Building AI Sales Agents for SaaS

Okay, we are live. We just wait for a second for a viewer or two to join in so we know that we are live.

We are. We are. Awesome. Perfect. So let's begin. Hello and welcome to Momentum Officers. My name is Yash, and I'm joined by my co-founders, Jay and Koushik, to discuss topic of the week, building AI sales agent for SaaS firms. How can SaaS firms build sales agents using AI? Our goal with these sessions is to provide you with actionable insights and practical strategies that you can apply to your own business.

Throughout the session, we encourage you to engage with us by asking questions and sharing your thoughts. This is a fantastic opportunity to learn from each other and gain new insights that can help drive your digital initiatives forward. So let's get started. Koushik, how are we doing today? Today is a Saturday, so I'm sure no meetings. Were there any meetings? I didn't have, I mean, I was supposed to have one, but like it got postponed to Monday. So I was preparing for the Monday.

Yeah. Awesome. So does that mean that you weren't prepared for the meeting today? Had it happened today, if you're preparing for Monday, it means that we had the meeting today though, no? Yeah. So it's basically I prepared for the entire week for the coming weeks together. No. Got it. Got it. Yeah. Awesome. So we're going to talk about building AI sales agents for SaaS. And before we dive into that, Jay, you're going to take us through

how and why and what to do with building AI sales agents for SaaS and so on. Before we do that, so tell us what are agents we've heard, what are sales agents and then maybe you can share your screen and we can go through the whole thing. Sure, yeah. So basically, I'm going to cover that as you can see 33 % of my slide on what are sales AI agents but just to answer that, that basically you can consider

Yeah, enabled software which basically assist in getting certain processes of sales done. How they are different from basic automation or regular tools. to have a little deep dive, understand some use cases and see through as well. So, Yeah. Awesome. So let me bring it up to the screen. Can you share that? Yeah. And maybe we could.

go through that because I just realized that I asked a question that spoiled your flow of the deck. So I'll just pause myself and maybe you can take us through. Yeah. So can you see my screen now? Yes, yes. And it's up on and live on the stream as well. So we're ready to go. Awesome. So as you already mentioned, this time's topic is building AI sales agents for SaaS, transforming

basically your sales with AI. So now we have AI in every process of our businesses or that's what we are trying to help our clients with. Obviously, sales is something where one should have AI utility and that's why we have been doing series of sessions where we also talk about how AI can be helpful. So last time we also talked about what AI tools you can utilize in each and every stage of your sales processes.

Now, this time is more about building AI agents on top of it. This time it's going to be more of an introductory part where we will discuss first upon three main things. One would be why should we use AI agents utilizing sales? And the next would be what they actually are and how they are different because there are a lot of confusion between people on what is the difference between basic automation and what is.

What will sales agent do in that? And then how you can utilize in your organization as well. So we'll have pointers towards that as well. So to get started, as we can see, so let me just come to the place where why do we need it? Since you already use AI for other things, it's not like we just need to do it for sales as well. But the primary reason is because of the increasing competition throughout.

especially from SaaS standpoint, CAC for getting new customers is increased. for example, you can say there's a rising customer acquisition cost. Certain reports showcase that it's already risen to around 70%. And that's why overall profitability hampers up. And growth strategies also need to improve for sure.

And that we can also see in lots of newspapers and things, all the news and all the places as well. A lot of companies are not able to survive well because of their CAC ultimately. The SaaS companies, which were very well funded, even are also now shutting down just because they have not been able to optimize in their profitability as well. So this is one of the reasons. The second thing would be the sales cycles, which were earlier smaller now because of the competition.

the potential prospect has a lot of options to choose from, and that's why it takes very longer timing for them to get into it. if there are any AI agents or if there is any process which could reduce that time, then that would be helpful. So this is one of the reasons. the third would be, sales processes when completely handled by human or large amount of it when handled by human could have some gaps. So for example, there could be too much of time in

You know, qualifying the leads, then there could be 78 % reports also showcase that 78 % of buyers go with the fastest responder. Personalization at scale is also tough because let's say you are supposed to reach out to 50 prospects or one salesperson is supposed to reach out to 50 prospects in a day. I mean, doing personalization for 50 of them becomes very tough, right? Because they need to, so let's say if there is a LinkedIn autumn,

there's a LinkedIn message that you want to do. You want to just reach out as a cold message, but in that also you would want some personalization. If I am supposed to do that, I won't be able to, every time, check on their feed, just see what last posts were about, and then talk about the same, because I would have other work as well. In the same way, I mean, it's not possible, right? So that's the larger chunk of it. So what if all of these things can be done through AI? So we have already talked about how AI tools can do it.

can also talk about how AI agents can automate these processes by making tools talk to each other and get the process flow done step by step. So the other benefits I would like to also talk about is with utilizing AI agents, 20 to 40 % of tasks can be automated. It will definitely have very faster response. And faster response would lead to more conversions.

and then you don't need additional head count to just scale. If you just want to capture a new market, then you can just replicate the same process through agents and you don't need a dedicated team just to handle that particular market to a larger extent. Am I audible, Yash? Yeah, loud and clear. And the points that you are making are also echoing.

So yeah, is 20 to 40 % of tasks being automated is and I think out of the four, while the other things are financial gains, the fourth one, is cleaner, richer CRM data is something that is of most appeal to me. I might not care about more money or more efficiency or anything like that, I have over the last 15 years of my work life, I've never seen a clean

and rich CRM, right? It's either clean and not rich, or it is rich and not clean, or it is neither. Having it both clean and rich is amazing, especially from my perspective. Yeah, definitely. So a lot of benefits for sure. mean, there's a lot going on. Kashi, you want to say something? No, was going to bring up this point that, are we, from the past three slides, what I'm seeing as a pattern that you're trying to make a point of?

Are we saying that most of the knowledge work that happens for a salesperson in his day-to-day tasks, those are something that will be the first set of things that or first set of tasks that will get automated or augmented by an AI agent? Is that what we are arriving at? So not the initial processes for a salesperson, right? The overall work of entire sales cycle, it varies from business to business. But to answer your question, it's not just about

processes before the main salesperson gets started. that what you're trying to ask? You mean that, know, pre-sales part of sale gets covered up and then salesperson will still be needed to get things done. Is that what your question about? No, I'm specifically asking about the knowledge work that a particular job does, right? Like for example, you mentioned that to create a personalized response for a particular, for 50 profiles, one has to go to every individual person's

Feed, what sort of a, you know, create a person for each one of them. Now, we are saying that the choice and knowledge work that you are actually seeing, understanding and coming back and creating a personalized, you know, message for that particular person. So are we saying that for those kind of tasks, which is a knowledge work task, the AI agents are going to be in some form of help for the salesperson? Is that what we are arriving at is my question.

Yes, 100%. 100 % for gathering the knowledge. That makes sense. So two things, right? Whatever you ask, answer to that is yes. But it is not limited to that, is what I would like to say. There are other things where AI agents can be helpful. So what we are talking right now is more about the enrichment part, lead enrichment, where we are talking about getting more information so that it could be utilized. But the other piece would be.

generating, having a communication, and also predicting. So one would be, first of all, finding out the right relevant data. The second would be about generating the content relevant to it. And the third piece would be also predicting. I mean, second would be predicting. And then third would be generating the data or generating the content to it as well. I hope that makes sense. That makes it clear. So I think the term.

Knowledge, if we just call it intelligence, would be more, it would make things a little simpler for us and for people to understand. So all the knowledge work may not be, like the knowledge work can be helpful, but think of it this way that as a salesperson, people will continue to buy from other people. So AI cannot sell or close the whole transaction, but for the salesperson, everything that they need to know, they will

know that and more to be able to facilitate the sale or accelerate the sale or augment the sale, whatever the case may be. So I think we can say that this will make things, it will make it more intelligent. Yeah, 100%. And hence it increases the efficiency of sale as well, right? Yeah, absolutely. And so one of the things that also happens is so that I may not, so here's a good example of this, right?

So as an example, if today you create a mobile app, your mobile app has to be as good as WhatsApp and Instagram and Telegram and whatever, right? Because your consumers don't know, I mean, they don't care about your lack of resources. They are using like, you know, all the number one rated apps in their own category 95 % of the time. And so their expectation from your mobile app is extremely high. Very similarly.

Purchase teams in organizations who are buying SaaS products are being pitched by other salespeople who are using AI agents. So that means that the purchase teams' expectations of how they are treated during the sale, what quality of documents, what quality of presentations, what quality of comparisons, what quality of quotes and proposals that they receive and what quality of emails and responses do they get from the sales team over the next three, four, five years is going to be so high.

that if you are using only humans and not humans plus agents, then you're much like how you will think that a mobile app is clunky because your expectation is so high using other apps. Exactly that way, the purchase team of an enterprise or a mid-market company that's looking to buy a SaaS platform for solving a particular use case will have their expectations set so high because your competitors are using

all of those AI agents. So their treatment has to be significantly better. So think that's another piece that I wanted to add. sorry, what do you say? Yeah, yeah, 100%. So the next part, just to add to that, impact, what impact does it create, right? If you utilize AI in your sales and especially have AI agents in place. So overall, 63 % faster replies could be possible because you're utilizing it in your chatbots.

in your websites, also in your products as well, wherever any questions are. There could be 22 % more conversions. All these numbers are basically from certain set of reports that we have gathered upon related to sales and just putting it here. So 28 % more demo bookings, 40 % after hours also gets improved. 52 % shorter cycles for sales of buyers as well. Now, one may ask on why and how sales of buyers

journeys can be reduced through ages. But consider it this way that if there is any, let's say, I'm using one of these SaaS products and I just started with my trial period. Now in this case, I know that I just want to check some feasibility and if it's helpful for me, maybe I want to go ahead and buy it through self-serve. I don't need any salesperson in this area.

if the product is having lots of features and if my use case is very specific. Now, for instance, if I'm using it, if there are certain agents built in a way which can, first of all, the use case or you can say the usage timing from my side on how I'm utilizing the platform. And if it was already, if someone like me was already done this 100 times and then went ahead with the purchase through certain features, then

certain things can prompt to a certain part of our platform and maybe from there a new pop-up can come through chat or any other ways and then this sales of journeys can be improvised as well. So it's ultimately all about utilizing AI wherever possible in predictability and based on that just creating something very personalized and just doing it.

like humans may not be able to do it that efficiently or up to that extent is what each and every slide ultimately tries to tell. that's why we should have more of AI in our platform. So yeah, to answer what you asked, Yash, in much detail. So basically, sales agents are basically we're talking about AI-powered software that autonomously performs sales functions. And obviously, it's not just going to be one, just one.

tool that we might be using. I mean, I would say deep integration between multiple tools and getting the outcome. So what it does basically is it utilizes NLP workflow automation and sales logic. It makes decision based on sales context and customer interactions as well. Unlike the basic chatbots, it can actually execute the action, then not just respond to the queries. So earlier chatbots were more about

some data is already feeded. So based on the question that comes up, it just answers over here. It's not just about answering, but also executing. So consider utilizing it as a ticketing system. That 13 queries are being asked. Response to that is already covered up. then if a new feature is requested in that case, sending out feature request form, but then also getting its entry added inside or just from the chat, entry being added inside our overall board.

So stuff like move my plan from silver to gold and add three licenses. You needed a person to sort of do that. And now an agent can just do it by like within the platform itself. Yeah, yeah, definitely. So Jay, does that also mean that the agent needs to have access to my knowledge base of apart?

Apart from the tools also, is that the case over here? Yeah, definitely. Definitely needs to have an access to knowledge base. once we move ahead, Shashik, we'll also define on what sort of, we'll also, I mean, not necessarily today, but later on when we have other series of plans, we'll talk about what priorities we should have. And then based on that, what needs to be kept up on. So I'll answer that in the next slides, for sure.

Yeah, and not just like knowledge, it also needs to know the user roles and rights within the system, right? So as an example, all of our company uses ClickUp. If ClickUp deploys an AI sales agent, my intern shouldn't be able to change our plan, right? So my intern shouldn't reach out to the ClickUp AI agent and say, hey, upgrade us from whatever grow plan to enterprise plan and add 50 licenses or whatever. It also needs to know that the person is talking to them.

actually has the authority to do it. Yeah, indeed. yeah, definitely. mean, for sure. So just to add to that, if we talk about from sales standpoint, what AI agents can do. they can be, it could be, you know, inbound qualifiers as well. I just said they can be instead of it can be. You see how it replaces humans with. Yeah, interesting.

Yeah, so inbound qualifiers, right? So understand, like it would screen and qualify the website visitors and leads as well. So this we have discussed it time and again, so I'm not utilizing much time there, but it just qualifies the lead in whatever way, or form possible. Outbound prospectors are basically identifies and engages the potential customer. So it just the enrichment part we discussed upon in outbound standpoint earlier, it just.

This one of the those use cases. The third could be product explainers. Fairly simple. It would answer and deliver personalized demos based on whatever information is feeded, but not just the information that is feeded. It also has a capability of, you know, utilizing more about your profile. So if let's say Cauchy is, you know, signing up on any of the platforms, it just doesn't take the three fields. It takes typically three fields of name, email address and some additional information.

then the next onboarding from onboarding part could look like what you are here for and things like that. But not only these, it also on the back end checks about what Cauchy what that particular email ID what information with that particular email ID who is it? So it's Cauchy get momentum. I mean, if if your email ID was Cauchy get momentum 91.com, then in that case, momentum 90 what momentum 91.com does based on that also it calls you certain information.

And then it will be more personalized product explainer video that is generated. So that's how we are talking about utility of AI in every aspect. The next is deal advances. So it basically follows up on proposals and handles the objections. So a human is not needed on that particular part, but those things can be created. Earlier follow-ups were also possible through automation, but now we are more talking about handling the objections, generating content, understanding what the

prospect is asking on and then just answering it accordingly.

And so deal advances is largely that happens on emails, it? Or even calls or? So I'm basically what I'm trying to know is when I'm doing a video call with a prospect, are we expecting an agent to join on that video call or is this following up on on LinkedIn, email, WhatsApp or? Yeah. So here it's more about.

So when we say deal advances here, is this an example given on proposals? So proposals follow up on the email message. It could also be on a call as well if that's set up we want to do. So now that is possible. As you also mentioned last time on AI work, so that is also possible. Whereas, as you mentioned about joining on call, so as we discussed earlier, there are AI tools which basically

help sales reps on improvising on their conversations right in the meeting. So it also gives you certain insights and tells you what next to talk about and things like that. So, you know, advances can be that as well, for sure. Talking about the capabilities as we talk, so like, it will have a natural conversation with contextual understanding. So that's the benefit. Integration with CRM and sales tools ecosystem.

It's obviously possible and that is needed. Otherwise, this won't be possible. The third would be execution of multi-step workflows with conditional logic. So this thing I would like to take up in much detail in next sessions where we are trying to build a workflow and setting it up in building an AI agent for a sales process. Then seamless handoff to human representatives when needed. So handoff also can be possible as and when needed for a certain set of processes.

Overall, things get covered up from capability standpoint. To quickly discuss upon certain use cases, here are some. So let's say inbound lead qualification. So it does the initial engagement. It needs assessment and qualification criteria verification. So just to give an example, Drift basically reduces its sales cycle by 33 % using just PIA qualification. Same way from.

Outbound prospecting standpoint, we have already discussed time and again on how hyper-personalize a message could be. Overall outreach would be how follow-up sequences can follow up and how meeting can be scheduled also very effectively just through AI. So outreach.io achieved 35 % higher responses rates with AI personalized cadences. So that in itself is an example of it. The third would be.

you know checking the proposals it is again which we discussed on the previous slide as well so DocuSign Increase its deal velocity 28 % with AI Assisted Follow-up. So what basically it does is if I am prospect who is supposed to sign some third party or some vendor has provided me a proposal with that if I have some questions I can just ask write in that document and those will get answered up and I don't need a third person obviously this

This does not mean that salesperson should not be notified about the same. They will get triggers that, OK, this person has asked a question, and if there is any additional help, you may want to do. So it basically helps the salesperson in a way, but it gets the job done as well. So Jay, just a question over here. So could you tell at which part in the sales process are AI agents enabling me to do something?

Similarly, at which part are they replacing a particular person?

This is very vast question Koushik because as we talked about in entire sales process if we see, only one instance, you can take any one instance for example you could just talk about outbound. Yeah like an example. Yeah so let's talk in outbound standpoint right so if things are very well set later on what happens is from finding the data to you know writing a messaging for that and then to doing the outreach then to

and handle the objections on getting the idea from outbound is to get a meeting book for any business, for instance. Then in that case, finding the data of right set of target customers, optimizing, mean, just writing down the message or sequence for that through email, LinkedIn, or whatever communication platform that particular target group is presented. And from there to getting meetings book, all of these processes can be done through AI, but what we need is the right workflow to be set.

and right context to be given so that this thing is well set. Ultimately, in the meeting, I would have to go because if you asked on where a human interaction would be needed, so I would say largely in the entire sales process, the main conversation that happens would be needed. But from sales. And even the first one, right? And even the first one, which is defining who our ideal clients are and what are their pain points and what is our value from, right? So the first piece, like having an extremely strong

very pinpoint definition of ideal customer persona. that is very important. Thanks for adding that, Yash. I missed out on that. that is very well needed because consider you have set up very good processes and built a very good AI sales agent. I had lot of great integrations from amazing tools inside it. if you're not able to position, if you're not positioned well, if you're not

targeted your groups accordingly, then it ultimately won't give the results we need. So all of these definitely. So it's more about reducing time, increase. It's all about, I would say, reducing the time and being more and more hyper-personalized and active or more efficient. But then ultimately, the strategy part is where human interaction would be very well needed. And the second part would be the main, if it's a larger ticket size, if it's not a

low tickets as I would say in larger ticket size also or for products where you know it's not just about some customization but then there is no predictability on what sort of questions or queries could arrive from prospect I would say in those cases human interaction would still be needed for closure. yeah.

Does it answer your question, Koushik? Yeah. So talking about how, right? So I, to be honest, we'll go very deep into this later on as well. But just to give some insights on how to do this. So we have some step-by-step processes on what you need to do. So first would be identifying the use cases, of course, determining the specific areas where AI would be helping the sales process. So for that, one needs to have a very

well-defined sales process and then define on where AI should be used in the same. Mapping these whole processes on how overall a prospect's journey is and then from the same web, what sort of workflows would be in place so that we can integrate AI solutions accordingly over there. Implementing the tech stack. So now since we understood what use cases are there, finding and deploying the right technology to support that is very much important. And then

The last part would be testing and optimizing it. obviously, one has to continuously evaluate and perform. So when it comes to implementing AI, it's always about testing and optimizing. And that's how AI in itself has functioned so far. So even it's not like you have established AI sales agent for your business. will work everything. It will replace everything which was there earlier. It's not like that. Eventually, it gets better. And then you can eventually utilize it.

by replacing the old processes. So just to get into it, just to give you some idea on core components which are needed, would be LLMs, orchestration layer, vector database, workflow automation, which is very much needed for building AI agents. N8n is a very good example for the same. We are going to showcase an example, as I mentioned earlier, in the next one of the next sessions as well.

And you need definitely integration APIs. The right places where you're currently everything like out whatever your outreach platform is, what your calendar is, what your CRM is, you need right integration APIs for the same as well. yeah, these are the core components, which we had talked much in detail about in last one of the sessions taken by Yash as well. So, yeah, we can refer to that if there is any further understanding needed in those areas.

Talking about integration points, obviously we need customer data, we need CRM systems in place for communication, email chat and messaging platforms to be in place, obviously knowledge based scheduling. These are very basic, so I'm just moving real fast, just let me know if there's any question with it. Automation logic is obviously needed, so prompt engineering is needed for specific instructions, right? So we need like

Retrieval augmented generation for product knowledge. As you mentioned, that product documentation would be there. But if there is a large amount of data along with the rules and each and every aspect, so we need that particular data. And then we need some RIG setup on that. Workflow triggers and conditional logic is also something which is needed. So these are just basic step-by-step things on what would be needed. From testing and deployment standpoint, from testing

framework, conversion simulation testing would be there. A-B testing would definitely be needed. Just like having basic sequence also, we have it here for the whole processes. We would like to have that and then see which workflow which we have planned upon works well. It feels like each bullet point is one, like requires at least one live session. Yeah, it's one session. Yeah, it's not, yeah, of course. So thank you for that.

Thank you for saying that this will be a series, right? So this is sort of a table of contents for a book like a song of ice and fire. Each of them, it's a lot of things, but it is still. Yeah, but this is more like highlighting on one should be aware of before actually getting started on like, know, so we can consider as a very, as you mentioned, table of contents on what would be needed, but.

Just to get started, are certain things one should be aware of, and then maybe we can take things forward. So yeah, obviously deployment strategy is also there. So just want to quickly discuss that as well, that we need to first have a limited scope. And then human in the loop should be there initially so that some supervision is there and see how things are performing. Because imagine you just say that, I'm

just utilizing AI sales agent for reaching out to 10,000 of my prospects. What if things were not said well and what wrong sort of messaging goes on from your brand? So it's a problem. So you have to start small, have supervision from human for sure, and then gradually you need to expand that and go ahead with your automation. this is just a view and this could be subjective from.

organization to organization. So initial first two months would be more of utilizing and mapping the processes. Third month would be more of implementing the tech leg and fourth would be pilot rollout and then you start gathering feedback and then start optimizing on that area. Yeah, so just to give a quick takeaways from this particular session. It's more about maximizing ROI. So one thing is that

We are utilizing it, it's very cost heavy for sure. But obviously the whole idea and the process is about maximizing the ROI. second would be... It's not cost heavy. I'll tell you what. I'll tell you what. It's just how do you think about it? So like as a business owner, if you are running a SaaS company that's selling to mid-market enterprises, imagine a scenario where you hire a salesperson who's either inefficient or you hire a salesperson who's not effective.

And then if you think about it, it's not cost intensive at all. Having a small sales team, a particular geography for a particular vertical, that is inefficient. If you just put the numbers, mean, it's so, that piece is so cost intensive that just thinking about it makes me fearful. My spidey sense starts to tingle.

But as you mentioned, if you actually scope your tasks precisely and match it with the right cost optimization that is needed, it is not that cost extensive. We actually have done those exercises while we tried to implement. So that is the first thing that we arrive upon. Yeah. And whether

like whether one chooses to go ahead or not with AI agents, their competitors are going to and they are going to be way faster in coming weeks and months of time. there is no... Yeah, this is exactly like building a good quality mobile app, right? So 15 years ago, all the media companies did not publish mobile apps. But people started consuming... Since media companies were not on...

they didn't have mobile apps. People started consuming media on social media. So they would follow Wall Street Journal or would follow New York Times on social media and seeing that they had to build mobile apps and very early first versions were clunky and then they had to become significantly better over a period of time. So it's the exact same thing. If you are not doing it, your competitor, the industry is going over there and it has to be, it's just a reality that we have to all...

essentially just an app. Yeah, indeed. So yes, I mean, as we can see, the key takeaways would just be on obviously, this is more for maximizing your ROI. And you need to start with small projects for sure, for the reason mentioned in the previous slide and leverage mix of expertise. Why this is needed is just one salesperson sitting and just building AI sales agent for them would not work. There needs to be some, you know,

information from different domains, I would say, which are relevant for the product. And then those information should also be kept along with for sure to make sure that output is efficient. So that's it. And as I mentioned, we can have next session for where we actually take you and walk you through how to build your own AI sales agent for your business. So we can do that in the next session as well.

And so surprise, surprise, we do those implementations as well. if you want to check it out, then you can consider reaching out to us. Happy to get on a discovery call, which is where we can help you identify what are the places where improvements can happen and what are the places where these things can be implemented. But until such time, just thank you, Jay, for putting this together.

for writing all the bullet points that we are going to cover in the sessions in the future. This, think, is a great starter for us to sort of have a direction in terms of what are the pieces of content that we will be covering over the next few weeks. for today, we are already at the 37-minute mark. And we didn't realize that we are going over time. So this was the perfect mix of sort of understanding and

setting the tone for future conversations as well. So thank you for putting that together. Thank you everyone for joining in, for people who are watching or listening on YouTube or LinkedIn. I don't have a joke or I don't have a dad joke for you at this time. So in absence of that joke, you just as a gift, right? So me not saying a dad joke is also a gift and as a return gift, if you can subscribe, that will be great.

It gives us great encouragement to keep on putting our thoughts and ideas and knowledge together and sharing it with all of you guys. Thank you for joining in and we'll see you again next time. Until then, bye. Bye.