Why Sales Automation Just Leveled Up: Inside OpenAI’s AgentKit

Stop asking AI for answers and start letting it act. We break down how OpenAI’s Agent Kit flips sales from reactive prompts to proactive, multi-step workflows that coordinate CRM, email, calendar, and call intelligence like an orchestra under a single baton. I walk through what launched, why it’s different from traditional automation, and how a visual Agent Builder, Chat Kit for human-in-the-loop feedback, and built-in evals make it practical for frontline reps—not just ops and engineering—to build production-ready agents in minutes.

We get specific with three high-impact use cases you can clone: an inbound orchestrator that checks Salesforce, enriches with public insights, drafts a reply, and summarizes in Slack; a deal health sentry that monitors engagement, multithreading, and silence windows, then escalates with context; and a proposal producer that pulls your last call, maps to your statement-of-work template, asks clarifying questions, and outputs a Google Doc v1. Along the way, I share a story from my early career about finding conversion “quiet pockets” and how the same insight now takes thirty minutes to automate end-to-end.

There are guardrails: keep the human touch for tone and senior outreach, expect occasional connector changes, and invest time upfront to define quality. But the upside is real and immediate—sellers who learn agents can 2–3x output and build compounding advantage. If you’ve ever felt buried in follow-ups, prep, and proposals, this is your way out through orchestration, not hustle theater. Subscribe, share this with your team, and tell me the first workflow you’ll automate. And if you want hands-on help, join our AI certification and leave with working agents and a repeatable playbook.

Chapters

• 0:00 Bots Versus True Agents
• 0:40 OpenAI Dev Day Breakthroughs
• 2:13 Why Agent Kit Matters for Sellers
• 3:35 The Agent Builder, Chat Kit, and Evals
• 6:31 From Prompting to Acting: The Big Shift
• 7:13 Jake’s Data Hack Story
• 9:29 Three High-Impact Agent Use Cases
• 12:05 Guardrails: Human Touch and Maintenance
• 14:01 First-Mover Advantage and Next Steps
• 15:35 Closing, Resources, and Certification Invite

Transcript

SPEAKER_00: 0:00

Many sales bots do one thing. They might reply. They might schedule. But with an agent, you can build a flow. So you can say, hey, if somebody who is a prospect in Salesforce emails, the agent can then check the CRM. It could one, obviously, draft a write-up. It could escalate something. You could say, hey, Jake, you should probably send this to your leader. It could immediately enrich with a third-party trend. You could create an agent that says, Hey, if a prospect emails me, I want you to then go and find three publicly available insights about that company that are relevant to what we do, and then feed that back to me in Slack. You could do that today. AI Powered Seller. What is going on, everybody? Welcome to a very special edition of the AI Powered Seller. I am your host, Jake Dunlap. Really excited for you all to join. Very, very big news on Monday. If you did not see it, um OpenAI in their dev day. They actually announced this dev day uh quite a bit, I think back in June, maybe. Um but this dev day turned out to be a lot more than I think what a lot of people thought was happening. And yeah, I like to describe what OpenAI rolled out. It's think about like you're a conductor of an orchestra. Okay, there's hundreds of instruments, they're all playing in time, all contributing to this beautiful symphony. Now imagine that orchestra is your sales org and your data systems, your sales tools, your CRM, quoting, follow-up, cross-functional customer support. Uh all of those things are now humming together. That the person conducting can be you as an individual seller. It can be you as a sales leader. And this whole concept of being able to bring all these systems together in a way that wasn't this like data lake of this and this, and oh my gosh, we've got to spend you know two years to clean up our data, seemed impossible. But on Monday, it became possible. And within one tool. Now, you know, for a lot of you who might be more advanced, there's there's tools, workflow automation, workflow connecting tools like uh N8N and Make and Zapier, even uh Workado, like you name it. These tools have been around for a little bit. But but there's something really special about Agent Kit, which is what was announced and launched on Monday that I think is going to have a massive impact on all sellers out there. Um, it is also really exciting. So we have our AI certification program jumping off. Make sure to sign up and teach you how to build in the new tool as well, too, or make if you have another tool that you want to use, we can help you there. Uh, but but this to me was really a wake-up call for anybody in sales. These tools are becoming easy, they're becoming so integrated that not using them is almost, you know, it's like I'm I'm gonna keep using the yellow pages when the internet exists, or I'm gonna keep reading books when I can go to a search engine. I really feel like this concept of what goes into being an innovative seller and an AI-powered sale seller has really changed. And for all my Chat GPT users, it's changed for you in a very big way. So let me start with the gist. Okay. Agent Kit, okay, is their answer. It's a complete set of building blocks that's gonna help you tape agents uh from prototype to production in this demo. It's pretty crazy. We'll link to the demo as well. They built two automations in eight minutes. They connected multiple systems and built autonomous agents in eight minutes. And that's why when a lot of you hear this stuff, you might originally go, Jake, I'm never gonna build in a workflow tool. I don't need to know this. My tech team does. No, there are many, many salespeople, you know, who are understanding how to build these things. There are many, many people out there that are realizing the future is now. The future is in these types of tools. And if I don't learn how to use these tools, I am going to be left behind. So let's talk about what goes into Agent Kit. The first is Agent Builder. It's this visual tool. Think of it like a Canva for agents, where you design the logic and workflow. If you've seen an N8N, you know, you're saying, okay, the first is this webhook that pulls in data from Salesforce. Then it goes to this agent that does account research and prep. And then it goes to this one from Gong and pulls in the call, and then it combines the new call prep and research with the Gong and then pipes it back to Salesforce. That's just an example of one of these workflows. Uh the next is a chat kit. What does a chat kit do? Well, it allows you to turn any of these into your own like little mini app. So let's say if you're a rep and you want to, you're tired of sending out recap emails. You know, you can then you know have not only the automation, but you could also have a little chat where it chats you and says, Do you like this output? And you're like, nah, make it a little more like this. It's like, okay, great, and now I'll send it for you. So it the ability to go back and forth in a chat environment is huge. This is pretty crazy. Okay. This is this is where we start to get a little meta, which is evals for agents. These are tools that evaluate the quality of the agents themselves. So think step-by-step uh trace grading of like the different instructions, prompt optimization, data set evaluation. So literally the tools now can evaluate each other as you build it. And then obviously, this kind of connector registry where you know I can kind of pull together all the different third-party APIs or get access to other APIs. So these these tools, these no-code tools have been around for a little bit. The the issue has been and why this is such a big deal, it's we used to have to, you have to used to kind of understand a little bit how to use these. Maybe the APIs weren't that good. But you know, the fact that this is really kind of moving everything into not just reactive chat or prompt-based chat, but truly multi-step actions that are being coordinated across systems, that is the next level. My my big takeaway for everyone is to think about this concept of where we are today. We're moving away from this world where we ask AI to do something to a world where AI does things for us. And that is where I feel like if you're in sales, I'm gonna talk through what I think are three very, very relevant use cases for you. That you as an individual rep, yes, you are 22 years old. You are sitting there, you are trying to figure out how to optimize your day-to-day. You can be doing this. I did a version of this, okay. Let me tell you the back-in-the-day version of this. When I was uh just out of college, I was 23, um, and I ended up uh at a professional sports team in baseball. They had an inbounds sales loop. And what would happen is that if you were on a call, it would skip you and go to the next person. And so I was already one of the top two performers on the team. And at the end of the season, I went in, and again, I was not like a math wizard, but what I was is I I love stats. And so what I went and did is I went and analyzed every single inbound call sale and what time of day it happened. And I found these pockets really early morning, about 11 o'clock, about two o'clock, like 410, these little pockets of 20, 30 minutes where nobody called in. And so by using the data, I realized, well, why don't I just make all my calls then? And so, and that began everyone else's uh week from hell, where I just sat there and I was doing my emails, I did my follow-ups, and then it would be Tampa Bay Rays. This is Jake. Next one, Tampa Bay Rays. So they they kept making calls and kept making calls, and I just let the phone loop skip all the way around to me. Then I made all my calls during the times whenever uh you know somebody didn't call in. And after about two or three days, people figured this out. And and I went and I showed them the scatter plot. I showed them what I had done. What do you think people did? Do you think they're like, oh wow, okay, that makes sense, Jake. That's databased. Um, I should do that. The answer is no. They kept doing, they kept just doing it. And so eventually, like over a week, they like changed the phone system, which is fine. But you know, it's a good example. I had to do that by hand. I had to pull in CRM data, normalize it into some type of Excel, um, look for the patterns, etc. But now with these types of orchestrations, I could literally have built that agent in like 30 minutes. And not only that, it would have probably been even more precise. It's like, oh, Jake, also you should do like a 10-minute call block here or a five-minute there. So that for me, I have always been the type of frontline seller who is always looking for ways to, you know, get ahead and to continue to outperform by still working extremely hard. And so if you're a frontline seller and you're like, well, Jake, this seems interesting. I'm telling you, my friends, this is interesting to you because this can be your game changer. This can be your two to three Xer. And so I want to try to give you maybe three tactical examples of the types of autonomous agents that you could build. One, you could do what I just said. If you're on an inbound loop or something like that, you could try to build that one. Like, okay, webhook into my Salesforce runs through this agent that's looking for this type of data, build it out into a table, et cetera. So you could build that one. Um, so the the thing that I would think about is automating personalized orchestration across tools. Okay. And these are the types of agents that could coordinate across CRM, email, calendar, and maybe like your conversational intelligence, right? So many sales bots do one thing. They might reply, they might schedule. But with an agent, you can build a flow. So you can say, hey, if somebody who is a prospect in Salesforce emails, the agent can then check the CRM. It could one, obviously, draft a write-up. It could um escalate something, you could say, hey, Jake, you should probably send this to your leader. It could immediately enrich with a third-party trend. You could create an agent that says, hey, if a prospect uh emails me, I want you to then go and find three publicly available insights about that company that are relevant to what we do, and then feed that back to me in Slack. You could do that today, right? In AgentKit, you can do this. Um, think about number two, real-time contextual like escalation/slash assist, right? An agent could observe one of your activities and suggest or perform actions proactively. So the agent's watching a live deal, it sees that you know there's rules in place. Um, there's only one person and you've had three meetings, for example. Um, there hasn't been any type of interaction in the last five days, something like that. And then immediately escalate, immediately slack you. Hey, what's up, Jake? Just so you know, you haven't had an email from Tim, you might want to follow up. Right? That's a really big one. And last but certainly not least, is just the ability to continue to scale yourself. Um, you know, I can see a world where, like, let's say the things that I do over and over again, you know, let's say uh again, I mentioned follow-up emails, but let's say I also have to create proposals. You know, I what I might do is I might, as I continue to try to optimize and scale myself, is say, hey, um, initial uh Zoom API pulls in a call recording from my number specifically. I then say, hey, here is my statement of work or how I like to structure my statements of work. I then process that, I then push back a couple of clarifying questions to Slack. So using the builder back to Slack, I clarify and then it spits out a Google Doc with a version one of the statement of work. Again, scaling yourself. Okay, so when I think about you know this type of world, you know, if you are a top seller, this is what you have to do. If you want to know how to do this stuff, make sure to sign up for our certification program. But again, eight minutes. Literally in the demo, they built two of these things in eight minutes. And and I do want to hit on the challenges here too, that maybe some things and risks that I want for you all to watch for as well, too. You know, blind optimism and this is going to do everything is not true. You want to make sure that we're keeping the human touch. So that's why, again, you're not hearing me talk about a lot of like automated automated emails. Now, you know, look, uh, you know, hey, create trigger if uh, you know, propose email equals proposal type to type up a draft that I can then potentially send. I'm all for that. Have it type up a draft and then you know, uh go ahead and you know try to make it a little more personalized the situation for sure. But again, make sure you you think about yeah, it'd be great if if a workflow could do that, but a machine, you know, might not, you know, do as good as I could do, etc. Um another thing is to understand that as APIs change, you're gonna have to adjust things. So, you know, maybe one of your connection tools, they change their API. Um, sometimes they degrade, sometimes they update their logic, and so things flow differently. Um, but again, make sure that you're staying on top of some of those things as well. And and number two is the time that it's gonna take you up front to invest in this. And I don't really think that that's a risk. I think it's a risk for the people that don't invest in it. And so if I was a rep, I was sitting in your shoes, this is, you know, I don't really know if there's anything that's come about to where I could actually double or triple my output. And this is no bullshit. This isn't, you know, oh yeah, maybe I could get a little better. This is double to triple my output than this. So um, that's what I've got, man. That is my quick little recap for you on open AI's dev day. I want you to think this concept, a couple key takeaways I want you all to run with is, you know, number one, agent infrastructure is the next wave of automation. We're moving from bots and APIs to autonomous agents, tools that reason and act. And agent kit is a major kind of out-of-the-box enabler, if you, especially if you're a ChatGPT fan. Um, these workflows, number two, um, the ability to orchestrate, assist, and iterate on these different workflows are literally gonna allow people to two to three X themselves. Um, and and three is the time to experiment is now. Early adopters are going to gain a massive, competitive and not only competitive, compounding advantage. You as a seller are gonna be able to experiment and start to realize the art of possible. And as the next thing comes out, you're like, I already know what to do with that. I already know what to do with that. And you as a rep are truly going to be in a different class as the rest of your sellers. So that's what I've got hot off the press, hot off the press, um uh just from Monday. And again, I want you to think about you know, the orchestra, right? The conductor, that agent kit and tools like it, you know, are gonna be the conductor force behind your revenue machine, right? And you actually have an opportunity as the seller to be the conductor. You have the opportunity to orchestrate these things together in a way that's not super techie. And I think that that's what stopped a lot of sellers, and that's what stopped us from moving forward is like, oh, it's technical. I don't know about this. Well, guess what? In today's day and age, you don't need to be that technical to build a lot of these out. And the first movers are gonna have massive, massive advantage. So that's a quick hitter for you on this Wednesday. Uh, make sure to forward this along to your teams. Um, if you're watching on YouTube, make sure to subscribe to the channel, sign up for the next one if you're listening. Um, make sure you go get in front of your computer immediately. We will put some links so you can go check out you know other summaries as well. But I want all of you to realize whether you're a Chat GPT fan or not, the concept of what we're talking about of agents and workflows working across tools is the future. And it's time to get up to speed and make sure you tune in every single week to the AI Powered Seller, where I do my absolute best to bring you what I think is the most relevant AI news for everyone in sales. Make sure if you want more hands-on support, you want to nerd out on this, sign up for our AI certifications. They kick off in a few weeks. I promise you, and I've said this again, if you do not leave that course saying, Jake, this has changed my life, 100% money back guarantee. 100%. Right. And we're gonna build on what I just talked about as well, too, which is really cool. Brian, our AI lead, was telling me, he's like, Jake, I think we should do this in the AI agent kit. I was like, Well, I'm good with that. So that's what I got, everybody. Appreciate you tuning in. Thank you, as usual, uh, for the support, the comments, the DMs. I appreciate it more than you know. Uh, and make sure to share with your teams. Uh, and I hope it was, you know, helpful and actionable for you on a Wednesday. Have a great rest of your week, everybody.