pitchbox, go ahead and set one up for your campaigns and let AI handle your responses and your link building while you sleep.