Your Pipeline Is Lying to You. This AI Fixes It After Every Sales Call

Anis Bennaceur: Coaching is a vitamin. People wanna buy painkillers, and the biggest painkiller was actually filling the CRM.Sam Nadler: Built this week, breaking it down. Built this week, we show you how. A fresh idea, a clever tweak you locked in. You built this week. Hey, everyone, and welcome to Built This Week, the podcast where we share what we're building, how we're building it, and what it means for the world of AI and startups.I'm Sam Nadler, cofounder here at RYZ Labs, and each and every week, I'm joined by my friend, business partner, and cohost, Jordan Metzner. How you doing today, Jordan?Jordan Metzner: Hey, Sam. Happy to be back. Another exciting episode. Lots going on in the world of AI, a lot of crazy news, Anthropic filing a IPO, and obviously lots of other crazy news. And, yeah, obviously, looking forward to this week's guest as well.Lots to talk about there.Sam Nadler: Yeah. Looking forward to introducing our guest. But before I do, please don't forget to like and subscribe. I think we've approaching the 28,000 subscriber mark on YouTube. New episodes out every Friday with great entrepreneurs, CEOs.And with that, I'd like to introduce our guest CEO of Attention, Anise. Anise, if you don't mind, give us a brief intro, who you are, what you're doing, and then we'll shift to the demo that we built.Anis Bennaceur: Hi, Sam. Hi, Jordan. Excited to be here, and thanks for having me. I'm Anise. I'm the co founder and CEO of Attention.We're a series b startup, about 70 ish employees. Our mission is really simple. We automate increasingly smarter work for sales teams. We work with companies like Scale AI, Lovable, Bridge, Bamboo HR, and and hundreds and hundreds of other ones that are just incredible. And we just help them make more revenue through more sales.We automate this work for their sales reps, for the managers, for the leaders, and for their enablement and revenue operations teams.Sam Nadler: Awesome. So we're gonna dive into attention pretty deeply. But before we do, I wanted to kinda highlight something I built with attention in mind. So just to give a little bit of context here, and let me provide the the warning that this may be completely off base and I did a horrible job. But Anis did open source his entire sales coaching stack, which is wild.And so I spent the day trying to build on top of it. So what the tool does is it scores agent conversations one conversation at a time. And what I tried to do was take the next step, which is take the whole pipeline and leverage a Monte Carlo forecast to put it head to head, the forecast against what the team committed to in the CRM. So and it came up with this dashboard. It could be wrong because I don't have that much familiarity with with your tool and I did my best.So we can we can kinda run through, poke through this dashboard. It's all fake data. So here the big thing is, you know, the AI forecast is 1,170,000. Obviously, a little difference between what was committed, the 1,440,000. You know, you can swing down to a couple deals like Helios, Northstar.You can click into it, and it has, a a promising, you know, a strong deal until the board level vendor risk view gate was introduced two weeks ago. So that's, like, a new thing that, has to be overcome. What I also thought was fun is it it gave us this coach a live call. So you got a live call. I'm gonna skip to the end so so we don't have to watch through the whole thing.But you've got the rep, in this case Jordan, the prospect, and it's as the conversation is going, it's scoring how the the agent is doing in this particular call. It got a 67 out of 100. We can also walk through kind of what's going on in general. So you've got just any recording thing where's attention really understands, you know, who the rep is, who the buyer is, who the role, CRM linked. It gives you these coaching reports, the scores, And then lastly, I tried to build on top of that this this, like, forecast using the score.So let me pull down here. Let me see if I here's some things. What happens in these scenarios, you've got a macro tank and your CRM changes based on, you know, what was the forecast versus the CRM committed. Obviously, wouldn't wanna see that happen. If this Helios deal falls through, we also have a bigger bigger tank versus the delta.What the CRO actually wants, you know, as minimal delta from the CRM as possible. Anyway, long story short is it's a demo. Wanted to, like, provide, you know, what what is maybe an additional product feature for attention. Would love for you guys to pick it apart. What's stupid here?What works? Watch the math. Beep beep beep. Running it. Running the Monte Carlo replay to get the 10,000 trials to try and understand based on the conversations how close we are with the forecast.Anyway, I'll pause there. Pick it apart. Let me know your thoughts, and then we'll take it from there.Anis Bennaceur: Sam, this is awesome. I love it. Just for context, everyone, I I released actually on Monday a an open source version and obviously dialed down version of some of the things that attention can do. And so I'm pretty impressed with what you did here in such short amount of time. So the things that I'm pretty impressed by, so the whole Monte Carlo simulation, we we've never done it.We've never even thought about it. I've actually done some of that in for some of my own outbound automated outbound piece to know if a message should be a go or a no go, but on this on the forecasting front, that's great. I love that the forecast is actually relatively close to what you have committed in the CRM, right, which is kind of like what most companies see. I love that it's using the the frameworks and the coaching and and the understanding of the conversations to tell you whether a deal is at risk or, you know, promising or so I would actually, for the band, use more sales related terminology, like, is this deal committed or best case or just, you know, actually at risk, which you have here. The actual if I look into some of the feedback, the feedback does sound very AI ified.I know that I released also a humanizer a humanizer that makes the text look less AI, less of an AI. But actually, as early as today, I updated the humanizer just by telling it right like Paul Graham, and now it it writes like a real human. So that's great. But overall overall, yeah, this is this is awesome. I think this is just one part of what leaders care about, which is the forecasting.The other piece is now that you have the information, what can you do about it? Right? Which is what attention does really well. It's giving you a list of proactive actions to save your deals. Right?Let's say, for instance, you know, that on a deal, the actual, you know, CRO or or VP of revenue that you haven't talked to is the final decision maker and your deal is at risk. You need to multithread your deal, and you need to reach out to them to save that deal. Right? What are the the what's the the probability that that person responds relatively low, but still you gotta shoot all your shots and you gotta do it? So and reps generally don't do it.So if you give them the ability within a deal, hey, click this button, it'll draft a message and all you have to do is review it and send it, that will actually help your reps by quite a bit because you're going from not just not just forecasting, but actually improving the forecast.Sam Nadler: Working out is easy to start, hard to stick with. HipTrain fixes that. Real coaches, real accountability, a plan that actually fits your life. No guessing, no skipping, just consistency. HipTrain.Amazing. Well, let's jump into attention overall. I know we quickly covered it, but give me that, like, high level. You've mentioned what attention does really well. You know, it sounds like it's coaching.It sounds like it gives you the next step, but, like, I would love to hear from you. What is, like, the core value prop? At least when you began, and I know that can evolve or expand over time, but, like, give me that need that you had to solve in order to make attention grow to where it is today.Anis Bennaceur: Yeah. I'll tell you how we started and then or or what made me start initially, and then what we got right over time, and just how attention simply works today. The first thing is, you know, in my the last company that I founded, it was Bootstrap. And so every single deal mattered in the limited space that we were selling into. And so I thought to myself, hey.What if we could capture these conversations in real time and tell the seller what they should be saying during these calls so that they optimize their conversion rate. Right? In other words, just one of their deals. So we started building that, right, just like a real time coach. I'm sure you've seen Cluely, you know, in all these other apps in the past year, but we were kind of like the first company back then to do real time coaching based on these conversations.Think think of it as kinda like your GPS. And our whole value was, you know, comfort the the coaching and stay for the productivity, which meant basically you would come to get coached. And once you're ramped up and didn't really need any of the the coaching in real time, the productivity would mean everything related to after your calls, making you more productive, preparing you for your next conversation, and making sure that you're actually just winning more deals. We realized, you know, after releasing the initial coaching feature that within a few weeks, users would not care about it. Buyers would not really pay for that.This was kind of like right after the ZERP era ended, and so it was really hard to sell any of this. So I was I was sitting over lunch with another another founder, and he tells me, hey, why don't you flip it around? Why don't you just make it come for the productivity and save for the coaching? And at the same time, someone else was telling me coaching is the vitamin. People wanna buy pain killers.And the biggest pain killer was actually filling the CRM. Right? I remember actually having dinner early dinner with a great investor, and he tells me when I told him, oh, you know, we pivoted from coaching to auto filling the CRM. He looks at me, and he's like, oh my god. This is gonna be a billion dollar company.Everyone hates filling the CRM. That was so obvious to me at that point, you know, in hindsight. At my last company, our CRM was never filled. And so we started with a really simple value prop at that point, right, which was we capture call recordings, we automatically fill your Salesforce, and then we started also working and integrating with HubSpot and so on. But the whole point was, you know, just auto filling your CRM, writing your follow-up emails at the end of your calls.And what I actually realized was I was using that product every single day, multiple times a day, and I loved it. Right? Whereas I I wasn't really using the real time coaching. I didn't really need the talk tracks and the battle cards. And so if you start building something that is really useful to you, that it's going to be extremely likely that it will be very useful to other people.Over time, right, we started including other forms of messaging. So not just calls, but also emails and, you know, chat messages and also Slack messages and so on. So increasing the number of inputs. We improved also the brain behind attention. Right?A lot of tools were doing very horizontal things. We just stayed focused on go to market and actually more narrowly sales. To build the best possible sales brain, we were the first company in 2023 that was able to analyze thousands and thousands of conversations at once. And then the last piece is, you know, in 2024, in addition to just filling the CRM, we started serving this information where you need it, when you need it, for the most important people that need it. Right?And so it could be in Slack. It could be an email report to your leadership. It could be, obviously, in your CRM, but also many other places. Right? And so we would capture all the information that matters the most to anyone and orchestrate it wherever it's needed.Today, actually, we have an entire proactive section that allows you to that just distills all the the the massive info that we gathered about your go to market that we ingested, that we processed, and then we would tell you as a sales rep proactively what are the next best actions for you to best execute on your execute on your deals and win them.Jordan Metzner: Awesome. Awesome. Okay. So alright. Well, that was I mean, that's a lot.But obviously, you kinda have some experience of, you know, building this into like what you need. I think nobody's CRM is up to date when it was had to be done fully manually. So I think, you know, obviously, like, bringing things in, like, email and all these other things into the CRM was just, the beginning of automation into those flows. You talked about some of your customers, which are, like, some of the leading startups in AI right now. And, you know, not to go into anything confidential, but maybe you can tell us a little bit about like, you know, what was the problem, what was their pain, how did attention come to fill that, and kinda how are they using it to win more deals?Anis Bennaceur: Yeah. Initially, everyone would just come to us with a very simple pain point, which is our CRM is always empty. And it's a nightmare when we need to pull any sorts of reports. Right. That's been the case.Even we see that all the way till today. I spoke yesterday with one of the fastest AI native scale ups at this point. It's like one of the biggest darlings in Silicon Valley. And I I could've, you know, talked about I was I actually did mention the agents. I I mentioned a lot of things, but the one thing that the buyer really cared about the most was, oh my god.You can fill our Salesforce? Yes. We can. Right? Now we have a we have a lot of buyers who have AI mandates and are looking to be agent agentic solutions and proactive actions and next best steps and and automated reporting.But the biggest biggest biggest thing is how do you feel our Salesforce. Right? And that's why people still come to us every single day today.Jordan Metzner: Okay. That's awesome. So and by when you say fill out their Salesforce, that means you're doing deep integration with their mail, calendar, phone, etcetera systems, and then using that with AI so that all those details are then reported directly into the CRM?Anis Bennaceur: Yeah. That's correct. It's not just a summary. Right? You you can do a summary, but you'll miss a lot of information.We'll fill every single field in your CRM. Right? And so let's say, hey. I wanna capture the dates of the competitor contract renewal. Right?Boom. It automatically gets captured and put in there. Right? Imagine when your your SDR is cold calling, like, one of your prospects, and the prospect responds, hey. We're still in contracts with your competitor for the next six months.The chances of your SDR going in and filling that information are close to zero, but Attention will capture that and automatically distill that and and fill your CRM based on that. You wanna know what is their tech stack. You wanna know, you know, information around, like, you know, their spending habits. Anything that you really care about that you wanna store in there, Attention will do that work for you. Right?And there are so many millions of of errors and edge cases that can happen that we spend the past three plus years just deeply, deeply, deeply making sure that things would work there. Obviously, you know, the alternative is you go in and buy an AI native CRM. But every single time I talk to a buyer who used one of these AI native CRMs, they hated it. Right? They end up having a new revenue leader that comes in and says, hey, forget this tool.I won't name any, but forget this tool. We're actually gonna switch to Salesforce or HubSpot. Right? I've seen that happen over and over again. And so at attention, we don't care about building a CRM.We don't wanna build a new CRM. We wanna focus all of our engineering efforts into building the best possible actions that for your entire team that are more and more insightful and increasingly smarter.Jordan Metzner: That's awesome. Okay. Cool. I think we like kind of hit it over the head a few times here now. That's super cool.It sounds like some of the you mean you're working with some of the best companies. So obviously, that's gotta be pretty fun. And series b also, congratulations. Let's see. Maybe we should jump into the news a little bit, Sam.Sam Nadler: I thought this was topical more as a discussion between founders, but, you know, Uber caps employee AI spending after blowing through budget in four months. You know, it's been a topic of discussion for us in the last three months. We've had a couple weeks where we've had a few surprises, and we're we're also grappling with, like, you know, we're we kind of lead an organization where we want our engineers to have pretty much full access, unlimited access, and just let them build. So, you know, it's been a co topic of conversation in the past sixty days for us, and I would love to hear hear your thoughts, Anis.Anis Bennaceur: Yeah. I can start with a story, and then I can tell you how we're thinking about it today. But that does not shock me, and I totally agree with what Uber is doing here. So that's a bit of a hot take. The first thing is, you know, I I wanted to measure the productivity of our engineers.This was, by the way, you know, in 2025. And so what I started doing was looking into their cursor, you know, cursor their cursor tabs. Right? Like, how how much they would actually commit in cursor using AI. And so I started looking into all the engineers.And I noticed something really interesting, right, which was the people who were actually spending the most money on AI on Cursor were either the bottom engineers or some of the top engineers and no one in the middle. And, you know, obviously, it's very it makes a ton of sense. At the end of the day, the top engineers who have best judgment will do great things, and the the bottom ones will just create a lot of slob. And so we end up having over here at attention this two by two matrix that's really simple. It's a, how good is your judgment, and b, how AI advanced you are.Right? And so if your judgment is really bad obviously, I mean, you're not gonna stay here for too long, but if your judgment is really bad and your AI build, it's one of the worst combinations ever because you're going to do a lot produce a lot of slop. You're going to make a lot of, like, poor decisions. You're going to go into a lot of the wrong directions and build stupid things and build another fifteenth dashboard that is completely useless and will have spent all that money for nothing. Whereas, like, if you have very good judgments and you're AI advanced, you're gonna build incredibly smart things.You're not gonna reinvent the wheel. You're you will actually just build the machine that builds the machine, and that's where we want the the token usage to go. Right? And so in our case, it's a bit arbitrary and some people might hate it, but we capped everyone at, I believe, like a few hundreds of dollars additional to what they normally use in Claude. The the the top people, and that's purely subjective, we will actually give them unlimited spend on Claude, but everyone else is capped.Right?Jordan Metzner: You know, I don't know if I agree with the Uber CEO or with Anise. I think, you know, I understand kind of capping it from the sense of kind of limiting your max outbound pain or spend. But the real question is, is like, how much tokens should a developer use? And, you know, I think that's like I think I think you're totally right about like, you know, I'm a crappy developer and I use a lot of tokens. And we see like, you know, our our best developers also use a lot of tokens and then some in the middle.There are some that are, like, top developers that don't use a lot of tokens actually that we've seen also that type of behavior where, like, they use it, like, more infrequently and only in the cases of, like, probably, like, performance improvements and things like that. Like, less, like, kinda core writing, I would say. But I I guess the point is is that, like, you know and this all comes from this Uber article in the first place, was that, like, they they blew through their token budget. Question is like, when do they mail the token budget? You know, these models came out in, you know, December and now we've seen new updates.And I think the other thing is that like, we don't know how we're spending using these tokens. They don't know how we're spending. I mean, I think like, we've seen from OpenAI, like, the codec dashboard has changed like almost every day for the past like fourteen days. So I think they're also trying to figure it out. And we're in this like kind of wide open space where no one knows the true cost of like tokens, how much tokens you need to use.And then the models are getting better, faster, cheaper with like more, you know, features inside the harnesses. Right? So, you know, is like using sub agents more efficient than like running like, you know, multi threads. Right? Like is that an efficient gain and that's a better way to use tokens or should sub agents never be used because like it's an excessive use.And, you know, I didn't go back and forth on this, and I've probably been rambling a little bit. But, you know, the codecs computer use feature over the past few weeks, like, definitely killed us on spend on tokens. And it took us like two and a half weeks to just figure out what it was that was causing all the spend. So does feel like the early days, almost kind of like when Internet was metered, and you could only get so many gigabytes of Internet on your phone or on your house, you know. And I don't know.I mean, the people that just kinda let it rip end up kind of taking a little bit of a competitive advantage. But anyway, that's kinda where I feel. But, anyway, I'd love to hear your guys' feedback.Sam Nadler: Yeah. The only thing I'm gonna say is, like, I do feel like it's early days. It Anis, maybe you're more sophisticated than us, but it did take us about a week and a half on a few occasions to really trace where the spend was coming from. It wasn't very transparent on their dashboards. You know, was it API usage?Was it did the billings, like, where the charges were coming from were all coded differently and, like, they were the where there was the console, there's the codex console. There's like, there's it took, you know, a lot of time just to piece it all together, and that was frustrating. But, you know, as Jordan mentioned, you know, we're we're are, like, debating with, like, is it time yet to try and, like, control their usage, or do we just deal with it and just monitor it? And if something gets out of control, then then we'll, you know, rein it in. But in the meantime, you know, it's an it's an ongoing discussion over here.Anis Bennaceur: Yeah. I think there there's two different things. The first one is how much your own product is consuming with these LMs, and then how much the second thing is how much your team is consuming for all other operations that are set that are outside of your product if your product is an AI uses AI. Right? In the first case, it is extremely hard to kind of, like, track things properly.We have an expound tier guy here at our team, John Liu. He actually spent a lot of time just building something that understand how much each client is spending or or costing us from an LLM perspective, how much each feature is costing us from an LLM perspective, and then how much each feature per client is costing us from an LM perspective. So we've been able to track this really well now. And so that allowed us also to do a lot of optimizations on, you know, like, which model to use and which cases and, you know, from a latency and and and intelligence perspective. The second thing is from a, you know, non product perspective.So from our our operations perspective, this is where the cast are helpful. Right? Because Claude we're we're an anthropic shop here. Claude will actually tell you within, you know, its UI, this is how much you have for the day. This is left for the day.This is how much you have left until, you know, your usage resets. And then if someone wants to use more, then they'll make a request. And either me or my cofounder will actually review it. It's not even one of the managers. It it comes through me or the my cofounder, and we subjectively approve that.One thing that you will should be thinking a lot about more is, you know, everyone's jumping on the bandwagon of of agents. The reality is in lot of cases, workflows, just deterministic workflows tend to work as well, if not better, than stochastic agents. And so what you wanna do there is just, like, let your workflows rip. It will not, you know, it will not incur any inference cost because it's just step by step things. Yes.Maybe some LMs will be called, but it there there won't be the the heavy thinking that agents will do. Right? And so if you can actually automate a lot of your business through worthless rather than agents even in 2026, then I think there's a lot of gains that you can have there.Jordan Metzner: I totally agree with you. Every time I go to build something, I start with, okay, I'm gonna try to use the agent SDK, like, through OpenAI or Clod. And then as I, like, reason through it, like, we always create some type of deterministic workflow and use, like, you know, just more of a classic LLM API calls and just kind of walk it through step by step to get the results rather than hoping the agent will do it. I honestly haven't found one use case where like we found the agent SDK to be useful for us other than building another agent SDK type of tool. Right?So, yeah, we're totally aligned. But, you know, it's only it's only one update away from seeing some some step function improvements here. But alright, Anis. Well, this was a really fun episode. Great to chat.Love to learn about Attention. Obviously, you have a great domain name, so tell people how to find you.Anis Bennaceur: Yeah. Attention.com, best domain name in the world. You can also add me on LinkedIn. Message me. Tell me that you've seen me on this podcast.And yeah, I'll I'll respond there.Jordan Metzner: Awesome. Okay. Great. Alright, Sam. Well, this was a great episode.Obviously, another exciting week in AI and awesome to see some of the things that Attention is doing and definitely made me think a little bit about our SDR pipeline a little bit, what we gotta do to make some improvements.Sam Nadler: Likewise. Thanks, everyone. Thanks, Denise.