Loading summary
Transcript59 lines
- [00:00]
Host (possibly Ross Simmons)
If 2025 was a year of AI agents, 2026 is the year of fully autonomous agents. We have Nate Folin from Perplexity who has built an entire Rev Ops team using autonomous agents built with Perplexity Computer. These are some of the most powerful fully autonomous workflows on the web for how you can scale your go to market. All of that and more on this episode of Marketing against the Green. Welcome to the episode. Nate, we are really excited to have you on here. You are going to show us how you have used Perplexity Comp to build a fully AI Rev Ops team. You are a team of one with a whole army of AI agents built all through Perplexity Computer. Why don't you give our audience a little introduction to Perplexity Computer and just why it's so powerful in terms of doing this work.
- [00:49]
Nate Folin
Yeah, thanks for having me on. So Perplexity Computer is a fully agentic multimodal orchestrator that leverages the best AI models to do end to end work for you. I'm excited to show you what it looks like. This is a dashboard that was built based on our customer information with Perplexity Computer helps me get very close to our customers needs. This is one example of what Perplexity Computer has built for me and our operations team. I'm going to show you how to build this and how to work best with Perplexity Computer.
- [01:15]
Host (possibly Ross Simmons)
Okay. I'm lucky that I've had a bit of a pre look at the workflow that Nate's going to share in terms of how he runs his go to market operations and his rev up teams. He's going to share some of the best workflows you're ever going to see from a Rev Ops go to market perspective and how you can use and leverage these autonomous agents like Perplexity Computer. All of that is coming up in this episode. You are going to actually show us a bunch of cool use cases and how you use Perplexity Computer specifically to actually run a lot of your go to market. I think you talked to one of my good friends in HubSpot, Rich. He was pretty blown away by all of the things you're doing with Perplexity. It's like an entire kind of Rev Ops team and so I'm excited to look at some of the use cases and how you kind of use your own tool to run the business.
- [01:58]
Nate Folin
Awesome. Well, thanks for having me on. I'm a big fan of the POD and excited to share some of our use cases.
- [02:02]
Host (possibly Ross Simmons)
Nate, maybe just for the Tee up here so people get the impact of all the workflows you're going to show. How big is your team today in Perplexity because you're using Perplexity to do a lot of these use cases and in their prior life what kind of size of team would you need to do all the things you're doing? Trying to give people like this scale of how much Perplexity Computer is actually doing for your org.
- [02:25]
Nate Folin
So my team is one right now. So that includes rev ops essentially enablement, some data analysis work admin for CRMs and other tools as well as procurement for tools and then just the strategy behind that as well. For the enterprise team I was at ramp before this. We had a team of six and that was just strictly systems folks under the ops team. And then on the go to market kind of revenue operations side we had a host of other folks that were aligned to different sales teams. So yeah, computer has vastly changed the way that we work and just the volume of output and speed that we can deploy new projects and test new experiments in the go to market space.
- [03:03]
Host (possibly Ross Simmons)
So in a prior world a kind of team size would be maybe between 10 and 15 to do the kind of work that you're doing. And now your team is like you and a bunch of perplexity agents. Perplexity Computer agent which that by the way is my dream. If I can live in a world where I'm a team of one, eventually I think that'll be a great day. How does it change your relationship to your work? Like you're working now primarily with AI versus like building teams.
- [03:29]
Nate Folin
I think I was talking to some folks at the HubSpot said my work today looks nothing like it did a year ago. Yeah in that just the breadth of what I can cover. It's very stimulating just to understand like okay, building a comp plan used to be a very long collaborative effort. Now with the help of computer and the network that I can draw from and talk to other folks and then just entering all that information and consolidating, you know, what should we do to build this? How can we back test it? Things like that, that would have taken months or are taking now days to get by in and deploy. And there is still work with humans. So it's not all, all agents all day. There's, there's a lot of co working that's, that's being done. But no, it's, it's, it's a lot of fun.
- [04:09]
Host (possibly Ross Simmons)
If you're enjoying this episode and you want all of the Resources, you can click the link in the description or scan the QR code to get them all, all delivered to your inbox. Okay, so this really is the episode for you. If you want to see what a true agentic rev ops team is in the future, there's probably going to be like fully AI rev ops team with someone like you who's kind of the orchestrator of those AI agents. So I just want to make sure I give everyone like the setup and how you should think about this episode. Maybe we can go into your first workflow and look to see, you know, an example of the kind of things that you're doing.
- [04:39]
Nate Folin
So one workflow I wanted to walk through is case that I find really helpful is thoughtful work. So work that I would like to bounce ideas off of a number of people. I usually reach out to people in my network like Rich over at HubSpot and a few folks there to get their opinions. But I also like to not trust one model. So I've talked to a lot of people that will jump over to Claude, see what the output is, somewhat trust it, and then maybe jump over to Gemini or ChatGPT to get another answer. Perplexity does that automatically. And that's just through something we call Model Council computer. You can prompt it to ask any models for, for answers that you'd like and compare and contrast them. So I really like this Model Council skill for thoughtful work. Like, hey, we'd like to host an exec dinner. Look through all the data that we have available in our CRM and let's pick a city. So starting this prompt off, we'll kind of review all the information that we're connected to. We'll come out with an output here that I can show, which I think is super useful. So it will show at the top here where models agree. So the checkbox is here and the evidence of why they agree, who they should target. Chicago carries a single rep concentration risk. A lot of Chicago deals are owned by one person.
- [05:50]
Host (possibly Ross Simmons)
It's obviously pulling from tools to know where you have deals and what stage you're at and quality of deals. Is it connected? Is that just pulling it all from your CRM, Correct?
- [06:00]
Nate Folin
Yes.
- [06:01]
Host (possibly Ross Simmons)
Yeah. Okay.
- [06:02]
Nate Folin
Yep. So that was part of the prompt here is I went with look through my team's open opportunities in our CRM. And yes, I think that's super important because if you just go third party data or what does the data say? It's less relevant to your work. And we can see kind of how this is working. I asked it just to pull together an output that would be simple to review. And I also asked for the latest models to go review here and it will give a detailed output. You can have this in any format you'd like, broken down by where the models agree, disagree and unique findings for each of those. And they have sometimes different lenses on what they care about.
- [06:40]
Host (possibly Ross Simmons)
Yeah, that's one of the advantages of Perplexity. It's like model agnostic. What if you asked it to basically do that, but then also plan out the outreach to get these folks to actually join the dinner. It'll be able to thread together who are the current contacts you have in your CRM and then based upon that, what is like a good outreach campaign for them.
- [07:04]
Nate Folin
So this is where it starts. You get information from these different models with all the information that you care about with these invites. And then we just launched a connection with Apollo. We have a number of different connections with obviously HubSpot for sending out these invites. And not only can it pick the audience through enrichment tools, but it can actually make the copy and insert the sequences or campaigns directly into those emailing tools. So I've actually two line built full Apollo sequences and added the relevant contacts to those with the right sender for those people. So that can be done all within.
- [07:41]
Host (possibly Ross Simmons)
If someone was looking at this and they were like, okay, this is pretty incredible, I can have Perplexity start to do a bunch of this kind of work. How do you build up your confidence level in terms of how autonomous you allow the agent to be? Let's take this end to end example which is pull data from CRM, then find best places to actually have exec dinners based upon open deals, probably some sort of deal stage and quality data ingestion it's doing, then you could say hey, like okay, cool, like you figure out how to like get these folks at the dinner, right? Like you figure out who I've got the contacts for based upon the contacts, personalize emails and invite those folks to this dinner and appeal to whatever would appeal to them. And you just do it all and send that. That is like the kind of, you know, optimal way to do it. But there's like this kind of trust factor you build up with AI models. Like I trust it to do this. Oh, now it's done a good job of that. Now I trust it to this all the way to like you just do it all. And you could ask perplexing computer to do all that and then likely email you the results as they come in. Right? Because complexity computer can run on a continual basis. So it can just keep checking in to see how that campaign is doing. Right. I'm not overselling it.
- [08:50]
Sponsor/Advertiser Voice
When a buyer asks AI for a solution like yours, do you know if you're showing up or not? At HubSpot we've got a great product that actually helps you get discovered in AI search. HubSpot AEO helps you show up in these moments with the right answers buyers are looking for before the first click, before the first fill. That's the moment HubSpot AEO is built for. Go check out HubSpot AO you can trial it for free and it's going to help you understand how you're showing up in AI Search and how you can show up even more.
- [09:22]
Nate Folin
Exactly. So this would then we translate this over to a skill and we can run a weekly job on this skill to say let's look at the past events, what the performance was and every Monday do an end to end skill which is essentially where should we have a dinner this month? Build the entire campaign for us and then ping our events team, say hey, this campaign is queued up, the people are in it, it's ready to send, we've got a place for dinner picked out and, and we're ready to to hit go on this. So that's kind of where we start is right. Typically have started with something simple and then it grows into an end to end flow where I have my check ins as to when to approve, when to review and then at any point we can kind of change the audience, change the city, change the messaging, etc. But once this is always on, it turns into a 30 minute check in of like oh, let's just approve this each week versus let's do a whole end to end project that takes a lot of time.
- [10:16]
Host (possibly Ross Simmons)
There's so many different cohorts of folks in terms of how they think about AI right now that everyone's experience is somewhat different. Like you and I are in the weeds and so we're like for the most part AI do whatever you can do and just check in with us. But I think you give a pretty good framework on like how you can actually get to that point, which is I would always think of an end to end workflow which I think you have said like there's an end to end workflow and then you're going to break it into stages and you're going to check each stage to see if you're happy with the quality. So you're going to say okay, well first of all pull back the data around the dinners and give me the logic of why you chose those group ends. And you're like, oh, like pretty good logic actually. Your logic is a little awry there. Fix that, do it again. And then, yeah, that's great. Go to the next part. And you go to the next part and then you do that until you get to the end. And then what you're saying is once you've got to the end, you've kind of had those quality checks, you're happy, you can just run it autonomously. Actually, you package it up into a skill, you run it as a cron job and you have, quote unquote, a little worker. So you've gone from like reactive agent who's doing things as you're talking to it to a proactive agent that's just doing work for you and is probably checking in with you to send you an update on Slack or an email of how things are going.
- [11:24]
Nate Folin
Exactly. Yep. And it gets smarter and smarter as the models improve and as our ecosystem of connectors gets broader and as you add more context layers above your, your CRM and things. So that's exactly how I'm thinking about it. I probably have tens, if not maybe hundreds of recurring jobs in the RevOps world to just continuously run each week. Whether it's looking at attribution and how these things went or sharing proactive meetings, that's something that we do quickly. And it's not only RevOps anymore. So in the past, just a quick tangent here is it used to be one RevOps leader would have to control some of these flows. One quick example of that is we have a finance lead that want to jump on some of our finance calls. In the past he might ask for alerts or report to get notified regularly. This is just a two line prompt today. So we just ask computer send a Slack channel alert when, when our sales team books a new meeting with a finance company of a certain size and he could jump on those calls. So little things like that that maybe a year ago used to take a quick intake meeting a ticket to resolve Computer can do really quickly.
- [12:31]
Host (possibly Ross Simmons)
Can we just touch on the visibility part? Because this is like one of the things I see a lot of people struggling with, which is you mentioned you have a whole kind of team of little mini rev ops folks through Perplexity Computer doing things. What is your like dashboard? Like, how do you manage them all? And maybe Perplexity is going to build something at some point that makes this much easier. But like, do you have A little dashboard where you can like kick off your workflows, see how they're all going to, you know, what's your control panel?
- [13:00]
Nate Folin
Yes. So I have my pinned flows in computer and then I also have a formatted Slack kind of summary that I get through a computer that attaches every single day with the things that I care about. So my control panel is computer. It's kind of where I do most of my work, whether it be in Slack, through the computer channel or the threads directly in Perplexity.
- [13:22]
Host (possibly Ross Simmons)
What does it look like in Perplexity? The way you're pinning them.
- [13:25]
Nate Folin
So I'm diving into perplexity here. You asked kind of how I prioritize the work and what I see. So I can see at the top here everything that needs attention and I can collapse these sections if I need to. Some of my scheduled jobs that are running and then pinned work just sorted by day. And so this is how I kind of keep my pinned items handy that are often scheduled jobs. And I also have dashboards that are built directly in, you know, hosted apps that I've built through Perplexity. So sales pipeline, for instance, I've actually built what I think is interesting, a super pack of skills that I'll share and we can maybe link out to it. These are all skills that you can choose your own tech stack here and then it will build agents for you essentially. So based on your role. So revops we can have a few of these that you can just copy these prompts in and run them as role recurring skills within Perplexity. So I use most of these as kind of my sub agents and our team on the sales side uses these regularly as well.
- [14:28]
Host (possibly Ross Simmons)
This is another thing that I think folks are really trying to understand, which is how do you have a shared set of skills across teams? And one of the challenges right now is it's somewhat the messy middle of AI. Every company is an AI sprawl. You know, AI is being democratized. We all want people to be AI native, but it means everyone's kind of building their own skills and doing their own thing. And there's not like a shared kind of this is the best in class skill to do that.
- [14:54]
Nate Folin
Right.
- [14:54]
Host (possibly Ross Simmons)
So everyone in sales just use this skill. It's been evaluated, we know it's the best, we've proven it. And so this is like the skill for this use case. And so you kind of need a shared skills repository that has had the evals done, probably have some sort of observability. So you can see does it correlate to actual results. Is this an internal thing or you've actually done this for external. External folks can go here, get this.
- [15:15]
Nate Folin
I had built a number of skills internally and I asked computer to look at what I've built. What other kind of leaders like yourself have built or talked about people on your podcast for instance, and built this in a way that it can be used externally and copied and pasted over. So this is based on how I work as well as other kind of rev ops and sales leaders.
- [15:36]
Host (possibly Ross Simmons)
That's very cool.
- [15:37]
Nate Folin
We've heard just from a number of customer conversations is this is so broad I want it to do everything. Where should I start? And in sales, I think this is a really great place to start.
- [15:45]
Host (possibly Ross Simmons)
How do you think about the shared skills internally? Like do you build skills for the sales team and then deploy them for the sales skill? Or like how do you think about that problem, which is how do we make sure that teams are using best in class skills?
- [15:59]
Nate Folin
I can actually share our skills library here.
- [16:03]
Host (possibly Ross Simmons)
This is awesome. So this is like your kind of todoist for agents, right?
- [16:10]
Nate Folin
Yep, exactly.
- [16:11]
Host (possibly Ross Simmons)
It's like your productivity tool for agents.
- [16:13]
Nate Folin
The other thing, some of these threads get relatively long. Once I notice that something's getting long, I will just create it as a skill typically and then run that skill regularly. And so that's where I get the most value out of skills. So of course there's template out of the box skills that work for a lot of companies. Those over time need to change or evolve based on changing needs of the company. So oftentimes I say review all of my skills, see what's changed, whether it's headcount and different people at the company or different organizations and update my skills or suggest updates to my skills.
- [16:45]
Host (possibly Ross Simmons)
Yeah, one of the things I've kind of started doing is I use color code a little bit. So I don't know if it's the exact same complexity, but you can put in the hook. So when I shut down, it will create a file of what I did and then basically suggest recurrent patterns that could actually be better implemented in a skill. And so like basically having a way that you compress what you do in these models and then it suggests itself what, what you should turn into skills because you just continue to do that thing and it'll be easier if you just had it all prepackaged in skill.
- [17:15]
Nate Folin
I love that idea by the way. Just that the ongoing improvement I have that regularly every Friday is how can I improve my my writing in Slack? So it looks at how I write in Slack and then what work that I'm working on is the most impactful to drive forward each week. That way it just, you know, validates usually what I'm working on, but it ensures that I'm not skipping things that really should be worked on each week. And because it's connected to all of
- [17:38]
Host (possibly Ross Simmons)
our sources, what I have in cloud code is like, you know how when you're working with it during long sessions, it will compress the contacts because it's getting too long. And so I have a hook that intercepts that. And so when it compresses the contacts, it pushes it to a log file and then something that will parse the log file for repeatable patterns.
- [17:58]
Nate Folin
Nice. So I'm in a demo account here and I just wanted to walk through our skills library. And so we have my skills, which I can create. Often these are created from a thread. I say, you know, save this as a skill and then we can uplevel those to. Org skills. So things that we always want everybody at the company to have available, we can share those. And we also have some example skills here. So lightweight in the demo Org, I think. But we do have a number of skills in our live production. Org and it's growing list every single day of skills that have been created.
- [18:30]
Host (possibly Ross Simmons)
Let's say you had a prospecting skill. Do you build the best in class prospecting skill for your sales reps and then mandate it or do you basically say, hey, like go wild, build whatever. You're a sales rep, you have access to the tool. Build whatever prospect and skill you think is the best.
- [18:48]
Nate Folin
My take on this is I do build something as a starting point and oftentimes we have improvements every single week. So people are sharing and going wild, so to speak, and building their own skills, sharing them with other team members. And we ask that that people, you know, dog food, our product, try to build things that work for them and share with the team. So I do build things like pipeline hygiene, lead leakage, check through any emails that I should be following up on, any stale deals that, that we should re engage and making sure that our hygiene is there and that the customers are getting the support from our sales team as needed. So there are some centralized skills that have been shared with the team, but also there's some really cool ones that people have built like, hey, let's look through my transcripts and figure out what a really cool gift someone might want from Perplexity Swag. Or let's look at customer Testimonials that could be good from this last week or quotes that we can pull out from calls. So everybody on the team is developing cool skills and sharing and then we usually harden them and share them across the entire org once. Once they get to that level.
- [19:44]
Host (possibly Ross Simmons)
Okay, cool. That's a very cool workflow. What would be another thing you would kind of recommend other Rev ops teams to be using Perplexity Computer for?
- [19:53]
Nate Folin
So one workflow I wanted to walk through is backfilling CRM, which is probably everybody's favorite thing to do as an analyst. So in this use case, we have a number of contracts or data living in different places. So we have a DRIVE folder with older contracts. One of the first things I did when I joined Perplexity is look through our past deals and then I wanted to make sure our CRM reflected reality. So in the past it'd be, you know, a bunch of manual work looking through individual contracts with computer. We have over 200 connectors in a growing ecosystem, so we can pull from a number of different sources. In this use case I pulled from Google Drive, we also have an ironclad connection if folks use a clm. And the goal I put here is to ensure that all signed contracts are represented as opportunities with the relevant contacts, amounts, close dates, opportunities, line items, et cetera, and make sure that we're reconciling that information. Now this is a great one off use case. It's also a use case that I see come up maybe quarterly or every month or so on Revops teams to say like, hey, let's see how many people are on auto renew. Let's get another data point into our CRM. How many people have, you know, natural uplift in their contract? So different contract information that really just doesn't live in a CRM is often asked to be pulled. So here in Perplexity you can choose your orchestrator. And then going through these tasks, it will kind of navigate and orchestrate to choose the right models to do the work. So kicking this off, you can really specify your output as it goes through. It will pull and attach to the CRM different opportunities. So I could link out to the OPS there. And then I want an export report. So I'll pull out the report of what it did, how many need manual review, how many were generated, what their opportunity IDs were. And this is a collaborative work, so it can work for hours on its own. And then, you know, if you need a different output, you can ask for different outputs there. And so this is one that I just Did a demo example just with a handful of deals, but this with hundreds of deals actually works at scale. So this could work for hours on its own. And then the output can be in any format. There is the element of human review as well. I may ask it to not auto upload anything until given the approval. And then adding an Excel file, it can actually execute those uploads as well for you.
- [22:01]
Host (possibly Ross Simmons)
And the real use case is it updating your actual CRM or so basically it will go through start to re enrich all the data itself within the CRM. It can work pretty autonomously for hours.
- [22:15]
Nate Folin
The other note that I'll add is that this kind of meets you where you are. So if I like to use Slack, I can run all this in a Slack channel that's shared with other folks on our team and then tag them in and say, hey, does this look accurate to you? Is this contract correct for this account, etc.
- [22:28]
Host (possibly Ross Simmons)
You've shown a lot of cool things. Maybe to round this out, is there like one flow that you've built or one skill that you've built, or one Perplexity agent that you've built via computer that is the one that you would recommend Every Rev Ops team who is want to try out Perplexity, this is the one they would build.
- [22:46]
Nate Folin
Yes. I think the main one that I've loved is our voice of customer dashboard. And this refreshes every single day. And what it does is it pulls transcripts and understands one, how are we showing up to our customers, what's resonating with our customers and prospects, and then two, what's important for RevOps team is what should we do about it? So we have these sections for customer love. What are the key themes? And then what should our product team do about it? And then what should our enablement team do about it? So these are the parts that not only are we extracting things and kind of reading the news of what's happening, keeping a pulse, but we're suggesting things that I can do and my team can do and our product team can do to get better. This was all built with Perplexity computer based on transcript information and emails.
- [23:32]
Host (possibly Ross Simmons)
So basically you're like you're ingesting all your customer transcripts, what feedback you're getting, and then you're distilling it into like a core dashboard and then actually strategically route into the different teams and what parts of this they should actually they have responsibility for and they can improve.
- [23:50]
Nate Folin
Exactly. And then because this is so prescriptive, we can dig into each section. So it's just one more query to say, hey, we found these really cool use cases across this legal space. For instance, let's work with our marketing team to understand and get some customer stories or get a testimonial and really maybe amplify this in certain channels like LinkedIn this use case that a customer has discovered. So on these calls, people are finding new ways to use computer to grow efficiency with their team every single day. And this is a way that we can surface it.
- [24:24]
Host (possibly Ross Simmons)
I love it. I would love to kind of close on. What's your advice on? You've had a great career, you work for some stellar companies, you've said that you've kind of drastically changed how you work and I suspect that the way you work today is probably how you believe the kind of future of work is done. What would your advice be to folks on how to make sure they have a great career in the era that's coming up where AI is intrinsically linked to how you do work?
- [24:53]
Nate Folin
I don't think that has changed too much. I think my general point of view is be obsessed with the customer and whether that's the prospect or the actual customer, that's that's using your tools or that's the internal customer and thinking, hey, what would our VP of Sales want to see today? What would our executive team like to see today? What do our SDR teams need from a RevOps team to be successful in their role and being really crisp on that? And now with AI executing on delivering on what they need, it just becomes that much faster and more accurate using tools like Perplexity.
- [25:25]
Host (possibly Ross Simmons)
Yeah, very cool. Nate. This is awesome. I think you've given a really great overview of how to use Perplexity computer, how to really integrate it into how you do work, how to build true agentic rev ops team, like a team of one with a ton of agents and given everyone that's been listening and watching really incredible workflows that they could go replicate. So thanks for coming on this episode of Marketing against the Green.
- [25:48]
Nate Folin
Thanks for having me. And we are hiring on the sales side, so if anybody's interested in joining but no big fan of the podcast. I appreciate you having me on.
- [25:55]
Sponsor/Advertiser Voice
Let me tell you about a show I love hosted by my friend Ross Simmons. Each episode hosts an in depth analysis of some of the greatest creations and creators of all time with deep dive conversations on the creative process that went into building companies, brands, stories and more. If you like learning about history, learning about the creative process, you'll love this podcast. This podcast is perfect for anyone who wants to learn how to systematize their creativity but still be organized. Ross just did a fantastic episode about Reddit and AI distribution in the playbook that everyone is missing. Listen, wherever you get your podcasts, this
- [26:38]
Host (possibly Ross Simmons)
data is wrong every freaking time.
- [26:40]
Nate Folin
Have you heard of HubSpot? HubSpot is a CRM platform where everything is fully integrated.
- [26:46]
Host (possibly Ross Simmons)
Whoa. I can see the client's whole history. Calls, support tickets, emails. And here's a task from 3 days ago I totally missed HubSpot grow better.