Turning AI Agents Into Revenue Workflows With Outreach

Outreach

00:00:00 Neil: So thank you for joining me on the show today. Can you tell everyone listening a little about who you are and what you do?

00:00:09 Outreach: Hi, Neal. Thanks for having me here today. Um, I'm Abhijit Mitra, I'm the CEO of outreach. Uh, for those who may not be aware, we are the leading edge AI platform for revenue teams. And I've been at outreach for three years now. I joined initially as head of product, and two years ago I became the CEO.

00:00:30 Neil: Awesome. Well, it's a pleasure to have you join me today. And if we look at the landscape at the moment, I think over the last few years, revenue teams have added AI features into existing tools. But many sellers report Loggins noise, fragmented information, a whole heap of other potential problems. But why has more automation not produced a better selling? I'm curious what's what's gone wrong here? We're we were given so many promises. But what happened? Yeah.

00:01:00 Outreach: Promises. Promises. Um, so here's the thing. Um, first of all, I would say that, uh, things are definitely improving if you do it the right way. Yeah. And what do I mean by right way? Throwing another tool or another technology into the mix is never the solution for anything. It's just complicates life for people even more. I just had the other day a customer who was telling me that, hey, um, because they have an AI mandate in the company, there's so many AI tools that they have to use that's just managing the interconnection between those tools is a nightmare. So people are super confused. The data is even more fragmented. The AI is pretty useless because it doesn't really have access to all of the information in one place. So it's really a matter of how you orchestrate the business process that you have. What problem are you solving? I do have the data orchestrated gathered together in one place. Are the agents able to act on the data and how it complements the human activities? That has to work together with the AI. If you haven't thought through your, you know, enterprise AI architecture, then it's a recipe for disaster.

00:02:04 Neil: So what is it that changes when an AI system moves from recommending an action to autonomously running a a multi-step revenue workflow?

00:02:14 Outreach: Oh, that's a lot of stuff in there. Yeah. You, you, you said, first of all, moves from recommending versus autonomously running. So that itself is a, you know, it's a book by itself. Yeah. And then then you mentioned multi-step revenue work and they use the word workflow. So there's a lot of lot of things in there to unpack. So let's start with the basics. Um, when you recommend an action, what does it mean that somebody has to now do that action. Right. Which means that you need, um, you still need human in the loop, which is not a bad thing for certain types of actions where it's, let's say high risk. Um, and you do want the humans in the loop. You don't want, you don't want AI to autonomously go and do stuff that may harm your business. Um, at least initially, you don't want until you feel that the outcomes are meaningful and hitting the target. Um, but then once you start autonomously running a process, like for example, one of the things that we allow our customers to do is, um, run targeted campaigns, uh, through revenue agents. So these agents will go and figure out who to target based on their ICP definition, um, get hold of all the information that they need, whether it's from your internal system or the ecosystem around you or from the internet. And then, um, craft messages and autonomously send those out. Now, many of our customers, when they initially set this up, they do what's called speed testing. So they do spot checks to make sure that this is actually not harming you. And it's actually you got the data, right? Because if you don't get the data right, then it could be, um, it could be quite damaging. But once you get that initial confidence in, then let it go. So now the AI is autonomously running these processes twenty four by seven. And it's just this productivity impact is just unbelievable. Um, like multiple X productivity impact, not percentages. It's like three hundred percent, four hundred, five hundred percent. That's what we're hearing from our customers. Um, and some of this we shared publicly on stage in our conference and unleash as well. Now that's from recommending an action to autonomously running, uh, a process multi step is a different thing. It's about how the outcome of one step influences the input and hence the outcome of the second step. So this is when you chain agents together so that one agent is able to feed into the other. For example, if you are in this example that I just said you're executing targeted sales campaigns, you need to research. That's one agent. And the outcome of the research needs to execute the outreach into those customers. That's the second agent. So if these two agents cannot work together then you've created silo again. So multi step is super important. And finally workflows. Workflows are business processes. There are workflows like outbound prospecting, inbound prospecting, retaining and expanding sales coaching forecasting. So these are all workflows. So when you have an AI system which is orchestrating workflows and not like single steps in workflows, then the value is much higher. So that's what changes in the end. Hopefully that was easy to understand.

00:05:22 Neil: Yes, beautifully done there. And we will have people listening that they hear all the hype around agents and how it can transform their work, etc., but not been brave enough to go head first in yet. So maybe to further bring to life the value that we're talking about here. You don't have to name any names, but could you share a workflow that an agent can manage today and explain where the human seller would still remain involved, just to make them understand the difference it could make to them, and also their role in the whole process as well?

00:05:52 Outreach: Absolutely. Um, so we have a number of success stories from our customers, um, that I will, um, that I will generalize and talk about and we'll go into details also as well in any of them. Um, first let's step back a little bit and talk about what we do, uh, at a high level. So what outreach provides are these AI agents which are orchestrating work on behalf of humans. So the humans are very much in the loop. Um, and these agents take over certain functions all the way from creating pipeline to closing deals to forecasting to sales coaching, um, both for new revenue as well as retaining existing customers and expanding your business with them. Um, so in this, uh, workflows or all across these workflows. Um. An example would be, um, for example, um, what I said, uh, you know, your ICP, you want to target those customers. So our research agent can go and research those accounts and prospects for you. And our revenue agent can go and autonomously, after you've done the initial training, execute those sales campaigns and get you meetings. And then where the human takes over and it's very important is in the relationship building with customers. Today, seventy percent of our reps time is spent in, um, the work that does not contribute to relationship building or closing a deal. These are administrative work, like researching, like updating, like, you know, you know, all of the notes and everything that you have to go and do in your CRM and all of that seventy percent of the time. Yeah. For reps time is spent in that. Imagine if AI could actually take over that, how much productive and how much fulfilling that job would be for a rep. So meetings are coming in. You're taking the meetings, you're actually building a relationship with the customer while in the meeting. Our AI agents are helping you with real time coaching. So they're listening in and telling you, hey, these are the things you need to watch out for. This is how you can potentially answer this question and address this objection after the meeting. These agents are essentially taking all your transcripts, summarizing it, helping you, sending a follow up email, updating your CRM for you, and at the end of the week or month or what's your whatever your frequency is, the AI also does the forecasting for you based on all of this data. And that forecasting is so much more accurate and predictable because a lot of it is based on data that is being entered autonomously by the AI. So this is the kind of productivity gains that we're talking about, and this is how it helps the humans. But in the end, the relationship building is something that you have to have people do, because that's the core of how business gets done. It's the transfer of energy from one individual to the other. And that the I will never take away.

00:08:43 Neil: And any conversation around AI, particularly agentic AI and agents. At the moment, the word context frequently appears. And I was reading before you joined me today that at outreach you argue that agents need sales context built from years of workflow data. So which forms of context makes the biggest difference between, I don't know, a generic answer and a commercially useful action? For anybody listening, just to give them the information they need there. On the importance of context.

00:09:14 Outreach: Context is super critical and context is means. Let me explain what context means. Context is about the of the business data that you have around who your customer is, what historical interactions you've had with them, what products they have bought from you, or services they have purchased from you. What's happening in their industry? What are the signals that you need to look for? Um, for the individuals that you are targeting? What's their background? What's their preference? Um, what have they said publicly about themselves, about their priorities and all that kind of stuff. Once you have access to that, and it's very difficult to get all of that context together in one place. It's not an easy job at all. But once you have access to that context, then you need what's called an agentic harness. And the Agentic harness is essentially making sense out of all of this data based on memory, based on, uh, surrounding information or context based on your own. What's changing within your own business and then helping you make the right connections with the right messages, uh, in taking the right actions. So this, this context graph is super, super critical. Um, a lot of times you hear that, hey, I use cloud or I use OpenAI and hence I have an agent platform that is going to solve all your problems. It is not because the context layer is missing in all of that. So. So you need to have the context of your own business. Incorporate it into your agent harness. Only then will the LM calls be effective and make sense for you. Um, and so that's what I mean by context and, and practical example of context. The signals. So what are the signals? Are you going to target customers or are you going to talk to customers blindly based on everything that's happening or specific signals that is perhaps causing them to take an action, make a move only then that's the right time for you to make for you to basically connect your products and services to what they're looking for, right? So what are the signals that you're looking for? That's usually a the part which has the biggest impact when you're looking at all of the data within the context.

00:11:51 Neil: And MCP and native integrations. These also allow external agents to access revenue context and then act on it. So how should companies decide which data and which actions that these agents are are permitted to use it? It must be a question. You get a lot.

00:12:11 Outreach: Yes. So governance is a security and governance is a big part of any agentic architecture. Um, because if you, if you build a bunch of agents and you don't think about what data they have access to and where and what did they do with the data, then, uh, also a recipe for disaster. Um, so MCP, both client and server is a very, very useful technology. It helps connects all the different, uh, platforms, aging platforms together. So we have both MCP server and MCP client and which means the client, our agents are able to get information from surrounding ecosystem and through the servers. Um, our agents are able to service the outcome of their, of their work into other platforms like, uh, agent for Salesforce, agent for org, uh, OpenAI or anthropic or any other platform, Microsoft as well. Um, the key thing is, like you said, the, um, the data access and the governance. So think of it this way. An agent is like a teammate, right? It's, it's augmenting your people and it's also doing some work by itself. So just like you give permissions and roles to your people, your agents also needs to have roles and permissions. If they are augmenting a person, then they should have exactly the same. No more, no less access to data and capability as that person. If they are autonomously augmenting your team and, you know, sort of like if they are the agent is the person itself, then you should treat it exactly like you would treat an employee. So the kind of roles you would give an employee, the kind of data access you would give an employee, you would give exactly the same to your agents as well. So this is super critical that you're managing your agents as if they are your employees. And if you don't have an infrastructure that allows you to have that level of granularity of control, stay away from it.

00:14:01 Neil: And if agents do become that main interface to business applications, do you think software providers that fail to expose useful context could literally be bypassed? What does that mean for the traditional SaaS model, too? We've heard a lot of rumors around the so-called SaaS apocalypse, but how do you see this playing out?

00:14:21 Outreach: Yeah, that's also a very good question. We all know what apocalypse is and what's happening in the market. So I don't want to I don't want to belabor on that. Yeah. Yes, definitely. If you're a software company and assuming most software companies today are SaaS companies, and if you're not using AI, especially agent AI, um, yeah, then, uh, you will be in trouble. Um, and this is something that I've talked publicly about at outreach as well. We were a SaaS company. Um, we were the leading SaaS company in our domain. And so two years ago when I took over as CEO, um, my job was to transition us from a SaaS company into an AI company. And this is not just in the products on how we build products and how we add value for customers by infusing AI into the workflows. Um, but it's also about how we run ourselves as a company internally as well. Yeah. Because in the agent world, so I talked about how you are essentially augmenting your team with AI agents. So in the, in the SaaS world, what did you do? You used to buy software, essentially rent software. So this is about renting software. And then you would assign licenses to people. And the people will do the work. And with the help of the software. That's how it used to work. Now with AI, with agent AI specifically, what's happening is that you're essentially hiring agents. And these agents comes with skills, some pre-built skills, just like you hire people who come with skills, but they also learn on the job and they come with capacity. And the capacity is what capacity is, you know, the tokens or credits that you're purchasing. Um, and there's a lot of talk right now in the industry about, okay, how expensive these tokens are becoming because it's like people are just using them all over the place. So they come with skills and they come with capacity, and then you allocate that capacity to your team. Um, you augment your team members with those skills and you give them certain capacities to say, hey, you know what? The agent is going to help you do this work. That is the new model. That's how it needs to work. And for companies who haven't figured out how to get this done, you know, they might become extinct soon actually. Um, and so now once you have these skills and capacity and you have the agents augmenting the humans. Then the question is really about like. Like, going back to your first question, uh, how does the process work? What parts of the process should the should be done by agents? What parts of the process should be done by humans and how they work in conjunction with each other? And this is what SaaS companies have to figure out. They have to figure out what they're good at, what their core expertise is, what their unique secret sauce is. Because the front end model is not the secret sauce. Everybody has access to that. And then double down on that and get the best ROI that you can get for your customers.

00:17:23 Neil: And when we talk about sales, they often depend heavily on things like trust and human relationships. So where does automation maybe free sellers to spend more time with their customers and build stronger relationships? And where do they need to be wary that they need to be careful they don't damage the relationship by overusing it?

00:17:43 Outreach: That's right. So it's also similar to what some of the things that I talked about earlier around, like the seventy percent of administrative work that you are doing today, which is really detrimental to your productivity and to your, um, to your motivation to work. Um, I think you can elevate a lot of that through AI and leave the human job of selling the relationship building up to humans to do because that's, I think what we like to do also. Um, so, and I don't think AI is ever going to replace that fantastic advice.

00:18:21 Neil: And the old mantra in it has always been, you can only improve what you measure. And there's been a big focus around return on investment on just about all tech projects at the moment. So what measures should a revenue leader use to determine whether that agent is improving outcomes, rather than just simply increasing activity. What should they be measuring then?

00:18:44 Outreach: That's a very good question because a lot of AI projects, um, don't go anywhere because people don't measure the baseline. So the very important thing, first of all is to measure the baseline. And we are heavily focused on that also through our AI maturity model. Um, this is the model that we share with our customers where our um, go to market AI advisors sit with customers and help them do the baseline measurement so that as, and when they're rolling out their agents through outreach, they've actually seen the outcome. And in the end, what is the outcome for a revenue leader? Its revenue. That's what the outcome is and it's revenue versus cost. So how much more money that you can make with how much less money that you can spend. That's what it is. Um, and so basically productivity gains is a big measure, um, that you would like to measure. Um, but in the end, there are leading indicators which lead to higher productivity, higher revenue with lower cost. And what are the leading indicators? Like how. is your win rate doing? Is your deal size becoming higher? Are you people attaining their quota? Is the pipeline progressing? Is the forecast accurate? So what we have seen with our customers, um for example Dennis Woodside from Freshworks was on stage with me at unleash. And he was saying that since they rolled out outreach, the deal sizes have increased forty five percent. That's not that's not small deal size increase forty five percent. Um, he was talking, talking about, uh, the win rates going up by thirty six percent.

00:20:11 Neil: Wow.

00:20:12 Outreach: Extremely, extremely compelling. Uh, outcome. Thirty six percent. Um, Cohesity is saying that their rep productivity is going up by thirty percent. Thirty percent improvement in product sales rep is making thirty percent more money, uh, thirty percent more productivity. And then, uh, Siemens was saying that, hey, you know what? We have standardized our forecasting, which is, which is basically in the, in the end, predictive. But it's standardized across one hundred and ninety countries on outreach. And imagine the productivity gains that when a global organizations can forecast on a regular cadence at that scale up predictably. Um, so these are, these are the kind of outcomes that customers are talking about. Um, and that I think other customers, the whole world should see this kind of outcome, not just the few customers that I talked about. Um, if they adopt AI properly.

00:21:09 Neil: Wow. There's some mouth watering figures and percentages you mentioned there. So thank you so much for taking the time to sit down with me and discussing Agentic AI for revenue teams, autonomous, multi-step workflows, sales context, application interoperability, and outreach as MCP server and client. We've covered a lot today in a short amount of time, so anyone listening would like to dive a little bit deeper into any of this stuff. Where? Where would you like me to point everyone?

00:21:38 Outreach: Um, so of course you can come to our website, outreach dot ai, and you have a lot more information than what you heard from me today. Um, and you can find me on LinkedIn as well. Um, I wish I was a lot more active on LinkedIn than other folks that I see, but I do go there from time to time, uh, to share my perspective. Um, for example, the other day a customer texted me saying that ever since I transitioned from your competitor to your platform, our productivity went up by twenty five percent. Wow. What? Um, thank you for telling me that. And can I post this on LinkedIn? Like, absolutely. And post this on LinkedIn. So that's why you see a post from me on LinkedIn this week about that.

00:22:20 Neil: Brilliant. Absolutely love it. And so many big takeaways, especially around agents when they become the interface applications must provide the context those agents need or risk being bypassed. So many big talking points I'd love everybody listening to check you guys out, check the links in the show notes, and, uh, I'll include as much information as I can, maybe even include that post that you just mentioned as well. But more than anything, thank you for sharing your story today.

00:22:48 Outreach: Thanks for having me here, Neil. And hopefully for the listeners, this was a useful podcast. Thank you very much.