Latest / State of the AI Union / Nirman Dave: The RevOps Reality & AI
Transcript
- 0:06Hello, everybody, and welcome to this week's episode of the State
- 0:09of the AI Union. Join with me is Narman Dave from
- 0:13ZAMS, and I'm your host, Laura Narman.
- 0:16Do you want to introduce yourself and maybe tell us a
- 0:18little bit about Sam's? Yeah, hey, I'm Nirman, founder
- 0:22of Zams. ZAMS is an AI command Center for
- 0:25BDB sales teams. It connects to all your apps and
- 0:27services and allows you to search and run automations
- 0:30across them in plain English. So you say something like pull
- 0:33my meeting notes from Gong and update the leads in my sales
- 0:36force and it just does the job for you.
- 0:38Set it up on automations and do your research.
- 0:41Any sort of, you know, sales task that takes up tons of time,
- 0:45tons of copy pasting, tons of tab swapping around Zams can do
- 0:49it in a in a second. Awesome, Norman, tell us what
- 0:52you were doing before Zams. Yeah.
- 0:54We started as a company called obviously AI, which was a
- 0:57predictive analytics company. So it was building general
- 0:59purpose AI models for customers. And we've now honed in on AI
- 1:03agents specifically for revenue operation use cases.
- 1:06OK, very cool. So your AI for Rev OPS, that's
- 1:09the use case and we met through your LinkedIn cold outreach.
- 1:14So tell me more about that. Why did you feel like I was the
- 1:17right person? Who you're going after?
- 1:19Who have been your best customers so far?
- 1:21Well, we mostly work with revenue operations teams, right.
- 1:24And one of the things with Rev OPS teams is that they were
- 1:27traditionally meant to be more strategic roles, more strategy
- 1:31roles. And today what's happened,
- 1:32sadly, is Rev OPS teams have become more of functional roles.
- 1:36They're kind of doing the heavy lifting of cleaning your CRM,
- 1:39pulling the kind of spreadsheets and notes and numbers and
- 1:42putting it together. And you know, what they should
- 1:45be doing is be focusing on the strategy.
- 1:48How do we help like the sales teams, how do you help help the
- 1:51AES have the right kind of ammo to get the best outcome?
- 1:54And that's what we want to help with Zams.
- 1:56We want to help revenue operations teams get back to IS
- 1:59strategy instead of the functional day-to-day of
- 2:01clicking through tabs and going through stuff.
- 2:03So that's how we got started and I think that's how we got
- 2:06connected as well, yeah. When you think about the the
- 2:09best case and the ideal, like Rev OPS team, what does that
- 2:13look like to you? Yeah.
- 2:15We've talked with, you know a little over 100 revenue
- 2:18operations teams today across different different customers,
- 2:21different leads, different people you've spoken to.
- 2:23And it's very obvious that the best Rev OPS teams are the ones
- 2:26that are very much strategically aligned with the CRO and are
- 2:29focused on the strategy. So an example is when you think
- 2:33of like any sales org, you have the CRO who's like very much
- 2:35focused on, you know, what are the markets we want to target?
- 2:38Like how do we get, you know, some of our biggest customers?
- 2:41How do you get like the big name logo, things like that.
- 2:43And then you have your account executives, SDRS that are kind
- 2:46of on ground typically focused on how do I hit my quota, right?
- 2:49How do I close like XYZ number of deals?
- 2:51How do I get this specific lead in in place?
- 2:53But you really need a function that is saying, how can I give
- 2:57the best kind of ammo to the team that we have?
- 3:00And that function needs to be very much strategically aligned
- 3:02with the CRO. So that's the Rev OPS function.
- 3:04And you know, some of the best teams that we've spoken with are
- 3:07the ones that are very much aligned with whatever the
- 3:09execution is, whatever the vision is with the CRO.
- 3:12And they're really kind of thinking of, OK, what can I give
- 3:15my team? Is it dashboards?
- 3:16Is it certain specific automation, certain structure
- 3:20that can help them move very fast without kind of getting
- 3:23caught up in, you know, the admin, the grunt work, right?
- 3:27So that's the kind of teams that we've spoken with.
- 3:29These ones are like that. Some of the ones that have a lot
- 3:32of figuring out to do, they're often kind of the ones that are
- 3:34stuck in that admin work and often find themselves being kind
- 3:37of the human glue to kind of get things going as opposed to a
- 3:41thinking strategy. So that's kind of high level of
- 3:44what I've seen with. But, but let's get more
- 3:47specific. So like I hear you saying that
- 3:49like the Rev OPS, the best Rev OPS teams you've seen are like
- 3:52really strategic. What what about like
- 3:55functionally? What would you say their
- 3:57functional responsibilities are? Yeah.
- 4:00So let me give you an example. One of the best examples that
- 4:02I've seen is one of the companies we work with, they
- 4:06said that it's very important for their.
- 4:08So this is what the Rev OPS person identified.
- 4:10They had A's that were going to customers and the A's often
- 4:13would go to these customers and they would have all the ammo
- 4:15that they need to close the customer.
- 4:16So they would say, hey, we're going to go ahead and renew our
- 4:19deal. The prices are going to go up
- 4:20da, da, da, da, but oftentimes they lack to the visibility of
- 4:24usage of that customer. And the reason they lack that
- 4:26visibility is that the usage was something that was reserved by
- 4:29the engineering team. And they had to often as an AE
- 4:32often have to go to the end team and say, give me a usage report.
- 4:35Oh, I don't want just like last month's usage report.
- 4:38I want last quarters. Can you give me this specific
- 4:40usage on this specific feature that we have so we can talk
- 4:43about, you know, this outcome for this customer.
- 4:47So every time they had a usage report that they wanted, they'd
- 4:50have to go to the engineering team and kind of set that up.
- 4:53Oh, as a Rev OPS team, some of the best Rev OPS teams that we
- 4:57worked with, they would say, hey, let's get in and go ahead
- 4:59and set up a dashboard for our sales team.
- 5:03So while they're on the ground, they have it on their phone,
- 5:05they have it on the computer, They have these dashboards that
- 5:07give them instant responses. So when they're selling, they're
- 5:10not stuck saying, oh, let me loop in our engineering and get
- 5:12back to you. They're right over there saying,
- 5:14oh, you know what? I saw that your usage has gone
- 5:16up here and we can, you know, expect that the usage is going
- 5:19to go in this direction. Let's talk about the the kind of
- 5:22deals that weren't closed. So specifically speaking, the
- 5:26decision to say, I'm going to provide my team with the
- 5:30dashboard that has this exact thing.
- 5:32That's a row OPS decision and the decision to kind of
- 5:34coordinate internally and say, hey, engineering team, let's
- 5:38kind of put this together. These are the resources I need
- 5:40from the company to execute it. That's the Rev OPS decision.
- 5:43What is happening sadly today with, you know, majority of the
- 5:46Rev OPS teams is instead of making that level of strategic
- 5:49decision, they're stuck actually going into the CRM, you know,
- 5:54filter, adding a bunch of filters, getting a bunch of like
- 5:56knowledge and copy, basing it into a spreadsheet and sending
- 5:59to someone. And then every time someone has
- 6:01a follow up, they're the ones who get stuck doing that job
- 6:04over and over again. So if I was to kind of
- 6:06specifically talk about it, right, like this is 1 very good
- 6:08example I've seen where when it comes to providing the ammo to
- 6:12your sales teams, the idea being that the robots person makes
- 6:16that strategic decision, orchestrates that internally and
- 6:19gets that output as opposed to the robots person being the one
- 6:21that clicks those buttons. And it's just like always inside
- 6:25a sales force trying to gather those.
- 6:27Those insides so you're you're kind of like main functional
- 6:30functional things that Rev OPS team would do is like provide
- 6:34provide important decision making data set to the sales
- 6:39team so that they have the ability to move at the fastest
- 6:42pace, you know so. Correct.
- 6:45So that's one of the tasks, right.
- 6:47Another example is contracting. So we worked with a couple of
- 6:50companies where every time the salesperson was going to go
- 6:53ahead and close a deal, it would take them 48 to 72 hours to just
- 6:58get the contract in place. Why?
- 6:59There are tons of templates and each template had like a bunch
- 7:02of different variations and someone had to go in mostly the
- 7:04salesperson because their job to hit the quota had to go in and
- 7:07say, oh, let me edit this line. Let me let me structure this.
- 7:10Or we give them a specific quote that includes these line items
- 7:13that's not in our templates. Let me just go ahead and do that
- 7:15right. One of the OPS things we worked
- 7:17with and they use Xams for this, where they connected their
- 7:20Google Drive and now the salesperson that they created
- 7:23this sort of a setup where the salesperson can just input these
- 7:26are the terms are provided to this lead and it would go ahead
- 7:29edit the the the document, give them the output very instantly.
- 7:33So it's that Rev OPS person's genius who said who first
- 7:36identified this problem and said, you know what my sales
- 7:38people are getting stuck putting these contracts together every
- 7:42single time and they have to wait on someone.
- 7:44They have to get approval Dah, dah, dah dah.
- 7:46I'm going to make that process easy.
- 7:47So that's what the Rev OPS person said.
- 7:49Then the Rev OPS person decided how am I going to do that?
- 7:52Well, I can't use traditional like if this, then that kind of
- 7:55workflows. So I'll have to come up with
- 7:57something more agentic. That's where I'm going to loop
- 7:59in AI right? And then they end up using Zams.
- 8:02But that's the, that's the typical idea where the Rev OPS
- 8:06team isn't just thinking dashboards or automations,
- 8:10they're really thinking, how do I find ways to accelerate my
- 8:13team? And, and that question can lead
- 8:16to I should do dashboards, I should do this, I should do
- 8:19that. And then that leads to what
- 8:21tools I want to use and things like that.
- 8:22And today, so I'm hearing a lot of my ref house peers also like
- 8:26using AI and dabbling in it. What are you seeing as far as
- 8:29like limitations in terms of where they're running, how
- 8:32they're using it today, and like maybe what the limitations of
- 8:35what they're using? Yeah.
- 8:38So let me talk about how they're using it today.
- 8:40Most common use cases are one of my favorite ones is, is lead
- 8:45routing. Typically when it comes to lead
- 8:47routing, let's say you have 100 AES, right, spread across
- 8:51different regions. And these A's are supposed to
- 8:53kind of when you get an inbound lead, you have to set up a
- 8:57system that decides who gets the lead, right?
- 9:00And typically, traditionally companies would either use kind
- 9:04of really bloated, expensive, you know, plug insurance or
- 9:07software that do that, or they would create these sort of
- 9:10complex workflows where they say if the AE is in the same
- 9:14industry as the lead, if the AE has a high closed rate, if the
- 9:17AE has da, da, da, da, da assigned the lead to them and
- 9:21then correct. Yeah, there's a great
- 9:23deterministic if this, if this, if this works.
- 9:27Now that's very challenging in many cases because let's say,
- 9:29for example, the lead has a, so let's say, for example, the AE
- 9:33has a history of working in logistics, right?
- 9:36And you're just not even considering that you want
- 9:39something that has more intelligence and, and kind of
- 9:42bring that picture in. So that's where they use AI.
- 9:44So one of the examples we've seen people use AMS is they
- 9:47would say something like, look at these criteria as you know,
- 9:50closed one rate industry in which the AE was before or any
- 9:54historical hobby that you can find about the AE from these
- 9:58different data sources. And then these different things
- 10:00about the lead intelligently decide which A is best fit.
- 10:05And that's where AI comes in because you can hammer down as
- 10:09many deterministic factors as you can, but it will never give
- 10:12you the best outcome or you might always miss out on
- 10:14something. And second thing is it's so
- 10:16brutal that let's say tomorrow if you have said if this then
- 10:20that and you don't have one piece of knowledge, it just
- 10:22breaks. So you need something that's
- 10:23more intelligent, can make it's own decisions, can do it's own
- 10:26research. And that's where lead routing is
- 10:28a really good use case that comes in where people use AI for
- 10:31it because it can work very non deterministically and it can
- 10:34give you the best scoring, you know, assuming you give it the
- 10:36right direction. So that's one of the one of the
- 10:39use cases I have typically seen as well.
- 10:41How about a use case where you know the Rev OPS team is already
- 10:44using AI in some way and you know they're running into like a
- 10:49limitation or challenge with with doing it in that way?
- 10:54One of the most limited AI use case I've seen, and this might
- 10:57be slightly counterintuitive to a lot of what you see in the
- 11:00market, is the use of AISDRS. OK.
- 11:04And it's this idea that I can send out a bunch of emails to a
- 11:08bunch of leads and they're gonna get this personalized, like copy
- 11:11of it. What has happened is that a lot
- 11:13of those emails become so similar sounding that when a
- 11:17lead gets them, they just kind of feel like, OK, I, I kind of
- 11:19can tell this one's more AI, but more importantly, like that is
- 11:23limited by how much knowledge the AI can find.
- 11:26And it's often scraping some LinkedIn info.
- 11:28So as a lead, you're just getting a message that says, I
- 11:31saw you worked at XYZ, right? And you probably know that you
- 11:34get that thousands of times. And so while people believe that
- 11:37there is a level of personalization going into a
- 11:39ISDRS, it's a little bit challenging to really nail it.
- 11:43And, and I think that's one of the use cases.
- 11:45We don't support exams because we're focused on revenue,
- 11:48operations, internal use cases. But often times that's where
- 11:51I've seen teams struggle a lot is they would have spent six
- 11:54months trying to like get the perfect outcome.
- 11:57And yet sometimes you know, it just doesn't hit hit the hammer
- 11:59on on the head. So that's kind of what we're
- 12:02seeing is, is a little bit of limitation.
- 12:04There are also a couple other limitations with scale that I've
- 12:07seen. So a scale limitation would be
- 12:10go through 100,000 leads in my Salesforce since the last 10
- 12:14years and find me all the stale leads and that works really
- 12:18well. If you're looking at a smaller
- 12:20time beer like the last quarter, last one year maybe.
- 12:24But the moment someone starts to do major scale operations,
- 12:27that's where AI starts to kind of hit the limit as.
- 12:29Well, and why do you think that's the case?
- 12:30Why does the why does the AI sort of like crash out at at the
- 12:33scale? Yeah.
- 12:34So a lot of AI today is designed with a context window in place.
- 12:38And So what that means is there's only so much information
- 12:40it can take as context and and kind of come up with conclusions
- 12:44on it. So if you say for 10,000 leads
- 12:47or for 50,000 leads, you know, in my sales force, go ahead and
- 12:51update all the records or go ahead and find me the stale
- 12:55leads and research those leads and tell me what specifically
- 12:58what they care about about regarding our product, right?
- 13:01So when you come to something like that, it has to go through
- 13:0350,000 records and each sales force account you probably know
- 13:06is going to have tons of activities, tons of like fields,
- 13:10a bunch of like custom fields. You probably have set up custom
- 13:12fields set up 10 years ago that people even forgot about.
- 13:15So it's going to go through all that information and that's
- 13:18where the limitation comes in today with AI of like it just
- 13:22doesn't have enough context to handle all of that info to find
- 13:25patterns. And what we do at Xams is kind
- 13:28of we try to fix that in different ways so we can get
- 13:31like a really good experience there.
- 13:32But that is the general limitation if if someone's kind
- 13:35of just plugging in some AI. Yeah.
- 13:37So like especially maybe like if they're using legacy AI of the
- 13:40existing system, right, the CRM or whatever it is.
- 13:43So tell me like this AI SDR problem does that XAM solve for
- 13:48that or how would it help the team with that something?
- 13:51Like that we don't. I think our core philosophy
- 13:53exams is we're not here to replace a bunch of AIS or SDRS.
- 13:57Our focus is to be that Rev OPS person style, you know, thought
- 14:00process where I want to support the team 'cause ultimately
- 14:03humans are the ones that are, you know, responsible for
- 14:06building relationships and driving those conversations and
- 14:09change forward. So how do you build systems that
- 14:11can support them and support the Rev OPS teams?
- 14:13So a lot of these cases that Xam simply solves for are things
- 14:16like leave outing, which is one thing that we talked about
- 14:18meeting briefings, like how do you make sure that those reps
- 14:20when they're going into a meeting, they have the right
- 14:22kind of information and knowledge.
- 14:24And also a lot of things around metrics and reporting.
- 14:26We recently had end of quarter last month and we had tons of CR
- 14:32OS on Xams constantly. Just like looking for metrics
- 14:35and numbers. Like a great example is if you
- 14:37want to get the average time to close for each of the Rep, it's
- 14:40very hard to do it because typically what you'll have to do
- 14:42is you have to go to your CRM or go through each and every Rep.
- 14:45If you look at the deals, you look at how quickly the deals
- 14:48are progressed from intro 2 closed one and then start to
- 14:52kind of map and sum those numbers.
- 14:54And now we're doing tons of different operations in the in
- 14:57the spreadsheet, going through a bunch of macros and it's it's
- 14:59heavy lift work. What Xams does is if you come to
- 15:02XAMS, you just say at Salesforce, give me the median
- 15:05time to close for XYZ Rep and it will do all the heavy lifting.
- 15:09It understands what's required. It understands where to look,
- 15:13connects all the dots and does the heavy lifting, gives you the
- 15:15number back. So what happens ultimately is
- 15:18that the Rev OPS person doesn't have to kind of do all that
- 15:21heavy lifting or kind of manage a bunch of macros.
- 15:24They can quickly run those numbers, double check a couple
- 15:26of things, pass it on to the CRO and start to think about more
- 15:29strategic things. What should the CRO be reporting
- 15:32in the board meeting? That's a better decision and
- 15:35that's a better strategic thing to be involved in as opposed to,
- 15:38gosh, how do I get this number? By clicking a bunch of different
- 15:41tabs and putting macros in the spreadsheet.
- 15:43So that's kind of something that we see as well.
- 15:45So tell us a little bit more about how Zams does this magic,
- 15:48like how is it able to be able to get all this information?
- 15:52And my assumption is that, you know, before ZAMS, you know, a
- 15:56Rev OPS team might be using Zapier or some kind of like
- 15:59other API integration to do that.
- 16:00Tell us a little bit more about how how the the technical side
- 16:03of it works. Yeah.
- 16:05So with language models it was very clear that we can do a lot
- 16:09of like language related work. It brings great user interface,
- 16:11a great brain, but the the brain needs a body, it needs access to
- 16:16apps and services, it needs a memory, it needs all these
- 16:19things that are hyper optimized to get it to work.
- 16:22What we've done with Xams is we've created something called
- 16:23AZ 1 engine, which is kind of our technology that hyper
- 16:26optimizes for Salesforce, HubSpot, Gong, all the sales and
- 16:31Rev OPS stack that you have. And what it does is it gives
- 16:35access to those things, It gives them memory.
- 16:38So it remembers specific certain things about your business.
- 16:41It also gives it scale. If you were to say here's my
- 16:44HubSpot to any AI and say give me the last 10,000 leads, it's
- 16:48going to start to kind of, you know, hallucinate and kind of
- 16:50fail. You really optimize ZAMS to not
- 16:53do that, to really work at scale with your CRM systems with, you
- 16:57know, something like Gong, right?
- 16:58If he said give me all the calls for my AE and tell me what he
- 17:01should do better in the next call.
- 17:03That's a ton of information. That's tons of transcripts and
- 17:07words that the system has to go there and always fails.
- 17:09ZAMS is really designed for that sort of scale.
- 17:13So it works with your GTM stack, it works at scale and it's
- 17:16really designed to have a sort of memory along with a language
- 17:19model of your choice. So you get to pick which
- 17:22language model you want to use, GPD, Claw, Gemini, things like
- 17:24that. And so ultimately what it does
- 17:26at the end of the day is it creates this beautiful system
- 17:29that talks to all your apps and services and just does the job
- 17:33as opposed to you creating deterministic workflows of like
- 17:35if this and that. And there are two example that
- 17:39we talk about a lot as well, which have existed for a while.
- 17:42And with the, the problem with tools like those is they're very
- 17:44deterministic. And they're great for workflows
- 17:46where you know exactly the steps you want, but where you want
- 17:50intelligence like lead routing or where you want, you know, a
- 17:53significant scale like finding me the median time to close for
- 17:56all of my reps. That's kind of where
- 17:59deterministic behavior becomes very limited because something
- 18:02like finding me median time to close for my reps just doing
- 18:05that determinant, just kind of if you were to just write that
- 18:08workflow down on paper, it's just so complex so you can think
- 18:11through it that let alone you're going to put it on ZAP here or
- 18:13automate it and things like that, right?
- 18:15So those are the kind of things that ZAMS does really well and
- 18:18can allow any team that's wanting to include not AI, but
- 18:22more intelligence into their workforce.
- 18:24That's kind of what what our our kind of focus on vision is.
- 18:27Got it. So on the on the data side
- 18:29though, as, as like you're asking all these questions, I'm
- 18:32assuming it's like conversation. I could just ask a question like
- 18:35what is the median time to whatever, right?
- 18:38And then the the AI answers it like, do you, are you like
- 18:41pulling the data out of those systems or are you going into
- 18:43the system and calculating the data?
- 18:45It's more like going into the systems.
- 18:46You're going into the systems. OK, got it, got it.
- 18:49And and what? What are some challenges that
- 18:51you have encountered as far as being able to test the system?
- 18:56Yeah, great question. Every CRM is set up in a very
- 19:00custom way, right? Or Salesforce could have 50
- 19:03custom fields, someone else's Salesforce probably could have 3
- 19:06custom fields. And and there's like whole sort
- 19:08of stuff that happens. We've noticed a couple of more
- 19:10things like like there are some teams that use Slack inside of
- 19:14Salesforce. There's something teams that you
- 19:16Slack outside of Salesforce. There are some teams that have,
- 19:19you know, HubSpot calendar as their default calendar.
- 19:21Some teams use a call recording feature inside of a CRM called
- 19:25close. And then there are some teams
- 19:26that have like enhanced notes in Salesforce that they really use
- 19:29as opposed to just standard notes in Salesforce.
- 19:32There's just so much, you know, set up that each of these teams
- 19:35have, which is kind of one of the most beautiful and
- 19:39interesting thing that we've seen and that we've really set
- 19:41up for. So when I think of like use
- 19:43cases on Zams, it's really what I'm thinking is I'm seeing teams
- 19:47that would say, hey, I want to look at the picture that was
- 19:50added in the enhanced notes in Salesforce.
- 19:53And to be able to go to that level of depth.
- 19:55It can only happen when you have an agent that's very, very smart
- 19:57enough to not only go into your accounts, but really figure that
- 20:00out. So when it comes to challenges
- 20:03of of the challenges we're solving.
- 20:04We're often saying, how do you build an AI agent that's so
- 20:07smart that understands your CRM, your product, your business at
- 20:12the core so that even without you defining every single thing
- 20:16you have in your CRM, it just comes in, figures it out.
- 20:18And when you say, you know, who are my top reps, it knows
- 20:21exactly what to do, where to go, what to look at.
- 20:24When you say, hey, look at all the notes and tell me which
- 20:27leads are leaning positive for this quarter?
- 20:30It will actually go into those enhanced notes, follow all the
- 20:33links that you have, figure things out and get you the
- 20:35outcome. So that level of detail and
- 20:38depth is kind of what what we really, you know, care about and
- 20:42optimize for and that's what we are, you know, revenue sales
- 20:47stack focus as opposed to doing kind of anything and everything
- 20:49because you know this much focus in one.
- 20:52Place and and so you have like a marketplace of connectors or
- 20:55something like that that you already already connected to.
- 20:59So the way it works is users would come into Xams, they would
- 21:01connect all the apps and services.
- 21:02So that could be Salesforce, HubSpot, Gong, any of these
- 21:05tools that you use. And then you have two options.
- 21:09You can build an ambient agent. That's the type of an agent that
- 21:12works in the background. So you can say every Monday at
- 21:155:00 PM, give me a list of leads from last week, give me a list
- 21:19of all my AES, tell me which AES are performing better and give
- 21:22me some feedback to give to those AES.
- 21:24So that is an amid agent's work, just works in the background,
- 21:27slacks you the details when you need to, for example.
- 21:29And then you have something called a command agent that
- 21:31works on command. So you can say hey, I just need
- 21:34you know, total number of arr from this week for example.
- 21:39You could just type it in as a one off command.
- 21:42It executes and gets you the answer.
- 21:44So that's how it typically works.
- 21:46It sounds like, you know, they're even smarter than some
- 21:48Rev OPS analysts that we're hiring.
- 21:51You know, being able to figure out like what do we do in
- 21:53Salesforce and just trying to navigate that.
- 21:55I mean, those are difficult things, right?
- 21:56I think about bringing on somebody in the team like their,
- 22:00their, their days and like weeks before this person is, is
- 22:04familiar with how the systems works, right.
- 22:07So tell us, tell me a little bit more about what you think about
- 22:10that. Like what's gonna happen to my
- 22:11job? Oh, yeah, great question.
- 22:13I think a lot of the jobs moving forward, regardless of just
- 22:16sales and dev OPS are gonna become more of like we as people
- 22:20are gonna become more of orchestrators.
- 22:22And the biggest challenge that's gonna happen is for us to figure
- 22:26out how do you align strategically with the goals
- 22:29that we have as a team and what can we think of in terms of
- 22:32ideas? We want to orchestrate the, the
- 22:35concept of going in and navigating your CRM and clicking
- 22:38buttons and, you know, copy pasting stuff from Gong and all
- 22:42that's going to go away because that's something what, what
- 22:44agents are going to do really good at.
- 22:45And what we've seen is, you know, a lot of teams we work
- 22:48with, when they hire their Rev OPS teams or when they hire new
- 22:51GTM engineers, they're heavily focused on is the Rev OPS
- 22:55person, is the GTM person that I'm, I'm kind of interviewing.
- 22:58Are they someone that can be very much aligned with our
- 22:59strategy and they can really execute from that standpoint as
- 23:02opposed to, you know, hey, does this person know XYZ tool stack
- 23:06and can they can navigate that? And so I think that's going to
- 23:08be a bigger, broader shift that we're going to see.
- 23:10So it's not going to be that the jobs are going to go away
- 23:13anywhere. The jobs are going to just
- 23:15change in the way they think about execution versus thinking
- 23:18about orchestration. And I think that's going to be
- 23:20the biggest shift we're going to.
- 23:21See in the world, what do you think?
- 23:22What is something that maybe like the customers always try to
- 23:26build in terms of like automation maybe maybe they've
- 23:30tried to do it with AI but it always breaks until they use?
- 23:35I think research is one of those things that that I found has
- 23:39been limited at scale. So it's it's one thing to say,
- 23:41hey, find me everything about Laura.
- 23:43I'm about to talk to her and give me some icebreaker
- 23:46questions. But it's another thing to say,
- 23:49find me buying signals for Laura on a specific product and cite,
- 23:53cite those signals and tell me where those sources are coming
- 23:56from. Give me the output now.
- 23:58Do it at scale. Do it for every lead that's in
- 24:00my CRM. Do it for every top 10
- 24:01enterprise account that I have. That sort of level of research
- 24:05is something that everybody really wants.
- 24:07But it's very limited right now because of multiple factors,
- 24:10because of the ability of AI to have the right kind of access to
- 24:13the information, come up with conclusions, but also the
- 24:16ability of the AI to bring real meaningful conclusions.
- 24:19Like it might see that, hey, you have a job opening for a
- 24:23designer and a design company that might be using AI would
- 24:26say, cool, you know, Laura's hiring a designer.
- 24:29That's not necessarily true just because you have a job opening
- 24:32that mean you have now an in to make your pitch.
- 24:34But if you see that, like let's say the AI agent looks at a
- 24:38signal using built with and it uses built with to find all
- 24:41these signals and it starts to see, you know what, your design
- 24:43stack changed three times in the last six months, Chances are you
- 24:47might be having challenges with design.
- 24:49Let me use those signals. So for an AI agent to really
- 24:52identify the right kind of signals and bring conclusive
- 24:54outcomes on it, that's a little bit limited right now, but I'm
- 24:57sure that's only going to be a matter of time where it gets
- 25:00better. So I'm hearing like it's not so
- 25:02much that the automation rates, it's more so that the level of
- 25:06intelligence that is being brought back by the AI agent is
- 25:10a little it's more sophisticated and you know, you can do more
- 25:13with with with it, you know, when when you use EMS, Right.
- 25:17OK, well, and yeah, go ahead. Oh, I was just going.
- 25:20To add and one other limitation I've seen is prompt engineering
- 25:24is a very new concept and it's very different than your if this
- 25:28workflows as well. A lot of people don't think of
- 25:31prompt engineering in the same way.
- 25:33Often times we would see something like a customer would
- 25:36use DAMS and they would say give me my top reps and the AI would
- 25:40go into the CRM, it'll figure out what the word top means.
- 25:43And for AI, it might be just the number of revenue closed, right?
- 25:46And I'll just give you a list and the customer would come in
- 25:49and say, you know what? I don't want just top based on
- 25:51revenue, but I want based on the win rate that they had, not just
- 25:55the amount of revenue. OK, great.
- 25:56So you got to define that to an AI or sometimes they would say,
- 25:59hey, you just give me a list, but I want a table of all the
- 26:03reps and like all these things, in which case, you know, you
- 26:06have to tell the AI that give it to me in a table.
- 26:09And so often times when it comes to prompt engineering, there are
- 26:12a lot of things that we just as humans assume the system will
- 26:15understand and know and just do. And in most cases, the system
- 26:19doesn't or it'll like do something that's more relevant
- 26:21to itself. And so I think prompt
- 26:24engineering is also one of those skills that is very important.
- 26:26And I feel like they have absolute leaders who kind of get
- 26:29used to it. And it's not super hard to kind
- 26:30of pick up. But if you pick that up, it
- 26:32changes the game a lot because now you can really mend AI
- 26:35models in the right way and the way you want it.
- 26:37So that's something. Let's pivot a little bit.
- 26:41Norman, I wanted to ask you a little bit about you building
- 26:44Zams as a company. What do you think was your
- 26:47driving factor in building Zams and maybe what is something that
- 26:52you chose deliberately not to build?
- 26:55Great question. My personal background is I I
- 26:58grew up in India. After high school I didn't know
- 27:00what I wanted to do, so I took a gap here.
- 27:02I built a bunch of different apps, taught myself how to code,
- 27:05ended up at A at a college here in, in near Boston, you know,
- 27:10studied engineering and behavioral economics.
- 27:12So hey, I'm a machine learning engineer by training like that's
- 27:15what I've learned in my life. What happened when we started?
- 27:18Obviously, obviously I was this complicated piece of software
- 27:21that allowed any machine learning engineer to build their
- 27:24own AI models in the most your own complex neural networks and
- 27:28and go really deep and heavy into it.
- 27:31What I learned starting that company was I, as the founder of
- 27:34the company, had to pick up sales myself.
- 27:37And so going from someone that was machine learning engineer
- 27:39all their life to selling to enterprises and selling to
- 27:43Fortune 100 companies and selling contracts, you know,
- 27:45close to half a million, $1,000,000, it started to kind
- 27:48of build Adna in me that was very different than what I had
- 27:51always been used to. And as that DNA started to form,
- 27:55it was very clear that there were challenges I had with my
- 27:58own sales process, with my own hiring process, with my own
- 28:01systems that I was like, this needs to be solved.
- 28:05So when we started to work on Zams, it wasn't about let's
- 28:08build a cool AI agent product for a bunch of developers
- 28:12because trust me, there are tons of companies doing that.
- 28:14We don't world doesn't need another dev focused AI agent
- 28:17company, But I wanted to build something that can truly solve
- 28:20the problems that I faced when I was going out and selling.
- 28:23I didn't want to do a bunch of like CRM updates, but every time
- 28:26I had to do quarterly reporting, I found myself wishing the CRM
- 28:30was updated, right? Like I didn't when we hired our
- 28:34reps I didn't want to always figure out a system that would
- 28:37route the best leads, but I always wished that there was
- 28:39some way to do it rather than randomly assigning them.
- 28:43And so that's where the idea started was can we build AI
- 28:47agents that solve real problems? And the most real problems that
- 28:49I was deeply connected with were the sales problems.
- 28:53And that's where it really started from.
- 28:55Very cool. OK, what's in your own graveyard
- 28:59of Rev OPS automations? Things that we wanted to build,
- 29:02what we did not build, there is a ton of those things, a ton of
- 29:06those things. One of the things that we wanted
- 29:09to build was ACR, sorry, was a CSV based agent, so a
- 29:14spreadsheet agent of sorts. And the idea is that the
- 29:17spreadsheet agent would actually go through your spreadsheet,
- 29:20which could be leads that you've talked with, emails, things like
- 29:23that, and does activity for each one of them.
- 29:26So you would go ahead and say, hey for here's a list of leads I
- 29:29spoke with. I met at this event
- 29:30automatically go through each one of these research and send
- 29:33out an e-mail to them for a call with me next week.
- 29:37We decided not to build that because as cool as it sounds,
- 29:39right, it's not a Rev OPS problem.
- 29:42It's, it's a nice to have a problem for an account executive
- 29:45and you know, they might be able to put a second or two in it,
- 29:48but A's are busy selling. They have very little time to
- 29:53even do something like this. And it's not a, it's not a Rev
- 29:56OPS problem. So we decided to not do that for
- 29:59for a second. Another thing that we also
- 30:02shelved was on a very similar line of CSV agents was LinkedIn
- 30:10connectors. So there's this really cool idea
- 30:12where, hey, what if I could connect to my LinkedIn and you
- 30:15can just surface all the leads that I could target and there
- 30:19could be a good score that we can give to them.
- 30:21And it sounds like a really cool idea.
- 30:23I would love to have that personally as well, but we
- 30:25decided not to build that immediately.
- 30:27We we most likely would pick that up at some point in the
- 30:30future. But again, it's not a Revos
- 30:33problem, right? It'd be nice to have system that
- 30:35can connect to your leads, but it's not something we decided to
- 30:38build almost immediately. So there are a bunch of these
- 30:40things that we are personally very I'm personally very excited
- 30:44about, but I had to put heavy stone in my heart and say we're
- 30:47not going to build it today. We might think about it later,
- 30:50but not today. And that's these are the ones
- 30:52that are showed. Awesome.
- 30:54How about 1 Rev OPS task you'll never let an agent touch.
- 30:58One Rev OPS task, I'm not going to let an agent touch.
- 31:02That's tough, I don't think. I mean, I would want my agent to
- 31:05touch a lot of my Rev OPS tasks. I think the strategy part of
- 31:10like what piece of information am I going to use while I'm on
- 31:14the calls? That's not something I want an
- 31:15agent to touch. I don't want the agent to
- 31:17interfere when I'm doing my sales calls.
- 31:19I don't want the agent to kind of come in and and tell me do
- 31:22this, do this, do this, even though it sounds very cool.
- 31:25I feel like every AAE has their own kind of process that they
- 31:28really kind of walk through. And that process is something
- 31:32that's really true to you. It's very human.
- 31:33That's the one that really builds these relationships with
- 31:35people. And if there's one part in sales
- 31:39where I wouldn't want an agent to interfere, it would be the
- 31:42relationship building part. And I think that's better.
- 31:46And so like the human aspect of things like keeping that,
- 31:48preserving that with female humans.
- 31:51OK, OK, What's your favorite It just worked moment?
- 31:54It had Sam's? It just worked moment.
- 31:57There's so many of them. I think the biggest 1 was
- 32:01metrics reporting and this is again going back to very similar
- 32:04metric example, but find me the median time to close for each of
- 32:09my Rep and slack me the details. When we onboarded one of our
- 32:12first set of customers and they were kind of very excited to try
- 32:16this on. We found that like when he typed
- 32:19this and he pressed enter and it took a minute and when it just
- 32:23worked and he got a Slack message, he was blown away.
- 32:26It was like this magic moment that he had like never seen
- 32:29before. And that was like, not only the
- 32:31moment we had a customer, but it was like one of those moments
- 32:34where they're like, holy shit, like, how can this happen so
- 32:38very well? And he was blown away almost
- 32:40immediately. His first reaction afterwards,
- 32:42let me just double check the numbers.
- 32:43So he opens up his, his HubSpot, he opens up like the Slack
- 32:46message he got. He's like discounting and double
- 32:49checking, going through a spreadsheet.
- 32:51And then I'm just staring at the screen for like 2 minutes.
- 32:53I'm like, let's, you know, let's go through this.
- 32:56And then two minutes later he's like, wow, this just worked.
- 33:00This was just out-of-the-box, did the job.
- 33:03And I think every time we onboard a customer, we see that
- 33:07reaction the 1st 1520 minutes, like once you connect your own
- 33:10data and you just kind of say go do this and systems does it and
- 33:13more importantly does it right. That's kind of the main thing.
- 33:17Because a lot of times what what we hear is people see a bunch of
- 33:22decks all the time about AI agents and they're not excited
- 33:26about that. People see a bunch of products
- 33:28when they use it for themselves, it doesn't work.
- 33:30And that's why we've like taken the conscious decision to hyper
- 33:33optimized for HubSpot, Salesforce and Gong.
- 33:37And we're really saying, look, if you're using any of these
- 33:39three tools and you're going to be using Xams, you're going to
- 33:42have a mind blowing moment because it'll just work.
- 33:45It's just dead simple to say it and it does the job.
- 33:49And that's kind of the key experience a lot of our users
- 33:52get. OK, fantastic.
- 33:53OK, last question for you before we wrap complete the sentence in
- 33:572027, the Rev OPS dashboard will.
- 34:03Run itself. So that's the the most obvious
- 34:04one that comes to my mind yeah. I don't think we're going to
- 34:08have a world where Rev OPS leaders are navigating through
- 34:1050 different tabs copy pasting stuff in sheets and macros and
- 34:15kind of going into the weeds every day.
- 34:16We're going to have a world where Rev OPS leaders are saying
- 34:19this like they're identifying problems for the AES, they're
- 34:22building the strategy and they're building the road map
- 34:24for where the company should go from an operational standpoint,
- 34:27from an operational efficiency standpoint.
- 34:29They're not clicking much tabs, so will will run itself most
- 34:32likely. I think you just to to add on
- 34:35that I think the Rev UPS dashboard will like disappear in
- 34:39that, you know, all the questions that you ever want,
- 34:41you can just answer. You don't even need a dashboard
- 34:44in that sense, you know? Cool.
- 34:46Norman, thanks so much for joining us.
- 34:47Is there anything that you want to leave our audience with?
- 34:51No, it's super exciting times to be in and if you're in sales and
- 34:54you use Salesforce Hotspot or bong you they should definitely
- 34:58check out sales. Awesome.
- 34:59Thank you so much. And to our listeners, if you
- 35:01enjoyed this episode, please let us know like or comment on our
- 35:04episode and so we can give you more.
- 35:06See you next time. Bye bye.