에피소드 120: AI Agents and the Future of Revenue Ops with People.ai

And it's an important part of the selling process. It's still people to people, right? What's not helpful and not useful is spending time entering information into fields, double-checking information, moving data around, analyzing data, all of these types of things. The version two is that this stuff should just be handled by the software and the AI. Systems are going to talk to themselves and that's going to take us out of the tabs and out in front of customers.

Hi, my name is Demetri Panici and I'm a content creator, agency owner, and AI enthusiast. You're listening to the AI Agents podcast brought to you by Jform and featuring our very own CEO and founder Aytekin Tank. This is the show where artificial intelligence meets innovation, productivity, and the tools shaping the future of work. Enjoy the show.

Hello and welcome back to another episode of the AI Agents podcast. In this episode today, we have Jason Ambrose who's the CEO of People AI. How you doing today, Jason?

Pretty good, Demetri. Glad to be here.

Yeah, glad to have you. I'm really excited to chat a little bit more about what you got going on there at People AI. But before we kind of get into that, obviously the world of AI is an interesting landscape. A lot of people have a cool background of getting into it, their experiences with how in general they felt like the AI world kind of came to them. What was your kind of story behind how you got into AI in the first place?

Yeah, it's an interesting question. So I've spent a lot of time in sort of the rev tech go to market world, particularly in the world of CRM. And I think in that foundation of seeing how people used that product and the premise of people having to manually enter information and try to find answers, I always felt like it should be a lot better and it wasn't improving. So then as I saw AI start to hit and saw the possibilities, that's what really drew me to the space and to People AI in particular is thinking about how AI could do so much more for us to make us productive in the sales organization. I had some exposure before that when I was in the world of fintech, but that was the primary driver for me to be interested and get involved, that a lot of the things that have been lingering for the past 15 years in rev and CRM or things that we can finally solve with AI.

I would love to kind of hear a little bit more about what RevTech is specifically. I think I have a pretty good grasp on it, but maybe some of the audience doesn't.

I think it depends on who you ask and how you choose to interpret it. What were you doing in RevTech then? What was your role?

Let's just talk about you. So to me, rev is everything around supporting the go to market motion. There's probably different categories and maybe there's some purists who keep me more honest, but there's sort of the middle office stuff of CPQ and how do you get deals done and moving orders through the system. But CRM at its core, if you think of that as a core platform, is how do we make sure that we understand what's happening with our deals and our accounts and how do we move that through the process and involve the sellers and everybody who supports them to do that. So I think about that as the anchor to RevTech, because that's ultimately what's generating the sales that turn into revenue. Then there's a lot of different niches that could hang off of it. And the marketing side, the top of the funnel, the bottom of the funnel, post deal support, whatever else. But to me, I think it anchors around CRM as the core platform for selling and growing revenue.

Okay, so that was like how you first got that experience in there.

Yeah, because tech feels like tech kind of made this natural transition to AI. Anyone who's in the space kind of adjusted. I was working in tech then I started working doing content for tech, now I'm doing content for AI and just completely kind of moved in that direction.

So then how did you end up at People AI?

Yeah, so I got connected to Oleg, our founder, last year. It was a fun coincidence that several people that I really like that I worked with were working here and they said you should really talk to this guy. Because of my background and interest in the space and for the reasons that I said, I was really interested in this company and this technology. So I worked quite a bit with Oleg on the strategy and looking at how much the space was changing and how that was creating opportunities for us. Then it just naturally evolved that there was an opportunity to come in and fill in for him. As I mentioned, there was just an incredible opportunity for him related to his home country in Ukraine and it felt like a natural transition for me to step into the role given how long I'd been working with him on the strategy, how aligned we were on that and our shared view of where I was going to take the space and what the opportunity was for us at People AI.

Okay, and what do you think is a phrase to describe what People AI does in a sentence or two?

Yeah, so we're focused on this idea of how to generate answers. Almost everything that we do in sales starts with the goal, whether it's the number for the quarter or it's a meeting or it's an email that you're sending. You may have a set of questions of how do I get to that goal? Finding those answers can sometimes be really difficult and it needs AI to draw from context to do that. Once you have those answers, then you take actions and have a set of outcomes. A lot of the energy in the space is around automating that back half of how do I take those answers, how do I write and send the email for you, those types of questions, and maybe how do we track the outcomes. But the answers tend to be a really hard problem if you look beyond the numbers and into what's the activity and what's actually happening in the interactions with customers. So our job is to provide those answers based on how your sales team is interacting with your customers that help you through that at all levels of activity.

Extending it a little bit, the next question is how did we get there and what's been our evolution? We spent a lot of time around activity capture and providing AI that matches that to the CRM records. In CRM usually, you have these opportunity objects and you're trying to track what happened in a meeting, what happened in a phone call. Post-COVID that got a lot easier because everything was happening on Zoom. Our evolution has been to work off of that foundation to add our own AI on top of it and surface that through our APIs or MCP connection or some of our products to provide those answers. So it's probably a little longer than the tagline you were looking for, but I think it helps.

That was a good explanation. It's a primer to force you to try to say it in as least amount as possible, but you did good.

The question I would have then is what are kind of the main customers that you end up working with? Is it sales leaders, sellers, RevOps? What is a common relationship look like between you and your customers and how they find you and that sort of thing? What are they struggling with?

It's really, we think about where you are in your journey with AI. We probably all feel like the middle of the curve is this experimentation phase. Then the more mature folks like Red Hat, one of our customers, is well down the path of how they're adopting agents and adopting AI in almost everything that they do. Then there are some people who are really just trying to get to experimentation and figure out what they want to do. We tend to have customers who are sort of in the middle or moving to that later phase of okay, we understand what AI does and we have some ideas of how we want to do it and we need to make this real. When they have the experience of trying to do that without us, without a platform that provides the answers that we do, they're the folks who get it and say, "Okay, we get it. We can't just throw LLMs at our emails and our calendar meetings and have that work. We need something else." What's interesting to me is we have a very wide gamut of customers, from AI companies that are as forward as you can imagine, big names that are out there, to an Iron Mountain that does document destruction and digital data services that you wouldn't think of as being AI forward or on this journey but they are. So that's been surprising to me, the mix of traditional businesses with probably the most forward AI thinking businesses.

Traditional is an area that I think could benefit from this a lot. This upcoming year is a great opportunity for every business but traditional businesses as well. It seems to me that we all assumed knowledge work type stuff immediately had the biggest opportunities to be time saved. So that meant tech companies and knowledge work focused companies, but traditional companies still need those things solved. Like manufacturing, we see a lot of chip manufacturers. Folks who have sales processes that don't follow the traditional sales stages or look like a SaaS selling motion. As a space, we've sold to ourselves as a starting point most of the time and so we've usually been the first adopters of new innovations that have come in technology related to go to market. But because of the way that some of these traditional companies sell and the fact that they haven't been up on this curve and they're not doing a bunch of stuff digitally, some ways that does prepare them to lean in harder and get more benefit faster from AI if they're prepared to move that fast, which is what separates the movers from the followers in a lot of these verticals.

Yeah, I totally agree. What do you think is a key differentiating factor between what you're doing and what other companies are attempting to do in the space?

I think it's two parts. One is we especially now are really leaning into this notion of providing the answers where you want them. A lot of customers want to get out of the walled garden idea. Companies that are trying to drive you into their user experience and force you to buy all the products underneath that and lock in salespeople into that experience. Customers don't want that anymore. The build-buy equation is changing, the best of breed versus one-stop vendor shops. That's also changing. They want openness. They want to be able to build their own agents or buy specialized AI that focuses in certain areas of their business and they want on their own to figure out how that should come together for sellers. The second piece is just the output, the quality of the output of what we generate. It's hard work. We have our own AI that we've built and it's very difficult to process a lot of this qualitative information and get answers that are actionable. For example, what's the risk on a sales deal that I have? You can take AI in a generic sense or with some platforms and it'll tell you something like you need to build trust with your stakeholders. What does that mean? What do I go do? But being able to process the information and say the CISO has concerns about GDPR and your ability to protect personal information, so you need to deliver them this content that explains our story of how we handle GDPR, and then you need to triangulate with your business side so that the business side makes sure that security is comfortable. That's an actionable set of answers. We're moving toward aspiring to that quality not just to humans but agents because as you try to automate some of these higher level sales activities, you need answers that are more qualitative and have deeper context than just yes-no answers.

Would you say AI is at a stage where a lot of people are questioning its ability to do those qualitative answers versus just basic generic tasks? What do you think the future is in your specific industry and what you're trying to solve for with the capabilities improving, and where do you think it sits now relative to where people would hope it is?

I think the biggest thing is, as you said, as they're coming out of this experimentation journey, they're learning it can't do everything. We're getting through that. What exactly should it do and what is it good at and therefore what is the need for having some of these answers? Our aspiration is that at all levels of the organization, they're getting quicker and faster action answers that are moving them into actions that help them derisk and achieve better success. Agents will take a lot more of this low-level activity off the plates as you get lower in the organization. The CEO at Red Hat talked about this is what they're experiencing as automation gets down at the individual contributor level and they're seeing a lot of benefit there. Bringing reasoning in and chain of thought models becomes a lot more important higher in the organization because they're dealing with bigger and more abstract questions. Those models need the proper context to be able to reason through the questions they're being asked. We want to take AI up from writing and sending emails to what's my plan for the year at a CEO or CRO level and what are the major risks that we need to address in doing that. So that end-to-end architecture moves from automation at the lowest levels, which is still important and impactful, to deeper strategic reasoning type questions. As these chain of thought models improve and people get proficient with it, that experience for executives and the kinds of questions and productivity that AI is generating will be higher level essentially.

That's a lot of sense. Every different CEO or person running companies like these has a unique perspective on where things are at, especially with how their customers perceive where things are at. Trying to bridge that knowledge gap between what is possible in AI with their expectations and what you can provide is always an interesting little battle I feel like you're fighting as a company like this.

Yeah, sort of. To your question on where are the gaps, it's aligning the expectations but also thinking a bit differently about how fast you can move on some of this stuff. Right now there's a pretty quick step to staying with a human in the loop model and starting to experiment with this stuff. But there are still people who feel like we can automate away everything and hand it all over. That disconnect is maybe stalling or slowing people down. If they focused on a more pragmatic approach to keep people involved and work this through and get us all habituated to using AI at all levels, there's plenty to do and plenty of benefit to get there with that.

Human in the loop is something that was talked about for a little bit. There was a level of pause in people understanding that that was kind of a requirement to be the case for things as models continue to get better. They started to make bigger assumptions on complete follow-through capability. Reasoning got really good so people were like, oh look how smart the thing is. At a similar level, just like with creative writing, we're still at like 85% of the way there with a lot of things and then it would take someone to do 20%. When it got really good with reasoning, the number of tasks expanded within that 80%. Now we're maybe closer to 85-90% with more tasks, not all tasks but more tasks. People are now assuming that since it's more capable that means it's past that 90 to 100%.

It's an interesting idea. As it's had these capabilities, they're drawing more things in that they're trying to solve and maybe those things are more complex. Mentally assuming that they're getting to 90 or 100 on the stuff they were doing, they're adding new things that keep you in that 80 to 85%. There's some truth to that in that as we go to more complex tasks, especially in an enterprise, the easy straightforward stuff that we first started using LLMs on, they could get good results because they're just finding patterns from public domain information. We're moving into more that requires proprietary context of what's happening in these enterprises. That's where there needs to be a complement. We talk about expert agents. You have a first agent, could be Claude or ChatGPT, that's your smart intern that will go find out a bunch of stuff, but it doesn't really know much about how to work in your organization. It needs expert agents in different parts of the functions to tell it how we price, how we do deals, how we generate leads. As customers go to more complex tasks, they need more support from those types of agents, which either comes from humans staying in the loop or a different set of AI capabilities that provide that to the gating agent at the front end of the process.

Absolutely. The director agents and orchestrator agents we've seen over the last few months, because agents wasn't even a term like a year ago when I started this podcast. It was minimal because I didn't even know what I was talking about. I want to dive into some potential clients you've worked with or stuff you've done. It seems like you've had some really cool and interesting success. You guys raised a couple years ago like 100 million in funding, is that correct?

It wasn't quite that much, but yeah, a couple years ago we had a pretty big round.

So you guys had the early timeline to build out something. I'm curious as AI has evolved, where you saw some progress, like you added this, you added that, and how you've managed to stay up with the trends of models improving and capabilities enhancing.

I have this fake brag joke of we were AI before it was cool. It creates a challenge. Oleg, our founder, was in the Y Combinator class with Sam Altman, so he could see what was coming. Even in those days, he had a sense this could be a possibility. The question was when and what would the timing be right. We started on this data element because we understood that capturing and gathering this activity was valuable even years ago for humans working within CRM. Then the question was what would be the progression for AI to take some of the use of CRM that would move out of the system and what would that mean for our product. We spent a lot of time on this data question, which is a hard problem to solve. Then we found it wasn't enough because the market wasn't ready for AI yet in a front-end experience that could see the benefits of having better data. So we went into visualization to show results and have humans interact with the information. Then when LLMs came out, we started figuring out how to use them paired with our data and our own models and AI to distill this information into useful answers for humans. The market was still trying to understand what LLMs did. We embedded it in our product and CRM. It led us to forecasting, a more direct application to say when you build a forecast, you want to know the risks and have your platform go bottom-up on a deal-by-deal basis to tell you what's happening. The big unlock for us in the past year has been MCP. At the start of the year, a lot of people couldn't spell MCP, let alone understand what it really meant. But when we brought it to life and had people use Claude to ask questions that then ask its own set of questions to our platform to get really deep and meaningful answers, that was the aha moment. We're just in the transition to lean into that, but there's incredible stuff we can do now that we're starting to see show up with our customers.

To your point on how we work with customers, Red Hat's Matt uses it every day. He's got Goose hooked up, an open source version of an LLM, and he asks questions at his level to understand what's happening in the business. We've done things like taking our winwire form, which says why did we win this deal, fed it in and said go fill this out for the sales rep, and it does an awesome job. These are tactical use cases we're now seeing open up with agent to agent.

Let's talk a little bit more about agents. When you guys were working through what you're doing, I want to learn more about how the product functions. How does the end user interact with the agents you have and where do they sit in the process? Big question is how do I interact with AI? Is it going to replace me at certain levels or overall? I'm trying to learn more about that.

It gets back to why we're opening it up because customers want different ways to interact with our system. Let's start with humans versus agents, two different ways to ask questions and get answers. On the human side, we embed that in places like CRM, sitting alongside the record. If you're looking at an opportunity in CRM, there's a chat interface where you can ask who are the key stakeholders in this deal and what's their position on our product. It's a way to get the full story of what you're seeing in the records. We have our own user experience called Glass for looking at overall deals, a simple interface with AI tightly embedded for people who want to start with structured data and understand answers from unstructured data. We also do it through APIs so you can build your own UX and surface it wherever you want. On the agentic side, first and foremost, MCP, so that agent can ask questions in the same way a human would in a chat interface. It's cool to see the chain of thought from Claude or others that interacts with it to get the true answer you're asking at the start of the prompt. We also have APIs so they can talk directly or trigger workflows based on answers and reasons. Agents might ask questions that prompt us to send triggers to workflows or the other way around. This is deeper in orchestration and automation where you might find agents.

Let's talk about MCP further. That's a big unlock for products like yours in the last couple months.

I totally agree. What's compelling about it for us and our customers is if you start with a question and an answer, you don't have to burn a bunch of roadmap to build capabilities. It's as simple as just asking another set of questions and letting it reason through how to get those answers. If I step back to software, we had to think about and understand ways to build user experiences and workflows and codify it in our products. Now what changes is the things I want as a user just start with a question and the AI figures it out. The chain of thought takes your one question into the right 40 questions to ask our model and platform, which is set up to respond. It opens up how we think about using AI because it's not a big rollout or new product. You're sitting in cloud and asking these questions. Same for pointing agents to it. You can point whatever agents you want and they can ask their questions and get what they need through MCP and whatever client might be. The reason you might have a different client in MCP working through different information to talk to our server or others is setting up a world of deeper interactions between AI that don't have to be explicitly built and connected like wiring an API. You just show up at MCP, it asks questions, understands what it can ask, and off it goes.

That makes me think more deeply on it. I love your comment about Claude interacting with it. An example I gave on a previous interview is I was trying to help more people like yourself get on more podcasts with my experience in the space doing a new PR service. I was receiving a lot of emails that felt like they were for getting people on the show. I wondered how well I could do that because I have great outreach capabilities with my AI-centered mindset and I feel like the PR space is a little behind. At first, I asked an employee to check out recent emails from PR agencies about potential guests and see what they're offering service-wise and find gaps. Then 20 minutes later I realized this was around when MCP came out for Gmail inside Claude. I just pasted my ask into Claude and it did it. I laughed and said I don't need you to do this. It would have taken him the whole day without AI. It parsed recent emails, scraped websites, then asked me what I was providing service-wise. I told it a couple things. If you change your mindset from voice note to AI connected to tools with your information, productivity is a different level. It's wild.

I had an interesting walk when I first started doing this. We wanted to write a case study on a customer, which is a lot of work. Someone has to talk to the AI and account team, find out everything, and write it down. Salespeople aren't paid to write reports, they're paid to sell deals. It takes a long time. I was curious and asked Claude to talk to our product and see what it could do. It built a beautiful case study I sent it to. Then I thought could I turn this into a post? I asked it to look at my style of posts and it drafted something pretty good. I said it would be a good story to share with a contact, so it checked my email interactions and drafted something appropriate in the flow based on source content. This walkthrough shows how habituating to using the tool makes things faster and easier. It's compelling when it works its magic on you. When you experience this properly, your mind is boggled. It's a very positive thing to experience and use.

There are concerns people have. I'm curious where you stand on what AI will do to the market from a job perspective. I've identified three camps: one is doomsday, two is it will save us time to do more high-quality thinking and work versus nonsensical tasks, and three is sure there will be lost jobs but it will create high-focus solo entrepreneurs or small businesses with niche products. The latter two are more optimistic. I fall between two and three. Some industries will lay people off, but people will make something specific and those remaining will do less grunt work. What do you think?

It's probably a different version of two and three. We believe people will work with people and AI does the rest. Especially in a sales situation, you're wired to spend time talking to people. That's an important part of the selling process. What's not helpful is spending time entering information, double-checking, moving data, analyzing data. That stuff should be handled by software and AI. Systems will talk to themselves and take us out of the tabs and in front of customers. The version of three is some independence roles like specialized content generation could be project-based. Human-to-human roles, reasoning, communication, and interaction are still fundamental. Even what we're doing here, you and me talking, could bots replace us? Maybe, but that feels like Terminator 2 and not interesting. We're social creatures and want interaction at all parts of the business cycle. Humans lean into that. AI is good at information retrieval and processing.

I imagine what it will be like in the future versus now. I wouldn't want much to change practically outside of removing nonsensical tasks and helping with little things I don't want to do. Terminator-type stuff I try to avoid. I don't think we'll get to doomsday from a work perspective. There's been predictions since the 1930s about decreased work and increased leisure, like Bertrant Russell's 'In Praise of Idleness' after World War I. He predicted the world wouldn't work much. This has been predicted for hundreds of years. Prior to computers, there were stages from mechanical to binary computers, mouse clicking, automations, APIs, now AI. I don't know if anyone will predict the next move where we still work 40 hours a week. This is the next thing. People thought the fax machine or internet would change jobs, but jobs just change. We agree more on two and three than one. History tends to repeat itself.

It's a nice positive spin because family and friends worry about AI. I tell them banks just moved off servers for trading platforms managing billions. I don't want to hear that every company will adopt full AI power in five years. People with money are slow to adopt. Physical paper still moves the economy. My dad works in finance and they just moved off physical server-based systems for trading. Big money in tech will handle the transition quickly but not everyone. Those who do will succeed but not everyone.

The pace of innovation is fast and the experience is impressive, generating excitement. But sitting with it and working with output shows limitations and harder work. The horizons are longer. Value will be super high short, medium, and long term. We're in a longer arc than it feels when LLMs have a new model every month and league charts shift dramatically in weeks. It's exciting but a longer journey.

On a personal note, what got you most interested in AI right now? What do you use daily that excites you? Is there a specific product you use all the time?

I wish I could say something that didn't feel like I'm a homer, but I'm excited about what we do with the product almost every day. For example, we talked with Kimberly, head of marketing, about events. The problem with events is how do you measure value? You go to a trade show or webinar and historically we had influence revenue because you touched an opportunity. How do I know the real impact? In what we do, I can say this person from this account attended our event, maybe a webinar or in-person event, and we have context of what they were exposed to. I can take that to AI and ask what happened after that? Did Jim Farley show up and talk about our forecasting product? Did he introduce us to others? How did this show up in interactions? How do we measure success differently than CFOs who want ROI but don't understand value? Now you can have qualitative contextual answers that help solve problems that were difficult before. This is not just our product but the idea of expert agents building and delivering context from proprietary enterprise information to solve challenges with a front-end agent making it easier to ask questions and work through to the right set of questions is exciting. I keep trying stuff and am surprised how well it works.

Outside your product, what do you like using?

I think it's fun to watch what's going on with LLMs. Everything happening at the front end is causing us to rethink how we interact with technology. For example, a salesperson versus a CRO. The CRO deals with unstructured data and doesn't want dashboards but wants a natural experience like Claude to ask questions. They need to get better at writing questions but don't want visualization. A seller works with structured information. How do you blend LLM output with structured quantitative data, which LLMs are bad at, to make something productive? As LLMs and front-end experiences mature, you'll see more blending of user experience between structured data and chat interfaces. You're starting to see that in creating assets like videos and presentations. The goal is a one-stop shop for all information needed. It's exciting to stay on top of and figure out how to apply it in our product. The pace of innovation from vendors is really cool.

I agree. The little nooks and crannies of capabilities vendors compete on is fun to watch. Claude and Google have had a comeback. Google had the wrong name first, then Gemini which was awful. Claude came out strong. GPT was leading for a while. GPT 5 was a debacle. GPT 5.2 is good at coding but not daily chatting. People are freaking out again. They fixed 5.1 but 5.2 is worse for daily tasks. Advanced stuff is cool but daily chat is rough. Sam Altman says he's in code red and releases 5.2. It's fun to watch.

We're watching the canvas get painted. Sometimes you see a pencil sketch and expect a finished product. Companies race to show they're getting stuff out. Maybe 5 and 5.2 are setting the stage for 6. The league tables change dramatically with new releases. Google earlier this year wasn't talked about much but now they're top of the list. People in the know saw Google's issues weren't that bad and they made strides with 2.5. Sentiment drives narrative. Google's 5 flop caused distrust.

This was longer than expected but really fun. Last thing, where can people find what you guys are doing?

People.ai. That's an easy URL. You can hit me up on LinkedIn if you have specific questions. Come find us there.

Thank you for listening to this episode. We appreciate your time. Please check out everything Jason and People AI are doing at People AI. That's literally the most simple domain. Thanks for watching and we'll see you in the next one. Bye.