Using Agentic AI Systems to Transform Business Development

[00:00:02]Happy Wednesday.

[00:00:04]I'm taking a minute.

[00:00:05]>> A lot.

[00:00:06]>> We're here.

[00:00:09]All right, lots of ground to cover today. This is the thing that we have once again accidentally had to become good at.

[00:00:16]>> All right.

[00:00:18]>> [laughter] >> And I would say that's been what one of the highest like topics of conversation over the past 2 weeks.

[00:00:25]>> Yes, and I partly blame Tracy cuz she got us involved in it like many months ago.

[00:00:31]Um you know, marketing and BD and sales have always had this very like skittish relationship with each other, which I get.

[00:00:40]But I think when you introduce technology to the conversation, technologists have been even more fluid between those three disciplines because those technologies tend to overlap the most, right? Like you need integrations between HubSpot and Salesforce, and you need integrations between or you need data visibility between your marketing emails and your sales emails and things like that. So, there's a very natural, I guess, through line between the systems that you build for your marketing function and the systems that you build for your sales and business development functions. And we're just finding more and more as we get into deploying agentic systems that the business development function is ready and desperate and wants to take advantage of it. And so, we're we end up getting very involved, and so we've gotten I mean, I literally just spent the last 3 hours working on a multi-agent system for a BD function for one of our clients.

[00:01:39]>> Yeah.

[00:01:39]>> So, it's happening all over the place.

[00:01:41]So, I guess I mean, is that the reason you think that it's just because the technology is there is a natural overlap in the technology?

[00:01:48]>> That and I think that marketing teams are just getting asked to help out with like the sales function, writing the emails, like um since Yeah, I think there is the overlap of the technology and then marketing being asked to support because generally the marketing teams are pretty proficient in different systems that they're using. So, like they're if they're using HubSpot that's for marketing, they are have awareness where it could cross over into Salesforce. And like you said, there are the integrations, so there's already a little bit of an understanding. So, it's an easier um yeah, it's just an easier ask, I think, for like BD teams to reach out to marketing cuz we tend to, right? Like a lot of the asks that we've had so far have been the BD team has reached out to marketing who has then reached out to us to like um help them with or build something because we built something for marketing. So.

[00:02:47]>> Yep, that's true.

[00:02:49]And I think, you know, the BD function is a really good candidate for agentic systems because one [snorts] of the biggest, I think, challenges that business development teams have, whether they're doing ABM, whether they're doing direct outreach, I mean, we work with a lot of SaaS companies, so B2B SaaS, long sales cycles, try to get concrete to carpet, very difficult, all these things, is just the vast amount of data, making that actionable.

[00:03:14]>> Mhm.

[00:03:14]>> Like we have struggled even with executive leadership and talking about enacting strategies, the reality that they're just getting caught up in they're just sitting with Claude all day or they're just sitting with ChatGPT all day. And they have tons and tons of ideas and tons and tons of data and none of it is actually like they're actually doing less.

[00:03:32]>> Right.

[00:03:32]>> BD teams and sales teams are actually doing less. They're on the phone less, they're sending fewer emails, they're doing less DMs, less connections, less networking. Um and I think it can be, you know, AI can be an extremely shiny distraction that allows for a lot of activity and very little effectiveness.

[00:03:51]>> Yeah.

[00:03:53]>> And so, it can be a trap. And so I think that's the difference between having a hobbyist AI person >> Mhm.

[00:04:00]>> and a professional [laughter] come in and build you a system.

[00:04:03]>> Right.

[00:04:04]>> Probably.

[00:04:05]>> Yeah.

[00:04:06]>> Not to throw shade to anybody. Although I guess that is shade.

[00:04:10]Okay.

[00:04:11]So what we thought we would walk through today, and I think this would be helpful for people, is to kind of showcase some of the things that we've built um and really where we're finding value.

[00:04:22]>> Yeah.

[00:04:23]>> So always for us it comes back to like is it actually is it actually valuable?

[00:04:28]Like okay, that was fun, but like did this actually make a difference?

[00:04:31]Like does this actually give somebody something that they didn't have before?

[00:04:35]Um because like you always tell me, otherwise just use what you already have.

[00:04:39]>> [laughter] >> And it's really easy to especially with AI like there's a lot of things we could create, right? There's a lot of things that we could build to make this work really well or just package it super really nicely.

[00:04:54]But on the other hand, there are systems we can integrate it into, which is like what conversation we had yesterday. So we can do the best of both worlds. But yeah, what what you've been working on is amazing and it can integrate into the systems that the sales team are already using because that's another aspect that we always have to consider is like where are they and where are they working and visiting all the time? We don't want another system Nobody wants another system they have to log into.

[00:05:24]Yeah, so building something that's usable um very low low entry point.

[00:05:32]>> Yes, absolutely.

[00:05:34]So I'll give a little showcase of just kind of the the system that we're building or one of the systems and then some of the tools that we've found to be like most successful. So this is just an a rough outline of a schematic that we've built out um that is like a multi-agent system that we've found to have like all the component parts are themselves functional and useful and then together they kind of have uh you know, an orchestrated like a uh you know, a multi-agent functionality that has like a net effect of ultimately being able to early identification identify the best opportunities craft the best messaging and have the right person send it.

[00:06:18]>> Mhm. Right.

[00:06:19]>> I guess if we were to simplify, right?

[00:06:22]So, this works really well for ABM. Um it works really well if you have an integrated system, right? Where you can plug in your CRM, you can plug in your you know, your salesforce, your Gong, your data enrichment tools, Clay, um ZoomInfo, whatever. That kind of thing.

[00:06:37]So, basically what we do is we build out, you know, kind of all of your intelligence agents. Um so, you build out all of the agents that are meant to kind of map your products to intent signals, which is super cool and like I don't know why more businesses don't do this first.

[00:06:55]Like this is So, yesterday we were on a call with a colleague and he was asking kind of like, "Okay, what's you know, first steps and things like that?" And I was like, "Figure out what in your business development function is standard." Right? So, like what do you have that's standard? And most for most business development functions, what's standard is ICP, right? So, you have a standard profile of like who are we talking to? What are their pain points and value drivers, right? Like we all have these very simple kind of frameworks that we operate off of when it comes to the sales function and the growth function. And so, figuring out how you can then use that a in an AI way, right? Is intelligent intent mapping. And figuring out how you can glean insight based on digital activity.

[00:07:40]So, what is the digital activity that sends the signal? And then training an agent to discern what those digital activities mean or could mean, and then craft intelligent communication. All right, accordingly.

[00:07:55]>> Yeah.

[00:07:55]>> Identify the opportunities, identify the people behind the opportunities, craft the communication.

[00:08:01]>> Yeah. I mean, so just that alone is super valuable.

[00:08:06]>> Also, I kind of wanted to break away there cuz it's like my most favorite thing that I've made, and I make things every week and people get sick.

[00:08:13]>> Yeah. Oh, yeah.

[00:08:14]>> This is my favorite thing. Okay? Like, I think this is the Maybe this is the best thing I've made all month.

[00:08:20]So, what we did is we were like, "How are we going to make?" Cuz I was like, "Okay, to really train an agent, I need like massive amounts of scenarios." So, if you've ever tried to train an AI or train an agent or whatever, you know that you don't just tell it what to think.

[00:08:36]Right? You need to give it scenarios. If you want an agent to behave a certain way, you need to give it scenarios. And you can't just give it like three scenarios. If you can give it 400 scenarios, you're going to be in much better shape.

[00:08:47]So, for as many products as you have, you need to, if you're going to do intent mapping, you need to build out a library of possible scenarios.

[00:08:59]>> Right.

[00:09:00]>> That then allow the agent to learn what intent could be indicated by what behaviors or signals or all these kind of things, right? So, it's like this really complex matrix, and I think you need hundreds of them. So, I built this out, and it's like this prompt that you basically plug in cuz I was like, "Okay, well, then what collateral do businesses have?" Right? Because you can't just have a conversation or just tell the AI to make it up for you.

[00:09:28]>> Right.

[00:09:28]>> Right, you need to have actual information. So, I was like, "Okay, the collateral that businesses have, like, call transcripts. You have customer transcripts, you have prospect transcripts, you have sales call transcripts. It's like all that is money.

[00:09:43]>> Mhm.

[00:09:43]>> In all of that is tiny little psycholinguistic cues of what somebody is or is not looking for.

[00:09:52]>> Right.

[00:09:53]Even like reviews of your product, like yeah.

[00:09:57]>> That's such a good point. I didn't have reviews in it. That's a good point.

[00:10:00]So yeah, anywhere somebody is talking positively, negatively, or otherwise >> Mhm.

[00:10:05]>> actually, I guess, about your product, especially in conversation. So if you can do it in conversation, then it can set scenario.

[00:10:14]So essentially what this is is it's like a playbook build and then it creates a format. So what we did is basically took and then this will create We took this, you plug this in as the prompt alongside all of that collateral, all that like juicy goodness, right, of transcripts, yeah, potentially reviews, things like that, as much as possible, and then you have it spits out this giant matrix with intent signals, with recommendations of like if this, then that. So agents love that because then it gives them some rules, right, some structure of how to behave. Um it gives interest signals, it gives psychological suggestibility, so it uses like behavioral economics principles. And then the other thing I was thinking about was um if companies have like a specific um sales methodology.

[00:11:06]>> Mhm.

[00:11:07]>> Like if they use Sandler or something, they could also plug those principles in here, and then the AI would learn that, and then it gives a presentation strategy of like here's how you could present. So essentially what this does then is it would take all of your products, all of your services, and then all of this data around how people communicate around those and create this giant playbook for you to train a BDR agent that's out there looking for opportunities in the world.

[00:11:34]So far the output has been >> [laughter] >> so cool.

[00:11:40]Yeah.

[00:11:41]I'm proud of this one.

[00:11:43]So yeah, I think I mean that alone, I feel like teaches BD teams well, first of all, it teaches them perhaps to think that way, which maybe isn't something they >> [laughter] >> think about themselves.

[00:11:56]>> Right.

[00:11:57]>> [snorts] >> But it also teaches them a little bit more about AI.

[00:12:01]>> Yeah.

[00:12:03]>> I think there's a big assumption that chatbots already know >> Mhm.

[00:12:07]>> and kind of infer and they obviously can't. So >> Right.

[00:12:12]>> this is a training piece is huge.

[00:12:14]>> Mhm.

[00:12:17]So with that would you take what the output from like this prompt say and then put it into the intelligence?

[00:12:27]Going back and it would be as like the uh a knowledge file?

[00:12:31]>> Yeah.

[00:12:32]>> Or how Yeah.

[00:12:33]>> Well, I think you could do it a couple different ways. If you were training a chatbot >> Mhm.

[00:12:37]>> you could do So like we have a we have our little sales chatbot now called Winely and she's trained on all of our products and then we have a giant matrix that's all the Winsome products and then it's mapped for all the signals and stuff um and so like that's in her brain.

[00:12:55]>> Mhm.

[00:12:56]>> Um so she has all of that inside of her.

[00:12:59]So if you're just training a chatbot that way, that would be interesting. But I think if you're training like if you're building out an agent >> Mhm.

[00:13:06]>> the agent probably this needs to be like in a skill file for opportunity mapping.

[00:13:11]>> Mhm.

[00:13:13]>> So yeah, it probably needs to be in the knowledge file, but it'll probably be like in a markdown file somewhere.

[00:13:18]>> Yeah.

[00:13:19]Okay. Cool.

[00:13:21]>> For opportunity mapping for sure.

[00:13:24]So I think that that is a huge way to get to, you know, getting to agentic is thinking about how are you building a brain for BD? What are you, you know, how are you building the kind of internal mechanisms and thought processes behind how BD is mapping opportunities, rating opportunities, because that's a big part of agentic, right?

[00:13:48]>> Yeah.

[00:13:48]>> You think about the things that agents can shortcut, right? Agents can shortcut data ingestion and data evaluation.

[00:13:58]And so, that is something that we prioritize when we create agentic systems, right? So, that relationship intelligence or that I think that's actually more of a data intelligence.

[00:14:09]And then the relationship intelligence side of it is super interesting as well.

[00:14:12]And I think that this is something that Ross has probably worked on more than I have, but creating systems whereby you essentially create relationship graphs across an organization and say, "Okay, who knows who?"

[00:14:27]>> Yeah.

[00:14:28]>> And map the relationships that an entire organization has and use that as the knowledge framework for an agent.

[00:14:37]So, that when an agent is going in, so you can see like every agent you create kind of has a different brain, right? It has a different reference point. And then it's it uses that to make different kinds of recommendations. So, if you have, you know, internally, "Okay, here's all the people Win some notes. Here's all the people join us. Here's all the people Ashton knows. Here's all the people Ross knows." So, then we've profiled each individual within our organization, their networks, and then the agent has understanding of all of that. Then when the business development people go to investigate or go to do ABM or go to do whatever, they're able to shortcut who knows who and who introduced me and who could be a warm introduction and all of that and how should I get into that conversation or get, you know, into that first touch kind of thing by having something an agent that's maintaining that knowledge at all times.

[00:15:30]>> Mhm.

[00:15:33]Can you talk a little bit more about like how the orchestrator sits on top and like what the purpose is of that?

[00:15:42]>> Well, so I think it's two it's two things, right? So, we've said like there are really two outputs when we create these multi-agent systems. So, the first is like is actually not the orchestrator. The first is like the target account straight to BD. Like we want to create agents that essentially take all of this knowledge, all of these, you know, like we've said all these little repositories with their little brains and their little specializations and create like a daily punch list of like the the the agent sends a Teams message or a Slack message directly to the BD based on the what they're farming, based on the accounts they're going after or whatever and says, "All right, here's the connection requests you need to make. Here's the DMs you need to send. Here's the phone calls you need to make." All right, so we connect it to Gong or HubSpot or Salesforce or whatever. And so, there's kind of this synthesized and then that information gets logged of like every single day and the agent is paying attention to like, "Okay, 3 days ago you connected to that person. Now go check, see if they accepted your connection request. Now go DM them." So, it's this kind of like continuous maturity of the relationship building, all of that. And the agent is a great assistant for that cuz it keeps keeping track, right, of however many you're maintaining at any point, 20 or more um people that you're building relationships with. So, that's like the actionable element of all of this where it's like instead of having your sales reps or your BDs get into living in research mode of like trying to find people and trying to read about them and try to make connections and figure out who can introduce them, they're just doing the work.

[00:17:17]>> Right.

[00:17:18]>> The agent's doing the research.

[00:17:20]>> Yeah, the agent's doing all the research, the agent's figuring out all the connections, identifying and evaluating the opportunities, and then just delivering the short chart.

[00:17:30]>> Yeah.

[00:17:30]>> Here's your list, every single day. And then they interact with the list, and then the agent takes what they've done, what they've reported doing, and then puts that data wherever it needs to go.

[00:17:39]>> Right.

[00:17:40]>> And so one of the places it needs to go is into the orchestrator, so it rolls up all of the daily information from all of your BDs, and then every single day that orchestrator goes in and says, "Okay, across all these target accounts, and then across all of my BDs, what was the activity?

[00:17:56]>> Mhm.

[00:17:57]>> What was accomplished today? What relationships have we moved along? Which ones are stalled? Which ones have achieved more? Which ones got all the way to an actual sales interaction?

[00:18:07]That kind of thing." So then they've got, you know, sales leaders, like actual executives, have a very clear line of sight into what's occurring on a regular basis, so there's an incredible amount of accountability, right? And oversight.

[00:18:23]And then on a weekly basis or on demand, they have this orchestrator agent that's sitting over everything and also has access to all of the data agents, all of the relationship intelligence agents, all that kind of thing. So when it comes to like campaign planning or, you know, forecasting or, you know, new target account goals, things like that, they're able to just go dialogue with that agent. You be like, "What's our next best bet based on where we've been successful, based on all these things?" And as long as it's also connected to all of their tech stack, right? It's connected to all of their their HubSpot, their Salesforce, their Gong, their whatever, they're able to have a complete view of everything that's occurring in the department and everything that's occurring in business development.

[00:19:10]>> Yeah.

[00:19:12]It's pretty powerful.

[00:19:13]Cuz then provides a lot more insight into planning because I think a lot of, you know, in general a lot of assumptions are made. Like we see a lot of assumptions made about even in marketing but also in BD. Um you know, where to what industry we want to move into and things like that. But this allows you to have like actual data and performance metrics to back that up.

[00:19:40]>> And I think just as much as there are assumptions about what will work, there are a lot of assumptions about why something isn't working.

[00:19:46]>> Right. Yes, that's a good point. Mhm.

[00:19:48]>> Like it's not working. Well, the sales people are just not doing the work or or these people just aren't interested. The sales people are doing the work. They're they're telling me they are, but they're just not getting any responses. Yeah.

[00:20:01]This way there's no you know, nobody it's it's nobody's word against anybody's word, right? It's literally just the facts of here's the exact activity that has occurred and here's what it has yielded.

[00:20:13]>> Right.

[00:20:16]Yeah, I think the the going back to the, you know, the call transcript agent, that also is so valuable because you can then build email sequences off of, you know, what people are saying and like real scenarios.

[00:20:32]Um and yeah, it's super valuable because again we make assumptions sometimes, you know, about like, "Oh, well, let's talk about this." Or, you know, we can map products, but this is like much larger insight into how we how product mapping should happen and like the most ideal scenario. So Yeah, it's pretty neat.

[00:20:58]>> [laughter] >> Yeah.

[00:21:01]Yeah. And I'm The value of Agentifying, I think, too, is creating those memory loops and those data loops. Because if you think about it, like creating a one-off intent mapping is great and short-sighted.

[00:21:15]>> Mhm.

[00:21:16]>> Because your audience will change, your buyer will change. I mean, we'd even seen that in the last couple of years with some of our clients where it's like, "Whoa, like this job title used to be like a pretty sure bet for ad delivery." And that job title almost doesn't exist anymore.

[00:21:29]>> Yeah.

[00:21:30]>> There's so much change that occurs even at the enterprise level or whatever that you've got to keep a cycle >> Mhm.

[00:21:38]>> going. And there's no way you're going to manually, let's say you're exporting all of your call transcripts into a SharePoint. Like who's going to go download all of that and then upload it into an agent's knowledge base? Give me a break, right?

[00:21:53]Like there has to be some kind of automation that's continuously feeding um all of your agents with the latest greatest information, otherwise even their targeting will be off.

[00:22:05]>> Right. All of your >> you know, goals and targets and opportunity mapping and all that will be a little bit off target. And I think that's, you know, the goal of this cuz like I was thinking about we were even talking to one of our clients a few weeks ago and she was like cuz when we devised this idea for their team and she was like, "The goal is not to figure out what's happening. Like the goal is to figure it out before >> Mhm.

[00:22:27]>> happening."

[00:22:28]>> [laughter] >> Which is a little bit of a misnomer because obviously you can't.

[00:22:32]>> Right.

[00:22:33]>> The idea is for it to feel like that.

[00:22:36]>> Yeah.

[00:22:37]>> Right? Like it was when we were it's like when retargeting ads first started.

[00:22:42]And everybody was like >> Yeah.

[00:22:44]>> creepy, you know?

[00:22:46]That this company knows I was just on that website. Like and everyone was like mystified by it.

[00:22:51]>> Right. Yeah.

[00:22:53]>> I feel like now is the time for companies to start to use agents for BD in a way that not creeps people out, but that almost has a similar level of like forethought or like foresight >> Yeah.

[00:23:08]>> and understanding of their client that kind of like almost predicts what they'll need and then gets ahead of it like with messaging and >> offers.

[00:23:19]Yeah, it's just I mean, we've had intent signals in various ways, but this is like a way to track it earlier and find patterns sooner or that we necessarily wouldn't want because like each BD person is their own person and you don't have those, you know, in-depth conversations all the time about, "Oh, this person said this." and you just don't catch up on catch on to those as quickly as like having an agent and that's its job, you know, to to catch all of the intent signals. So.

[00:23:54]>> Well, and like large language models and artificial intelligence chatbots their literal function is autocomplete.

[00:24:03]>> Mhm.

[00:24:04]>> Like that is literally what they are designed to do is to like you start a thought, it finishes the thought.

[00:24:10]>> Yeah.

[00:24:10]>> They're just fancy autocompletes. So, if you think about it, this is the ultimate application of that >> Mhm.

[00:24:16]>> is to devise a scenario in which you give it the beginning and it fills in the end.

[00:24:22]>> Right.

[00:24:24]>> And so, I feel like it's just such a powerful opportunity for businesses to give it the right information and then give it the right remit and then say, "Go. Do it."

[00:24:36]and then deliver the right to-dos to the people, the right intelligence to leadership.

[00:24:41]>> Mhm.

[00:24:42]>> Your department's going to spend its time doing much more useful things.

[00:24:46]>> Yeah.

[00:24:48]Get back to the relationship building that we have kind of lost.

[00:24:54]>> [laughter] >> And I think it's like the exact opposite of what most people think AI is doing in sales and BD.

[00:25:02]>> Mhm.

[00:25:02]>> Like I think most people look at AI in sales and BD and they think like, "Oh, we're just going to use chatbots to like send to shoot out 2,000 emails a day and we're going to use it to automate DMs and we're going to like that is the worst possible use >> Mhm.

[00:25:18]>> of AI for sales.

[00:25:19]>> Yeah.

[00:25:20]>> Don't do that. Like that just makes you look cheap and >> Spammy.

[00:25:27]>> lazy.

[00:25:28]>> [laughter] [gasps] >> No. Use AI for all of this like strategy and opportunity identification and all of that and then go be human.

[00:25:37]>> Right.

[00:25:39]>> Then write everything manually and do the outreach and build a relationships and go show up at the events and like Yeah.

[00:25:47]>> Yeah, cuz I mean that's that's what is winning the customers is that relationship piece because there's so little of it. Like we get tons of emails about the most random services and everything every day and I'm sure most people in their company do.

[00:26:05]Um it's just like constant. So to have it's very refreshing to have a human reach out and have that relationship uh or establish the relationship. There's much more of a trust component that's built because it's I think everyone is skeptical skeptical at this point of you know, just like cold outreach all the time, spammy emails, but if you're taking that step to like pick up the phone or to send a LinkedIn message and like really be genuine then it makes a big difference.

[00:26:40]>> Right.

[00:26:42]>> And now like using an agent it you don't have this long list. You have like a very targeted list of these are, you know, here's my five people that I can reach out to because I know that they like fit in our ICP and have all the intent signals.

[00:26:58]So, it's not doesn't feel like wasted time either cuz I think that that's another issue that BD teams run into is they're supposed to do prospecting but it feels like a waste of time because it's like Here's Here's your big list of or your ICP reach out to everyone. You know, I mean, it's not that simple but you know, it does still feel like a very daunting task. So, this helps to just alleviate that and really like hone in on best possible contacts.

[00:27:30]>> Yeah, I totally agree.

[00:27:33]Yeah. Yep.

[00:27:36]Something new every week.

[00:27:37]>> [laughter] >> Yep.

[00:27:40]Uh I don't know if we should be launching something new every week but yeah.

[00:27:43]>> [laughter] >> It's just like it's just like as much as technology is evolving everything that you know, we have to evolve as well. So, you know, that's just kind of the the cycle that we're on right now.

[00:27:57]>> Our willingness to move fast I think finally paid off.

[00:28:00]>> Yes.

[00:28:01]>> [laughter] >> Also, when Ross is out of town we can get away with things.

[00:28:05]>> [laughter] >> We're like, he's gone. Quick, let's do new stuff.

[00:28:11]He'll be back Monday and be like, what did you guys do?

[00:28:14]>> We're going to need an hour an hour staff meeting instead of >> Right.

[00:28:19]>> [laughter] >> for 5 minutes.

[00:28:20]>> All right, all right, catch me up. What are we selling now?

[00:28:24]>> [laughter] >> All right, have a good day. Talk soon.

[00:28:26]>> Okay, bye.