Sell Like A Leader – Episode 43

David Kreiger sits down with Denise Willett, Vice President of Sales at Coconut Software, which builds appointment, queuing, and video banking tools for financial institutions, and former CRO of Nest Wealth. 

Over the past year, Denise has been building AI agents hands-on for her own sales team, and in this episode she breaks down what it actually takes.

In this episode, they cover:

  • What an AI agent is, and why a good prompt and a good agent solve different problems
  • The two-day enablement format that had Coconut’s sales and marketing teams building agents
  • Why building the agent isn’t the hard part, and the planning template Denise’s team fills out before building anything
  • How her team cut account planning from three hours to eight minutes
  • Agents that surface next best actions and the accounts reps should focus on each week

If you feel like you should be building AI agents for your sales team but aren’t sure where to start, this conversation gives you a practical place to begin.

About Denise Willett

Denise is the Vice President of Sales at Coconut Software, which builds appointment, queuing, and video banking tools that help financial institutions connect with their customers. She previously served as Chief Revenue Officer at Nest Wealth, and before that spent years at Achievers, where she built out the company’s EMEA operations from scratch.

She brings more than two decades of B2B enterprise SaaS experience, and more recently, she’s been hands-on building AI agents for her own sales team.

Podcast Key Takeaways

  • The hard part isn’t building the agent. It’s choosing what to build. Before building anything, Denise’s team fills out a template. It covers the problem they’re solving, whether an agent is even the right solution, what the output should look like, and examples of good output.
  • Frequency isn’t the only test of impact. A task you only do twice a year can still be worth an agent if it saves you eight hours each time.
  • AI can gather the information, but people still own the strategy. Account plans went from three hours to eight minutes. AEs still edit them and present them to a cross-functional team, because AI can pull outdated information, miss nuance, and can’t handle the relationship side of selling.

Connects

Connect with Denise Willett: https://www.linkedin.com/in/denisewillett/

Connect with David Kreiger: https://www.linkedin.com/in/davidkreiger

Subscribe to the podcast and follow our Podcast LinkedIn page so you don’t miss any episodes!

Transcript

David: Welcome back to another episode of the "Sell Like A Leader" podcast, the podcast for revenue leaders who are on a mission to cultivate a high-performing sales team within their organization. I'm your host, David Kreiger, founder of SalesRoads, America's most trusted sales outsourcing and appointment setting firm.

Today, we are bringing you another great revenue leader, Denise Willett. Denise is the Vice President of Sales at Coconut Software, which builds appointment, queuing, and video banking tools that help financial institutions connect with their customers. She previously served as CRO at Nest Wealth, and she brings more than two decades of B2B enterprise SaaS experience to the show today.

Welcome to the show, Denise. Thanks so much for coming on.

Denise: Thanks for having me, David. It's a pleasure.

David: So [00:01:00] I would love to spend most of our time today on something that you have been living through firsthand, which is building actual AI agents for Coconut and the sales team there. I'm excited about this because I think a lot of sales leaders hear about agents and feel like they should be building agents, but they're not exactly sure how to start or where the best use cases are. I love that you've gotten your hands dirty, and I would love to dive into that with you.

I'd love to start high level. For someone who has never built an AI agent before, can you describe what it is in practical terms and what it actually takes to build one?

Denise: For sure. I'd like to start by saying that no matter where you are in your AI journey, you probably feel behind, because I feel behind. I might be doing more than others, but wherever you are, it's okay. There's lots of time to get started. So, what is [00:02:00] an agent in practical terms?

An agent is not a magic employee, right? It's not going to just do your job overnight, as far as I can tell in sales right now. Also, everything I say probably can and will be held against me in six months when everything changes again. But an agent is really designed to do a specific job.

When I think about an agent, it's a repeatable workflow, so it's something I'm doing over and over again. I think most people are pretty used to prompting AI with questions, or getting into a conversational mode with AI for that one-off. But an agent is really for when you need something to happen again and again.

A good prompt gets you a one-off response, but a good agent gets a job done over and over again. It gets that job done with the desired output. And equally important, it also knows what not to give you or what not to do.

David: Yeah. [00:03:00] That's great. Can you give us some context on what got you started building your first agent? You were probably prompting and things like that, but how did you dip your toe into building an agent? What got you interested in that, and what were some of the challenges you hit at the beginning?

Denise: Yeah. Like you said, I've been in technology for over 25 years, so I am always excited about the next thing. I partnered up with our VP of Marketing here at Coconut, and boy, it was a year ago now, which is a long time ago in the AI world. We hired a company, Joel from Polaris Reach, and he came in and talked us through how to create a good prompt and how to create an agent.

Truthfully, I'm not sure I even knew what I was getting into. I was just like, "We've got to do some learning and development on AI." So we did that with the team, and we walked [00:04:00] out of that session learning how to create a good prompt, but also how to create an agent. We had an AI enablement session that was one day.

But I think we did something else that was really the game changer for us at that time when we were learning: we made that enablement session two full days. We had one day with Polaris Reach, and then the next day we went to the office and all got together, sales, marketing, the BDRs, the AEs, the account managers, and we spent a day doing the work.

So we put it into practice the very next day, not really knowing exactly what we were doing, but following the guidance we were given. If I remember back to one of the big aha moments a year ago when I was sitting in that room, it was, "Oh, wait, so you don't just Google something? You actually talk to it like a person?" It was just really odd. And when we talked about how to create a good prompt, it was, "Tell AI what you're trying to do. Tell it that you're trying to create a prompt, and ask [00:05:00] it to help you create the prompt."

David: Uh-huh.

Denise: That was a bit of a mind bend at that time. I'm much more used to it now because I've been doing it for a year. We did have some help to get started on agents, but if you needed to do it on your own, the biggest learning I took away is that you can use AI to help you build an agent. "Tell me what I need to do. I'm trying to build an agent that does these things." I think that can be really helpful.

David: Yeah. I'd love to talk about what you said at the top, and then maybe help people crystallize the use cases and how they can figure out what they might want to do with an agent. I think the key thing you said was that an agent, at its very simplest, is something that you are doing repetitively, specifically a job that you're doing repetitively, versus a prompt, [00:06:00] which is more questions and answers.

Through that prism, can you talk to us about how you approach thinking about which jobs you might want to use an agent for? Maybe you can give a few use cases to whet people's appetite for the things that can be done here. And then, once you figure out what you want to do, what are the mechanics of actually taking that specific use case and putting it into reality?

Denise: Yeah, and I think building the agent is actually not the hard part. The hard part is the work upfront. What do we need to do? What are the repeatable tasks? Which ones should we tackle, why, and how, and what's the impact? In our early stages, like on that second day, we got in a room and just threw ideas on a whiteboard.

What are all the things that annoy [00:07:00] you? Like account planning. It takes so long for each account, right? It takes hours to build an account plan for each account manually. Then there were things like our content database. We have so much content, so many customer stories, so much information, but how do you find it easily?

How do you make sure it matches what you're looking to talk about? How do you find what's a fit? Things like call prep, right? Call prep externally, call prep internally. How do we prep our CRO to join us on a call when he has so many calls that he's joining? There were so many examples thrown out across marketing and sales of, "These are the things I spend a lot of time on."

Going to their annual report, looking at something, pulling that back out, then reviewing my notes from the last call or the recording from my last call before I go into that meeting. There are so many repeated tasks. There's so much information in sales that we sift through to [00:08:00] do our jobs every day.

Everything from planning where to focus this week, to how do I prep for this meeting, to how do I manage up and report my forecast. There are so many pieces to that. So we threw all those ideas down, and then we prioritized them. And then, and I think this is the piece of the workflow that you talked about not to skip, we go through a series of planning.

Before you start building an agent, define the problem. What is the problem we're looking to solve? Is an agent even the right solution? To your point, do I use this twice a year, or do I use it daily or monthly? Even if I use it twice a year, does it save me eight hours each time I do it?

Start to think about what the impact is. Define what you'd like it to do. I think an account plan is a really good example. "Here's the template. Fill it out for me. Go externally, look internally, fill it out." In the early days, we'd [00:09:00] have to give it the data.

At least in our organization, we don't have to feed it the data anymore, because we have a great solution for that, which we can talk about as well. But define the behaviors you want and what the output should be. I know what I want an account plan to look like. I know exactly what pieces I want in it.

So define how you want the output to look. Give it examples. I can give it an old account plan and say, "Every time you return an account plan, no matter what account I ask you to do it for, I want it in this format." And then you really just have to test it and say, "Yes, this worked," or, "We need to tweak this piece."

So document all of that. We had a template that we filled out to document all of that before we even started building the agent itself.

David: Yeah. I've found that with AI in general, and especially with agents, because it can take some time to do the setup even if it's not that complicated, you've got to do it, test it, and make sure it's right. The key is choosing the right problems to work on. [00:10:00] Sometimes you choose the fun ones, the things that would be kind of cool, but they don't necessarily add the most value.

They don't save as much time, and sometimes you can go down a rabbit hole. So I do think some form of rigor, for individuals and organizations, around how you choose what you're going to build before you spend a whole bunch of time building it and rolling it out is really important.

It's great that you had that type of brainstorming session to figure that out. So once you chose an idea, and as you choose ideas now, can you talk to us about how this works? How do you actually go from an idea, whether it's an account plan agent or something else, to actually building that agent?

Denise: Yeah. We have done it multiple ways because, as you know, AI is moving very quickly. When we started, we were building what are called Gems, which is Google's version, and it's pretty natural language [00:11:00] based. Now we've moved over to a solution called Glean, which is a game changer. Glean allows AI access to any document that I have access to in the organization.

For example, if I want to create an agent, I can pull data from anywhere I have access to. I have access to decks our CEO has written, and pitch decks from our head of product and engineering on our new go-to-market products.

I can access all of that, down to my reps' documents that they've shared with me. It's all permission-based. Now, this starts to sound a little like a security and data conversation, but it's all permission-based. It's just the things that I have access to. So I can now pull from any of those things. I can also pull from all of our systems that are integrated, like Salesforce, for example.

So now I can start to do some really interesting things with agents. I can move away from, "Hey, build me an account [00:12:00] plan," to, "Go in and look at this quarter's deals and suggest some next best actions." So now we're moving from efficiency to actual output, productivity, and revenue.

As for how you build those, it's really about taking that template. With Glean, we've moved to a place where we can take that template and build the agent with natural language. I can literally go to Agents, click New Agent, pop that template in, and I've got a new agent. It's as simple as that.

Now, I'm not going to oversimplify it, because those are the ones that I can build. If we really want to get deep into connecting to tools like Salesforce, Apollo, Salesloft, and the other tools we use, RevOps does much deeper, coding-type integration work to bring that to life.

So I've leveraged lots of agents that were built by RevOps that go much deeper, and I've also built some [00:13:00] on my own that use standard fields and things I know from Salesforce. I just have to tell it where to look. So the answer is, it depends on your tools.

I've built in Gemini, or Google. I've built in Glean, which is super easy. It's literally natural language, so you can use your template. And I've built in ChatGPT.

David: Of all the different things that you've built, what has delivered the most value to your team, or what do you like using the most?

Denise: I think that's a really great question, because that's really our next evolution. How do we track this? How do we make sure we're using the agents we build to create consistency across our approach? When I think about moving from efficiency to effectiveness, being effective in your role as a salesperson, that's where the value comes in.

I don't know if there's one, but it's that move from "help me plan for a meeting" to "look at my deals and tell [00:14:00] me what the next best action is." "Look at my territory and tell me which five accounts I should focus on this week."

Surface the contacts, right? These are all agents that we have built. They fire weekly: "These are the top five you should work on this week. These are the top five you should work on next week." So it really starts to automate. We're starting to move from "make my life easier, help me write this email" to "help me do my job," which I think is really exciting.

That focus, and knowing where to focus, is really our next layer. We still have work to do, though. We started with, "Jump in, learn, just build some agents, learn how to do this." Now we've aligned some of those agents with our sales process. So I started asking, "Which stage do these belong to and get used in?"

Our next step, and what we're working on now, is mapping the agents we have across all of revenue, from first touch to advocacy. How are they being used? How [00:15:00] do we measure them? We're not there yet, David, but that's exactly where we want to go.

David: Yeah, that's super fascinating. And obviously all of this is evolving as we go. How has your team reacted? Are they excited about using these? Do you have to force them to use them? How does it work when you build these and roll them out to your team?

Denise: The effectiveness of an agent becomes pretty obvious pretty quickly. I'll give you an example. There's one we have that I think was built for anyone across the company, and it's called "plan my day." When I use it, I'm like, "Oh, this is a really great summary for my boss," but it doesn't really help me.

It just summarizes what I already know is in my calendar and my email, and what I need to do. So you realize pretty quickly that nobody's using it. Nobody on our team is using it. It becomes obvious. Then there are other ones, like our next best action agent, which is being used heavily.

We also have [00:16:00] a discovery scorecard and coaching agent, and a MEDDIC one, and that one is so helpful. When you're going through the sales process, you're not getting all this information on the first call, or even the second or third. It's a collection of conversations. So it really helps summarize: What do I have? What do I know? What do I need to know to follow our process?

We haven't gotten to measuring them yet, but we know by usage. I think our measure right now is just usage. Where we want to get to is: How many hours have we saved? Have we increased velocity? Have we increased win rates? What impact have we had there?

David: You've mentioned a few things that I think are interesting and would be great to clarify for listeners. There are agents where you're still prompting, in the sense that you're typing things in and the agent starts. It sounds like you also have things like "What are my priorities for the week?" that get Slacked or emailed to you, and maybe certain things that just get updated in the CRM.

Can you talk about the outputs of these agents and how they differ?

Denise: Yeah, absolutely. With the account plan, [00:17:00] for example, you go to the account plan agent and write in the name of an account: "Build me an account plan for ABC Company." Whereas one of the simplest ones, and I don't want to say simple, because I didn't build it and that might not be fair to the person who built it...

One of the ones that has created a lot of value is literally a daily Slack. It shoots to us daily. We're an FI-focused company, so it shoots to us daily the mergers and acquisitions in the space, competitive observations, just a snapshot of what's going on in the industry, right at your fingertips first thing in the morning, without you having to go to multiple news sites.

It consolidates it all. The output is consistent, and it follows the things we want to know about. That's been a really effective one.

David: That's awesome. And just on the mechanics, you've mentioned Glean. In the context of Glean versus something like ChatGPT, is it another app that's on their [00:18:00] screen, or is it running behind the scenes? Or is it like ChatGPT, where they're interfacing with it on a daily basis?

Denise: Glean is my new internal Google, but one that I talk to. I can't tell you how many things I would have Slacked someone about: "Can you tell me where I can find the...?" Now I go to Glean and say, "There was a training enablement on X. Can you surface it? There was an all-hands presentation that talked about Y. Can you bring that up for me?"

These are things I would have Slacked people about: "Where can I find this? Where can I find that?" Even questions about our product or our positioning. I can ask it anything, and it will return that to me. I personally have it bookmarked, and I'm on it every day.

That's for the daily things that come up that I would have Slacked somebody about. And then with the agents, I'm [00:19:00] using them regularly, but where I want to get to is everyone using them regularly, in the same kind of process throughout the sales cycle.

But I live in Glean like I used to live in Google.

David: So it basically serves as your ubiquitous knowledge base for the whole company. And it's able to read everything because you've connected Salesforce, you've connected... Are you on Google or Microsoft?

Denise: Google. Yep, we've connected it.

David: And I guess your Slack, so it can search through all of those different things. Does it take a long time to search through all of that, or is it surprisingly fast?

Denise: Surprisingly fast.

One thing I would say, though, all that said: we need to edit. It is very obvious, or it can be, when someone uses AI and doesn't take the time to edit. So we have to have human oversight on these things. We can't just have it writing our emails for us without reading them and checking them, [00:20:00] because we have to check for accuracy.

One of the challenges I've seen is that you have to start with good prompts. If I don't say, "Look only in the last two years," I might get data from four years ago that's no longer valid. So you have to be really specific, but you also have to have that human oversight. I think that's really important.

David: Yeah. I think that's the challenge people are starting to surface and think through, and maybe it's a training or culture thing. We need to give our teams these tools, and we need to think about where we can add the most value, where they can do things more effectively, do things faster, and surface better insights.

But how do we make sure they don't just become somebody who's copying and pasting from AI into an email? Take the account plan. I think account planning was really difficult to do previously, and AI makes it super simple. But we need to make sure that it's [00:21:00] a partnership with AI.

That the AEs are thinking through the account plan, challenging it, going back and forth, and getting ideas, versus just clicking a button, getting the account plan, and not even really internalizing it.

Denise: 100%. What we do is pair that with an internal meeting where we invite cross-functional folks and the AE presents the account plan. Because I look at an account plan not as just something we have or need. It's a rally cry for the whole organization to help. Do you have contacts?

Can you help? Do you have ideas on this strategy or next step? What can we do? So we rally around that and take that information. What used to take us three hours now takes eight minutes, and then we have to edit it. But to your point, what it's really good at is collecting all the information. It might get something wrong because it picked it up from something that's out of date.

It also needs help on the strategy and the [00:22:00] nuance. On the next steps, it can suggest some really good ones, but we often find we have to add to them and edit them. It's not one and done, but it certainly saves hours of collecting things like, "What's their tech stack? Who are the contacts?"

David: Yeah, I love that you...

Denise: That stuff.

David: ...do that. I think it's a really important thing with these types of tools. We want the human's context and background, which the AI doesn't have, to inform the output. But we also want them to internalize it, because there's something to be said about writing something yourself. You internalize it, and you think it through.

So having a process that forces that, like you have with that presentation, is super critical, and I think it's important for organizations to think through. We all have processes for coaching calls, for certain types of training, for account [00:23:00] plans, and now we need to think through what that process is with AI.

Where are the thinking aspects we want to keep the human in? Not just because they have context and ideas the AI might not have, but also because they're the ones who have to actualize it. They need to really understand and internalize it to bring it out of the bits the AI is producing.

Denise: Yeah, 100%. I disagree with AI a lot. I'm like, "That just sounds like a bot. I'm just going to write this email myself." I like to be more casual sometimes with my language, and it can make it too formal. So I think you always have to be providing that oversight. Also, it's not going to solve for the relationship piece, and that is so critical in our business.

I'm a firm believer that if you really invest in the customer's success, you'll get where you need to [00:24:00] go. That takes relationships, understanding a lot of what's happening, and knowing how to navigate sometimes complicated, massive buyer groups. AI is not always going to be able to tell you those pieces.

David: Yeah. So it's important with AI agent design to design not just the AI agent, but the human interacting with the AI agent. I wanted to ask you: if you had one piece of advice for a sales leader who hasn't built anything yet, what would you give them before they start?

Denise: I would say just get started. Start getting your ideas down. Make a list of the things that annoy you. I love AI because I now get to work on the pieces of my job I love, which are spending time with the team, being with customers and prospects, coaching, all of that.

Pick a task that frustrates you or your [00:25:00] team. If you're a leader, pick your team's. You can just use AI to help you create the template you should use before you create an agent. Fill it out. Spend some time defining the problem, defining what you'd like the output to be, and how often you think it might be used.

Just pick one problem that's repeatable and go for it. As far as I know so far, in all three of my examples, it's always on the left-hand side. It's going to be called a Gem, an agent, or a GPT. Just click New and write in natural language, because that's where we're at today.

We're able to build agents with natural language. Pick something simple. You can't ask it for Salesforce data if it's not integrated with Salesforce. Pick something simple, with data that you have, that it can return an output for, and just get started.

Because I had no idea what I was doing when I started. I was like, "Wait, what?" It felt like an agent inside of an agent inside of a prompt. I was like, [00:26:00] "Wait, I ask it to do the thing for me, and then it does the thing? I don't really know what's happening." So just get started, and it will become clearer every day.

David: That's great advice. Thank you. Let's move into a segment we call Rapid Fire Questions. I'm going to ask you a few questions, and you can just give us your quick thoughts. Starting with the first one: what is one thing people don't give enough value or attention to in leadership?

Denise: There can be a lot of pressure in sales, right? It's not for everybody. There can be a lot of pressure. And as sales leaders, we can be driving toward our targets pretty heavily. Take the time to motivate your team, create clarity, create focus, and have some fun, because we're all in it together. I truly believe a highly motivated team creates the best outcomes.

So take the time to do that.

David: Love it. What's one skill [00:27:00] you advise everyone in sales to master, leaders or non-leaders?

Denise: This is a fun one. It's listening in discovery. I love sales reps for this, because their goal is to sell. But in discovery, listen. Pull on the threads. Ask all the big questions, and ask the little questions. Don't even think about what your product is or what you're trying to do until well into the end of the call.

And if there's alignment there, share it. But pull on the threads, and don't micro-pitch. Listen. Ask more questions, and ask great questions to understand their business objectives and problems. Sometimes I see people get really excited, like, "Ooh, we can fix that!"

And it's really early in the conversation, and it's one of ten things we needed to find out about.

David: Favorite business, leadership, or sales book?

Denise: Oh, it's a classic. I've read it [00:28:00] multiple times and recommended it about a thousand times: Chris Voss, Never Split the Difference. I love any book where I can learn and feel entertained while reading. I find most business books very boring. I really enjoy learning and being entertained at the same time, and I felt like I got both of those from that book.

David: Favorite quote, mantra, or saying that inspires you as a sales leader?

Denise: I may have already said something to this effect, and maybe it's from my days in customer success, but: invest in the customer's success, and the commercials will follow.

David: And finally, what's the most important goal or project you're working on right now?

Denise: We're taking a new product to market, which is super exciting and a lot of fun. It's a really big lift, but we're getting some fantastic feedback, and it's been really fun. But what I'm thinking about pretty much every day is how I can help evolve our AI strategy and how I can contribute to [00:29:00] that.

We came in from a revenue perspective full force: "Let's just learn, do whatever, have some fun, create some things." Now we're trying to map all of that and bring it together as more of a strategy across all of revenue. That means being more strategic with AI, to your point, being able to measure and understand it, and making sure we're using our time valuably and not just building agents for the fun of it.

Helping create that AI strategy is on my mind every single day, and we're going to be spending a lot more time on it at our next leadership offsite.

David: Love it. Denise, thank you so much for coming on the podcast today, sharing your insights, and giving us a great overview of what an agent is and how we can use it. If people have more questions or want to connect with you, what's the best way to connect with you or learn more about Coconut?

Denise: Connect with me on LinkedIn, and I'd be happy to chat. I'm sure I have a lot to learn, so if anyone has anything to share with me that I should be doing, I'm open to it. AI is literally moving [00:30:00] so quickly. I'm just open to sharing ideas.

David: Awesome. Thank you all for joining us today. If you have any questions for me or want to connect, I'm always active on...

David: Until next time. Thanks again, Denise.