Transcription
Hello everyone. Thank you so much all for joining. I am Goresh. I'm the co-founder and CEO of Sibil.
Today we are essentially putting Sibel up against one of the long-standing tools in the sales B2B sales market which is gone. But honestly, what the key goal today is not to convince that we have more features. Feature comparison in the AI world are kind of pointless right now. Whatever is on the checklist for one person, everyone will have that same feature in about 3 months or these days if you wipe code, probably 3 weeks.
The more useful question and the way we think about how the stack is shaping up is how do you even evaluate revenue AI when it comes to features and product changing every single month? Like how do you tell real innovation from just a very great marketing platform? And that's essentially what we wanted to give you today. It's a way to evaluate any revenue AI platform and even to some extent any AI platform that will still hold up an year from now, two years from now. And then we want to stress test it live. We want to show Gong and Sibis side by side with the same deal, the same data and show why and what exactly we do and how that stands up from an output quality within CEL. Colin, do you want to kick us off with the poll first?
Yeah. Well, before we get to that, I want to give a special welcome. We had seven competitors sign up for this webinar. So, special welcome to those of you who are with us. I see folks from Gong and Attention. Attention folks, congrats on the series B. We're so glad that you would make the time to to join us.
Yeah. Let's, before we jump into the one for one comparison, I want to get a single question sitting in your brain. Here's a question. Is your AI, you think about the AI that you're running, is it progressively getting smarter with every single deal that you're running? I just want you to sit with that idea for a moment because I think the answer to that question is where the future of a lot of AI is headed. I think we're trying to go. It's at civil is your answer to that question will really, really matter.
Now let's move on. On that note, I want to talk a little bit about how we typically evaluate and think about tools like Gong and Cibble. So, if you've recently kind of bought or been on the buying committee for any sort of revenue AI recently, you've probably noticed something. Every demo is starting to sound suspiciously similar. You're hearing things like AI native co-pilot agents, forecasting, coaching, CRM updates, AI follow-ups, call recordings, MCPS, deal execution, context, intelligence. All these different logos and different websites and different brands, but they're increasingly all kind of making the same exact promises. And as Go sort of hinted, that shouldn't be super surprising that we're getting the spot where almost all of tech is has underneath it a lot of these similar foundation models and that therefore features that used to take a year to build can be shipped in a few weeks or even a week.
And so the question that comes up is if everyone is building the same exact features, how are these products actually any different from one another? And I don't think differentiation has gone away. I just think it's changed. What we believe makes these products different is something that has moved below the surface underneath the feature layer. So maybe the question is no longer does this tool have forecasting or does it automate deal execution and send follow-ups or does it update my CRM. The question is starting to become what makes this forecast or this email follow-up or this CRM update better or more intelligent than the one this other company over here is offering. And what you'll notice is that part of the reason we didn't go into a big feature battle here is Dong and Civil share a lot of overlapping features. But what we would argue is that what's under the surface is very, very different.
Here's the difference in kind of in one sentence. We are built on fundamentally different foundations which then produce fundamentally different outputs. So again, instead of comparing features, let's go beneath the surface a little bit and talk about the infrastructure that supports all of these things and makes their outputs intelligent or not so intelligent.
Okay, so Colin just made the case that the difference has moved below the feature layer. I will actually take you down there and showcase what is that differentiation like and what are the five layers. And I I promise this is the least boring architecture walk through that you have gone through this year because every layer as we think about changes what the product actually can do for you.
Okay. So layer one first is signals. So here's the thing about signals that surprises a lot of people. Tools like Gong and Cibila start in the exact same place with what signals, what data that capture that they capture. And I really want to be fair to Gong here. They capture calls. So do we. They integrate with your CRM, your emails, your calendar, the Slack. And so do we. It is the same raw signal that gets captured and it is also honestly every single vendor that Colin showed you will capture that same raw signal. So collecting data is just table stake these days. Nobody is really feeling there anymore. The interesting question for us comes as what happens right after that data comes into your own infrastructure your own system and that's where the two systems stop looking alike completely from one another.
So the first piece of this infra comes in the form of a very boring technical problem and that's our first fog but it is actually not really a technical problem we are solving. It's very much where the intelligence start. So let me make it concrete. So we are talking here about identity resolution. Picture one buyer across a six-mon deal cycle. Let's call her Priya. In Salesforce, she is called Priya Raalo. On calls, the rep just says Priya. Her email is some other address. A calendar invite says P. And somebody in the meeting notes or an internal conversation for them is she's just CFO. Now, Gong is great at linking activity to CRM records, but that is kind of the ceiling here for Gong. The understanding itself is anchored the records that it's looking at. We started from a very different question 18 months ago. Not what record is this, but who is this person actually? And what we ended up doing is we built one single persistent identity for Priya throughout her entire journey and interaction with your team. And this identity that we built for PI follows her every single place. So within email, within the CRM, within Slack, within Zoom, it's the exact same person. The system will know that it's her and it will carry that relationship forward 6 months, two years from now, even in a new deal. And once you have that, that is where things start clicking because now you're not just reading four different contact fields, you know exactly that Priya is your economic buyer, you know that she joined during the discovery phase, that she's missed the last two meetings, you know that she responds very well to ROI related arguments. And every workflow, whether it's forecasting, whether it's coaching, your CRM update, your follow-ups, every specific thing can reason from that exact same understanding. If your AI thinks one specific individual is actually four different people based on four different records that different four different systems are storing, everything downstream will start breaking a little, right? So you have forecast which is differently captured from stakeholder map from coaching from follow-ups, all of it. So fragmented identity means you will have a fragmented AI and that is something we fixed first before doing anything else.
Now that the system knows who everyone is across your entire deal journey. Next question that we answer is what does that mean across a lot of different activity with that same individual and across individuals. That piece is called the context layer. This is the second piece that we look at. Now I will caveat it with saying everyone in the AI world, every single company, every single person is now using the word context. And what they usually mean with context is basically we have retrieved a few extra documents before calling the model and we give that to the LLM as a prompt. That is retrieval. It is not context. Gong calls this a knowledge graph. It's basically your people, your account, meetings, email all connected into one single searchable network. That is useful. That does give you visibility. We think of context as something very different. And the simplest way I can put the difference is is something like a knowledge graph or just retrieve documents is basically connecting the records to each other. It's like data points which are connecting to each other. Context graph is connecting meaning of one entity and one kind of entity to another entity and another kind of entity. And what what does that mean for sales or in sales space is that the context that you imagine to define deals is almost never written down any at any single place. It's always in someone's head and you don't open up Salesforce and type, hey, Champion is losing internal sport or the CFO just quietly checked out. Your reps regardless of how much server you want them to will never do that. Right? Those signals, however, are absolutely present in the data. They're just scattered across different kind of calls, emails, who showed up, what did they talk about, how did people react, and the context graph itself is what stitches them into an actual read on the deal. So a John B, for example, isn't a single contact. He's the economic buyer who once he got access to the dress center went quiet and Sarah didn't talk about security or pricing or ROI. She raised an objection that was unanswered and it is now blocking consensus on the deal. That is the difference between just looking at John and Sarah as singular contacts with titles versus connecting them to actual meaning of what they went through. And by the way, you can't simply put call transcripts into claude of charges strategy ideas. This is where despite claude being a very high quality reasoning model, it only can understand what exactly you give it, what you hand it and your transcript or your emails is one set of conversation. It doesn't fully connect back to what is happening before or after and because of which you will start losing meanings and that meaning is typically set in people's head for the most part and so that's where civil is get smarter by understanding who everyone is and not just that but also what everyone is going through.
So that forms for us the next layer which we call as the institutional memory. So if I had a person that was a champion six weeks ago, did that person show excitement about certain set of features and then 6 weeks later when I reach out to them on email because the deal is stalled, I can actually understand what does what do they like and then I can write an email specific to them. That person is a detractor, that's a separate kind of email. Now what institutional memory is for us is once you have this kind of connected context, do you also know what typically to be sent to the champion based on the company profile and based off all the other thousands or tens of thousands of deals that your entire company is doing. Maybe your counterpart, a different rep, has already figured out what exactly is the most useful thing that you can say to a champion when the security is a blocker in the deal and that information on what that 99th rep said in a deal 6 months back is what forms as part of institutional memory. So the question here for this layer is out of all of the activity the interactions the context itself what context should be stored in the institutional memory and which can be accessed by the LLM when it's trying to do a specific task like forecast or follow up or building a strategy for a deal. That makes it really interesting because think of how your own brain operates like you can have every single interaction, every single email. You can have all of it searchable but the brain doesn't remember every sentence of every conversation you have ever had. You remember the parts that really define how you are thinking about a certain specific deal that really define okay I come across this competition multiple times. When I say this kind of line, I always end up clearing that objection. And when I don't say that line, I do not end up clearing that objection. Civil works the same way. After a deal closes, after a big activity happens, the system is continuously trying to learn, what really moved the outcome, which objections were showing up which were consistent and what competitive patterns got repeated, what coaching corrections actually worked in the deal process. And this is a big reason how the best reps playbook stop being for tribal knowledge that is in someone's head. It starts becoming a company asset. Your enablement people are trying to do this on a very manual basis re-watching recordings like understand like taking gut reactions talking to people but civil but that information doesn't go into the LM when we build the playbook in this way it is much more accurate much more nuanced and it gets built much quicker so it becomes the asset for a company which is really hard to replicate and this stores history. It curates what really matters within all of the deals, all of the conversations that you have.
Now once you have your system with these right information the signal the identity resolution the the memory and the context graph and then the memory the next layer is where it starts becoming in self updating loop. This is where both your context as well as your memory continues to get updated again and again and again. It's like a daily update and this is basically getting where the system gets smarter from lessons that you did not keep. Okay. So remember that polling question that Colin asked is your AI getting smarter with every deal. This is the layer where that either happens or it doesn't. There are other tools that learn. I want to be clear but the learning that they define is across thousands of companies. It is broad patterns. It is general best practices where if the rep speaks 40% the deal is 23% more likely to close things like those. What it does not encapsulate is what is the specific pieces that every single customer is essentially requiring to be done. It assumes that every customer is running the same that in all scenarios the rep speaking more is doing a bad job and the rep speaking less is doing a good job. We flip that question. We focus on what works for your company, for your product line, for your ribs specifically. And so every deal becomes a feedback loop into the context, into the memory, every single win, every single loss. And this means even the small stuff. So for example, if a rib edits a f an email which the AI generated, the edit and the original email which the AI generated, the gap of those is something which tells the AI for future that the rep likes to write the email in this way. So if my email generated was best regards, goes and I edit did to remove regards then the future edits will not contain regards. And this is big because you can essentially do this across your entire team. You can do this across your processes and you can do this across any AI artifact. And all of that feeds into your single institutional memory and the next forecast, the next coaching call, the next strategy and even the next email. And that one system now combined makes it such that it will accumulate your actual competitive advantage.
Now all of this stacks up in what we call an accumulated judgment. Identity, context, memory, and learning. None of those layers is a single product in and out of itself. But the combination of those makes the system super autonomous and much more powerful than any single individual component itself is. So essentially the way we think about it is every single forecast, every coaching recommendation is ultimately a judgment call. And the real question is what is that judgment based on? So with a tool like Gong or even claude your every request more or less is starting fresh. It will pull the current deal. It will figure out a window of recent activity. You ask it a general purpose like you ask a general purpose LLM to reason over it. That's good AI. Don't get me wrong but it is generic judgment and you will get generic responses on it. Now if you are below the curve then maybe generic is good enough and that works but once you kid into an organization you really need very specific judgment and very specific clarity on what is to be done and that's where civil will bring the whole stack to every single question every single action every single outcome. It knows who is in the deal which is identity. It knows what exactly is going on in the deal which is context. It knows what mattered in every deal before this one which was similar to this specific deal. That itself is memory and it knows like how this deal's outcome will impact future deals moving on and it will keep on creating these three loops which is your learning. If you are putting all of those together, the AI is not reasoning from scratch. It's actually reasoning from experience. Very similar to how a experienced RIB learns and reasons on top of a deal. This is why two products can have the exact same feature list from the outside and they look very similar but it will give you completely different answers in this age of AI. It's never the feature, it's everything underneath it. This is the same scenario which you would experienced with chat like different model layers on Opus Sand Fable and even the previous versions of chart GPD like it it's all chat box and it you ask it a question it will give you an answer but the answer quality varies significantly based on how good or bad the model is that's the whole architecture it's the same inputs has gone on the left it's the same outputs or features at least on the face of it on the right but everything in the middle is where the difference really comes in hits between identity the context the institutional memory and the long-term learning groups and because that middle is different the outputs aren't just generic AI or just better summaries they are actually high quality decisions that can be made.
I'll stop talking theory. Colin, show where the difference actually lies and how it comes.
Now, I want to recognize something that goes kind of already hinted at on its own. This idea of infrastructure isn't valuable, but it determines the quality of every single AI output that you run. The thing we're observing is people will run these tools side by side and they'll say cybles AI is just better. And everything go just laid out is effectively the why behind all of that. Now, we get screenshots from Gong users truly all the time. I get them in my DMs. We get them from prospects who are evaluating civil and they will show screenshots of they ran a query in Gong's ask AI and this is what they got. Now, we considered showing some of those screenshots, but it gave us a lot of kind of hairy privacy reasons. So, here's what I want to do. I want to give you a very basic real example of it is one that a customer reached out to me and walked through this of a prop that they ran both on Gong and on Cibble and how the two outputs varied based on that. So here's the example. This person was working a deal and we'll we anonymize the names and information here. This is not exact UX. This is just the idea. They said, "Hey, did this crea this this prospect did they ask for customer preferences?" And they ran the same exact prompt in Gong and Symbol. Gong and Syble have access to all the same input as Goish described. And Gong's answer was no. Priya did not ask for customer references during the meeting. And then Symbol's answer was yes. Priya asked for customer references during the call. Specifically, she requested customer references for LMS usage at scale. Why did we get these two outputs? There could be a lot of reasons. It could be that Gong's AI couldn't search it very well and didn't have the infrastructure from it from like the way the data is structured. It could be all the infrastructure reasons that we described there. The reasoning could be bad. Here's my hypothesis in plain person's terms. It's because Priya never specifically said the phrase customer reference. Instead, she would say something like, "Hey, do you have some other like LMS customers, you know, that have already rolled this out?" A human salesperson immediately knows, okay, if I want to win this deal, this person's going to need a customer reference and I'm going to need to provide that. That's going to be that's an action item for this follow-up. But that difference between just retrieving language, building building keyword trackers and understanding the intent is the fundamental difference between gong and civil. Gong is grounded primarily in the words and the semantics of the conversation itself. and just look strictly what was said. Civil is able to reason and understand what was meant both because it understand this person. It understand all the deals leading up to now and it understand the it knows who Korea is. It knows she's the economic buyer. It understands the buying stage. Again, just the sophistication with which it's understanding the deal is like your best rep might think about and understand the deal. Again, if your AI is only retrieving language, it's going to miss signal like this, which is why we get screenshots like this.
We wanted to again hop in here and show a bunch of different outputs directly in Gong and Cibble. There's some limitations about our ability to do that. And so to stay above board, I'm just going to show some of the ways that Stibble reasons. I'm going to show you some of our product. And there's an opportunity here. If you're currently on Gong, we regularly advise and will help Gong customers connect Gong to Syibble, pipe all the calls over and then just compare the two side by side. This is not something we're scared of comparing. We feel like we're on very firm ground on this comparison, but I wanted to show some of the actual outputs you get in Syibble. So, give me a moment to pull those up. I'm not going to cover Syble exhaustively. I'm going to focus on the AI and some of the reasoning that's here and just give you a couple different examples from different roles that you might imagine in the company.
So here is let's imagine someone from your sales or CS team is thinking about their expansion targets for for the quarter. I had said, "Hey, look at the account that you think this cybble thinks is ripest for expansion and create a proposal deck for that account to expand their civil usage across sales, marketing, CS, citing their spoken paying points, the direct quotes, benefits they've gotten from the platform, and other relevant factors." And so Cibil found a paying customer and then covered some of the functions that it could expand into and kind of the the stakes behind that motion and it put together this pretty in-depth proposal of how we could expand Syble across this team. This is something that is nearly rip and ready to send. It folds based on the industry that they're in the for customers in edtech. It pulls quotes from their team and all the collective kind of conversations we've had with them over time. It pulls from the gaps that they have stated and where they currently are using this tool and where they could expand and then it makes a case based on everything they've said about why they should consider expanding civil to more of their team.
Here's another example. We were seeing a lot of revops pop up in our conversations and I wanted civil to really think about our go to market motion. I wanted to think big picture, why we're winning, why we're losing. And so I said, "Hey, based on all of our active deals in the last 6 months and conversations across sales and marketing, build a slide deck proposal for strategic shift towards targeting revops in our sales, marketing, and outbound efforts grounded in findings from our close one deals." And so the deck gave kind of this highle summary who's actually buying based on our calls, some closed one proof, revops pain points, some gaps in our go to market, and then a 90-day road map all into a deck. Now you could get this plain text, you could access this in claude. The value here is that it is again deeply understanding and reasoning our deal. It's all the judgment that go talked about and you could access that in Slack with cyol. You could access that in clawed with RMCP not just by raw transcripts but like the actual reasoning all of those are accessible. So again here I was able to say hey based on your deals 60% of your acting deals have revops as a stakeholder. Here's some concrete examples of those people. Here's some proof from our close one deals again citing some of these exact people. So here's an example here of this head of revops was focusing on how we beat gong on CRM data and autosync. And that was that was again one of the reasons that this person ended up becoming a champion for us. Again, the the output here is not just that it's pretty, it's that it actually understands our business.
Now, let's talk more at like the ground level. You're an AE and you're working your deals. You have goals for this quarter. I just asked Sibble really basically, hey, pull the top 10 accounts from the last year that are ripest for re-engagement and draft re-engagement emails for each. And Symbol was really quickly able to draft a personalized engagement emails with subject lines that I could open and fire off directly from Gmail. So, here's one and another one and another one. But this one, look, a signal based on a Kaya contract ended. Sorry, outreach folks. There are real opportunities based on what these prospects have said and what they meant that Civil is identifying as real opportunities for engagement. Another example here is you know like I can say hey you know look at all the deals from last quarter that wanted us to re-engage in Q3 and create event invites for each of them. So Syibble actually pull those different accounts and I could open this in Google calendar to send that person an invite with a little agenda based on our conversation and you know I'm not a sales rep so silly me I realized I can't just put an event on someone's calendar. I need this as a message. So I asked Sibble, hey, could you draft some re-engagement emails for these accounts too and Syble did that. Here is again more and more personalized engagement emails written in my writing style that are built on the foundation of everything that has happened up to this point in this deal and re-engages where it actually makes sense. It's not just static generic re-engagement. It's really based on what that person has said.
Now, here's a prompt that's a really interesting one. It's going to be overwhelming at first, but I think it's relevant. So Elaine Velby is a CR of Tofu. She's one of our longtime customers, major Sibble power user. She has a really basic prompt that she runs in Cibble that that fires to Slack into her email every Monday morning. And it's this. It's go through every single open opportunity and list two bullet points. She just like wants high level. This is a CRO. She's busy. She's got a lot of time, not a lot of time on her hands. She asks what is the single next step we we should take and what is the biggest risk to this deal? And it gives, as you can imagine, it's a lot of text, but she combs through this and she's able to know really quickly like what are very tangible things that I need to coach my reps into. And then she runs this into quad where she had a dashboard running for some of their active deals and action items for those reps. She could also build that same sort of dashboard in Cibble based on like urgent deal risks and what needs to be done. But then again, just thinking from a CRO's point of view, a lot of times you want to you have this problem where you're removed from deals, but you're trying to get in the weeds with them where you can this ability to query across all of your data at once and for simple to really understand not just very generic next steps, but a true next step. So like this one, it's pending this CEO's approval and gong is an active comparison.
Now, we recently launched dashboards, which is really cool. This is a very long prompt that one of our AES actually wrote, so that's why I left it long. You don't normally need to go this long of prompts, but he wanted to build a July 20 26 sales forecast dashboard and he ran this prompts and civil was able to create a really beautiful visual dashboard forecasting for this month. So, it's it's highlighting, hey, there's some urgent things. There's some targets that you're behind on. You have this open pipeline. There's obviously a gap. We're early in the month, so I'm not terribly concerned, but I'm able to filter through by by stage, deal categories, risk, and analysis. The deals that are slipping and kind of what's going on with them. All of that is readily accessible in that dashboard.
Now, I I'm a product marketer at Cable. This is one of the my favorites is I we run a competitive intelligence dashboard. I have this running two ways. One, I have a competitive intelligence prompt that runs to Slack every Monday morning and just says, "Hey, what are the top look at our deals from last week and what competitors came up? What was said about them? What do I need to know?" I have that prompt running to Slack so our whole team can see it. And then I'm able to view this dashboard. As you can see, gone comes up a lot. We have a really strong win rate against them, but they are by far our most frequently mentioned competitor. We can look at some of the objections that that come up in those deals with by by the individual competitors that we're looking at. So, gong, fathom, attention getting out of the individual IC level, thinking high level about how does civil understand your whole business is this ICP dashboard. So, I asked build an ICP dashboard just based on our closed one data from the last six six months. And Stibble gave me this really beautiful, simply broke down dashboard on who are the industries we're winning with, what are the value props that are that are winning those deals. In this case, I had to analyze 50 deals. It shows our main segment that we're winning in some the individual demographics of like where where are these people coming in from, the competitors that were displacing, psychoraphics on what what's actually the most acute pain points that they're experiencing. And I mean, this is just very visually satisfying. Again, all this data that Sibble is not just spinning together something that looks pretty. It's something that is thoughtful and it understands what's actually happening in our deals. Here's another really cool one on interactive sales pipeline that our CTO likes likes to run regularly. Just this really beautiful deeply understanding dashboard and then a customer health radar. We could keep going on and on. But the point that I'm a risk of overproving here is yes, you could have you could run a lot of these same prompts in Gong. I don't think you can create a dashboard prompt to dashboard in Gong, but you can run a lot of these same prompts try to get a lot of the same deal analysis directly in Gong, but the outputs are going to be noticeably different. They're not built on the same level of understanding and the same level of depth of what's happening in your deals. And that tends to show in the quality of of the actual outputs.
But before we kind of wrap, I want to just invite you to change the way that you are evaluating your AI. We asked this question at the very very top of of this conversation. And we said, what is your AI getting smarter over time? And I think that's a really notable way to evaluate if you're thinking about like revenue AI and what you're going to actually run your whole or on looking for just individual features of like AI notes, CRM updates, recording, forecasting. If you're living there, I think you'll miss out on some of the core value and the things actually make these products different. You go to our website, you go to Gans's website, we're going to say a lot of these same features exist, but the engine underneath them is very different. Does this product actually learn how you win? Does it remember what matters over time? Does it not just look at the aggregate of sales reps who speak 40% of the time or x% more like no like on your team when match does this it produces that that level of understanding? Does it get smarter? Does it make better decisions as more and more time passes? And will every workflow ultimately get smarter because of it? That's I think the future of how we should be evaluating this tech now.
We gave some proof in the outputs. I wanted to talk a little bit about our customers and what they are typically saying. We have a lot of customers who switch from gong and the overwhelming consensus is kind of what I shared earlier is that you can you can let run some of these same prompts. Gong is saying that they have this ask AI but the outputs are just very very different. This one from Natalie said civil is so much better than gong. My manager asked me, "Can't you just do this in gong, but cibbles is way more accurate?" Or Jeff at SmartCow saying, "For the catalog press, you'd expect Gong's AI to be mild ahead, but Cibil is easily the stronger one." Or I like Adam saying that Cibil is like Gong on steroids. There overwhelming feeling here being that there is that what's on the surface these things looked similar under the hood they ended up being very very different. We when we actually are in active analysis with deal with c gong customers meaning they actually enter the cycle where they side by side show us the quality of the they see the gong outputs and they see syibles outputs they connect gong into syibble and they can compare the two we have a 75% win rate the win rate is strong and I said that not as like some sort of like flex like oh we are we've cracked go to market part of the reason I worked here is the product is genuinely great and that's why that's a big reason why I've stayed at Sibla but the point here is that I teams that are running on civil. When your AI actually understands your deals, that makes your whole team more efficient. You're not you're not editing all your follow-up emails into oblivion. The CRM autofill is actually helpful. So, RevOps isn't having to go chase down more accuracy. Reps can actually lean on the prep rather than have to tweak it and challenge it and dig dig more deeply. And the difference just shows.
Now, I want to talk briefly before we get to Q&A about a little bit of just like what are the options here. So if you're on Gong, what are what or if you're on even another tool like what are options are sort of available to you? There's basically two main paths that we're seeing. One is you rip and replace Gong with Dibble. And so this is where a lot of times the costs tend to go down. The rep usage tends to go up. We see that rep usage is much more widely adopted in cyble versus Dong. So part of that's because Dong had been really good at building a product for VP of sales for a long time. But it all the AE work was sort of an afterthought. And so we noticed that that when people come over the actual adoption increases. They're not just pulling transcripts from gong into cloud and they move deals forward from there. Like civil capturing all the deals, but it's also then it's one product replacing all the multiple Gong agents you need to run. The workflow has become simpler. Ultimately the cost is lower and the intelligence is deeper.
Another option here and sometimes this is just depending on your org structures. We'll see some orgs will layer Syibble on top of Gong. Maybe you're curious about this newcomer Syibble and you're like, "Hey, we want to give this a run." We'll sometimes see customers who will layer it on top and keep the two products running side by side because AI is just worth it. That difference is notable and they will pay two subscriptions to make that work.
Now, wherever you're at, our invite to you would be to run a comparison on your own data. We can set this up in like a day. Will backfill three months of your gone data in Sibble totally free and let you see for yourself how cyble and gone compare. You can look at your own deals. You can look at the follow-ups. You can look at the way it analyzes the deals and see for yourself the quality. If you're interested in that, you can email myself, Colin.ai, and I will get back to our team and we'll get you moving in the right direction.
Now, at the risk of sounding a bit presidential, I like this this idea of like not asking what your AI can do today. That is the wrong question. But asking what it will know after the next 50 deals. And I think the the best AI is not just going to automate the work. It's going to learn how and why your company wins. With that, let's do Q&A. I'll cue some of these up to you. Goish. I've got a couple of them kind of already handy. I'm going to scroll back. I think David had the first one, and I think you'd be able to answer it. David asks, "How does table decide what is worth remembering? And can RevOps or sales leadership tune that logic?"
So David, that decision happens over a few different layers essentially. It's not all at once. It's decided across. So the way we do it is we have three essentially separate context layers and we call it differently. So one is your customer data and that happens on a per individual a per company and on account and a per opportunity/deal basis. Then we have a team layer which is on a per rep per manager per executive in that team. So that stores the drafting preferences it stores the strength and weaknesses the kota the goals objectives of that individual across the entire sales order. The third layer is the institutional memory layer which is what are the product lines, what are the different playbooks, what are the budgets, what essentially has been various competition that have come up and what are the market verticles. So things like those which are defining across the company how deals should be done and then every single information every single interaction every single activity and every single outcome is getting filtered into one or more of these different graphs and institutional memory stores that we have built. So if you start getting for example a big let's say you start seeing a new competitor happens very frequently in sales space. So if you start seeing a new competitor in the market and you start hearing the name competitor ACME again and again and again like after the 10th time system will start tracking that Acme competitor as a separate node and it starts keeping track of objections and queries and questions that the customer is asking and also over a period of time it will start understanding what are the answers to those objections that are getting you the right outcome or not the right outcome. So that level of filtering is happening across an organization on a daily in fact multiple times a day and that filtering that goes into these separate areas to your point of whether this can be tuned in via the admins or the team to some extent. We are now building the capability, you should see that in July to actually import any and all kinds of documents and drive data or notion data or any enablement zone. I see folks from later.ai here, but also like seismic or highspot and those can then enrich your actual institutional understanding of the deal or of the company or the customer. So like that can be in inferred from those data set and they can be used as a starting point or as a continual tuning of the system. What you cannot do is which is very practically impossible is go into the specific memory and say ame is not priced at $5 but it's priced at $6. That that piece is really really tricky to do.
Okay. So the next one next question we had is about dashboards go just are they all created with AI prompts are we able to edit them love to hear how and then there's this kind of supplementary question of how they feel like how they would love to hear how they're better or different from claw dashboards that they're using today. Um yes you can edit you can create dashboards with a single prompt. You can edit the dashboard in the same chat window. You can say hey I like to see my task list not as individual task on this dashboard but also as cards. So you will get that option. Syil is able to smartly infer if you say I need a coaching or deal strategy dashboard versus coaching dashboard. Sibil can understand what kind of dashboard would be ideal for your company for your situation. So it will be able to create that dashboard. In terms of why it is different or better than cloud dashboards, uh few different things. One is the data itself is refreshed in every period. So you get a refresh data. You don't have to ask the system to refresh the data. The second part is that the data actually comes from your underlying systems. It's not just very basic call and email which is or activity which is what you get from MCP. You can also get internal information like Slack messages or internal conversation feeding into those dashboards. So again go back to the identity resolution example for instance if you have prio in four different places and if you ask without identity resolution if you ask clot to create a dashboard it likely will keep all of those four contacts in four different rows and that is going to create confusion on what needs to be done. The third big difference is that civil will capture and combine both structured and unstructured data together. So you don't only get the objective number on this is where the ARR is at today or this is how many deals I have closed this quarter. You'll also get a red yellow green on what is the likelihood of hitting your target or your kota or what is the forecast probability of closing that specific deal. So it's a mix of both things and that really makes it very powerful because you're not just relying on the top level number. It's going in a level deeper and giving you the context.
Swear we got three more questions. One is Manan asks when we discussed about accumulated judgment and compare that to kind of Gong's generic judgment. Their observation is the inputs like from a volume standpoint would technically be the same but is it that stibble is doing the analysis at a deeper level. Also example like 10 plus parameters of those 60 calls 500 emails which connect back to similar parameters from other data sources or put how is the context different in both the data memory.
Yeah. So man what I understand is where is the how is the final processing happening and is that dependent on like getting the system to do a much more in-depth reasoning process versus there being something fundamentally and structurally different. The way we solve for the problem that you're describing over here is we actually build in between layers from your input raw data to your action layer to get to the final levels of context like you cannot get to context in a single understanding because the system would not otherwise know where to look at like an example I love to give is if there was some individual that you were working with let's say John you working with 6 months ago John was a champion And now 6 months out the deal is stalled. You write back to John. If the system doesn't know that John was a champion, it would not even know where to look for that information. Only at that point of interaction with John can you figure out that Jon is actually a champion or a detractor. And then your system also has to have the context of what should we say to a champion versus what should we say to a detractor. And both of those information combined will end up forming the email that you sent to John which gets you to reply.
Britney asks if if someone is locked in in a competitor contract, does your team offer any incentives to buy those out? Yeah, we we would once the customer has decided that they want to actually move forward, they have tested out both, we would very frequently buy out existing contracts so that they can avail the team can a lot sounds like partially case by case, partially when they actually get to that spot, but yeah, that's a that's a real option.
Britney David again asks, can I inspect and edit the metric logic behind dashboards? So for example, what counts as pipeline created, how the same same lead close loss is excluded, how source segment are normalized. Kind of similar I think to the question before of like having that control.
Yeah. So with dashboards you can do some bits of some bits and pieces of this because you can define to in the prompt very clearly on how how exactly do you want the deals to be counted. What you may not be able to do is very complex like statistical or probabilistic analysis which most people typically do not need to in their deals. But for basic stuff, we have a code execution sandbox behind each dashboard that gets created. So essentially it's not a prompt that's running. The prompt is creating a code to run the dashboard. So the code is what gets edited and you can make that edit.
All right. And last one they have on here is do you use the same model for driving conversational intelligence forecasting etc as you have embedded in the memory semantic query and context graph layer.
No the model changes significantly across different layers both on the product feature standpoint but also on the building of the entire institutional memory. In some of the cases we have our own in-houseu SLMs and models. In other scenarios, we will use open-source LLMs like Llama. And in even more different scenarios, we will use paid LLMs like Charg Claude. And it depends very much on the use case and the need of like how much processing abilities are needed at each step of the memory layer. Because if you use the same model, you end up blowing your cost budget by in our case that would be probably by 100 or maybe thousand x given how much processing is happening underneath.
All right. On that note, there are no further questions. Right on time. So I want to say thank you to everyone who came. This is a hefty webinar at locked in an hour. So if you stayed the full hour, you're committed. So thank you. Thank you for staying.
Gish and I are both on LinkedIn. over to add both of us or my email was colin@simple.ai if you have some specific questions. To answer the question everyone asks, yes, it will be recorded. So, we'll get you an email out either this afternoon or early tomorrow. Thanks y'all. Have a good day.