How To Choose The Right AI Vendor For Sales & Marketing With Ananthakrishnan Gopal

In today’s rapidly evolving business landscape, the integration of artificial intelligence (AI) into sales and marketing strategies has become imperative for staying competitive. AI offers unparalleled opportunities to enhance customer engagement, personalize marketing campaigns, and optimize sales processes. However, with a multitude of AI vendors flooding the market, choosing the right one for your sales and marketing needs can be daunting.

To guide us through this we have, Ananthakrishnan Gopal, currently serving as the Co- founder and CTO at DaveAI. He is a distinguished expert in artificial intelligence (AI), machine learning (ML), and natural language processing (NLP), specializing in human-machine speech communication. With an MSc. (Engg.) from the Indian Institute of Science and a Bachelor’s degree in Engineering from Vishwakarma Institute of Technology, Ananthakrishnan has cultivated a wealth of experience across various esteemed organizations such as Vertis Microsystems, Samsung R&D, Artivatic Data Labs, and Niki.ai. Passionate about leveraging technology to address societal challenges, particularly within language data analysis for information retrieval, education, and healthcare, Ananthakrishnan’s approach is marked by a knack for identifying and solving complex problems with available resources. His adaptability and versatility, honed through diverse experiences including active involvement in a music group during his university days, enable him to excel in dynamic environments and deliver impactful solutions tailored to the needs of the organization and society at large.

Ananth covers the competitive landscape of Gen AI in sales and marketing, exploring why entrepreneurs are drawn to this sector. He discusses the challenges a Chief Digital Officer faces in selecting the right tools amidst numerous options and highlights the importance of impactful solutions. Additionally, he addresses the criteria for choosing a vendor specializing in sales and marketing integrations and outlines key characteristics to look for. Ananth also touches on declining client requests based on fundamental capabilities and the qualities Chief Data Officers should seek in vendors. Lastly, he anticipates emerging trends in identifying the right vendor in the evolving landscape of Generative AI.

AIM: Why do you think there are so many competitors in the sales and marketing Gen AI space? What is it about the use cases that is enticing entrepreneurs to take up that role?

Ananth Krishnan: The concept of Gen AI has been around for maybe around 20 years or so. When I was doing my PhD in Sweden, this was an upcoming branch which was very much in the research area. A lot of us felt that there’s a lot of potential for it but it was very limited. Most of the answers that could be generated would be very limited and there would be a lot of hallucinations as the term has come in now. But in those days we used to call it randomness or high temperature for these words. One of the biggest things that has happened is the revolution in terms of the GPU’s and because of that we have been able to process very large amounts of data and train these models.

But having said all of that, it is still a tool and what Gen AI actually does is to say that “Here is a set of inputs and if you’re trying to predict what could be at the set of outputs according to this set of inputs based on all the data that it has received.” Although recent updates, like considering context and paying closer attention to details, have had their impact, the framework’s essence remains focused on its core elements.

When you talk about a tool you will use when you feel that this tool is useful, but also cost-effective. If there is a tool which is not yet cost-effective, most of the cost centers will not use that tool until it’s established that it actually reduces the cost. Whereas sales and marketing is actually a profit center, which means that whatever tools they use may not be only seen from the point of view of its optimization or cost but to see how much more sales they can do, how much better marketing, how many more people they can reach out.

In the B2B space, especially for businesses they are the ones who adopt new tools quickly. They were the first ones to adopt automation and internet. Until 1999, internet was solely used for making websites about your company, which is sales and marketing. So that’s pretty much there. It’s just the thing about the adaptability of this particular tool and why and how business and still early we don’t yet know its limits, we don’t get all the factors. It is still not yet cost effective. So that’s the reason why sales and marketing are the first people to adopt it and that’s the reason why a lot of startups in this space.

AIM: Given Gen AI’s significance in quantifying ROI, particularly in sales and marketing, where technology plays a crucial role, how should a Chief Digital Officer (CDO) of a Fortune 500 company begin looking for the right tools amidst the plethora of options available, considering that a simple Google search might not yield the most suitable results?

Ananth Krishnan: As we know, it’s a comparatively new tool which we know is very effective and also there’s a lot of potential. This is something that is completely established and that’s why CEOs have in their mind that it is very effective. Now there should be two types of strategies. What is what we call as a short-term strategy and a long-term strategy. So because of the fact about Gen AI, this is a tool which can only be effectively used by extremely large companies because of the cost of computation that is required to wield these tools and secondly the cost of creating anything new from this and it’s extremely expensive to create something which is comparable to what ChatGPT does today. It’s at least a few billion dollars to be invested into building something similar to that.

In the short term, there is no way that people can look at it as doing it in-house. This is almost a dream because you will not exactly be able to do this in the short term and because of a monopoly of a certain set of vendors with four to five people who really are at the top of the game here in this case where it can be used at a business scale. What most of these businesses have to do is align with one or two types of solutions and figure out people who can customize this solution for their problem, the way that they are, doing it. So that should be a short term but also should work on a long-term strategy to create and own some of these things.

To start with they should also explore other Open Source models and tools that are available here and they should also invest in having a workforce internally who understand these tools, what the limitations and advantage of these tools are and then engage with internal or external researchers who can help them to use these tools specifically. So there should be a short term and a long term strategy here.

AIM: How do we choose the right vendor based on our desired use cases? As tech professionals, we’re focused on impactful solutions. Can you share some of the key use cases Dave AI addresses and those already making an impact in the market? Many providers mention only 10% of solutions reach production, with the rest in pilots. How do we identify high-impact use cases if unsure where to start?

Ananth Krishnan: We are working with Gen AI in three or four directions, one of them is to improve communication between the user and the business entity. We had already been working with avatar-based solution as well as well as chatbot based solutions which was pre Gen AI. this was more of the older NLP based solutions, that was where we were and some of our customers already were using our solutions there. They were free text in the sense that you could type in whatever you want. It would identify the intent all that was there but it wasn’t just click but at the same time it had its own limitations and so on. We were working on our own Gen AI models at this point when Open AI just blew the market open. So this is the context.

Going from there we had more tools at hand, open source libraries that came up and we adapted some of these libraries with our own research and we have some of our own systems there. But I would very easily say that in any such system that we are building they are not at the same level of accuracy as what you would get from ChatGPT including the larger open source tools like Llama 2, they are no way close to what ChatGPT is today. The other aspect that the use of ChatGPT is still not yet clear for many so it can do something great but a lot of people use it for ad hoc stuff. They don’t necessarily use it in a specific use case. We use ChatGPT like a extended Google Search, So previously what we did is we did a Google search went through many articles, tried to understand the problem and then maybe find a solution. Now what people do first they go to ChatGPT and then get a summary of what they are looking for. This is what people are doing more or less and it’s great at doing this because it’s able to take vast amount of sources condensed it into something of a few paragraphs and then give you a very succinct answer to that.

Now this is something that people feel has a great potential but when it comes to industries, they don’t want the variability that comes with this. ChatGPT is a fantastic tool and there’s a variability in the answer that ChatGPT would give or any such generator AI tools. And this is not something that enterprises actually like a lot. They want very consistent answers which is aligned with their legal aspects and also with the way they want to promote. They’re not used to this kind of variability. So what we have done is we have done various pilots with the enterprises that we have worked with. These are largely Indian companies. So maybe there is a bias here for me because I have not worked a lot with the american companies. Mostly our customers are in Europe and Middle East and India, but around 20 to 30% in the other geographies. For all of them together, we see that that consistency is something that they desire. Even if there is a small change in this they don’t want a very large change because sometimes it could be the answers that GPT produces for exactly the same question could be quite different. It could be varied types of answers. Second, they’re very keen on the accuracy of the answer, which is actually a problem again when using Gen AI. They’re not very accepting of inaccuracies which I think is for good reason because when it is for personal use you might be okay with an inaccurate answer. For example if you are looking, for GPT to create code for you and if the code doesn’t work, you can try a few more times and then get the code working. That’s typically the way that as a person I would interact with GPT because it’s not accurate I would find out a way to get more accurate answer. That’s not okay for a business use.

So what we have been doing of late in the last eight to six months was to find a way to make these answers more accurate and find a way to make it more succinct and precise, without a lot of variation. This is what we have been actually focused on. Out of that we have around 30% these pilots that have gone into production for us. It’s a slightly higher rate because of the fact that these are our customers which already were using some of these tools, and we just made an advancement in these tools with the help of GPT. But then again the new pilots that we did around 10% of them are pilots. There are very small scale, they’re not at a large scale like how we used to have those with Chatbots and avatars together. We used to be actually connect with more than a million customers a year in the past. This was a pure NLP based system.

That’s not the case here because if we did 200 million interactions with people they would be spending millions of dollars on this so they don’t want it to be at scale because of the cost. That’s one of the biggest reasons why pilots are not getting into production.

So the ones are going to production were the ones that were laid on top of our NLP system, which means that 80 to 90% of the queries were solved by existing technology, that is pre Gen AI technology and then only around five to ten percent were actually using the Gen AI in reality.

AIM: How can a company determine if a data science vendor specializes in sales and marketing integrations or modifications? What key characteristics should they look for in such vendors to ensure they have the capabilities needed?

Ananth Krishnan: Almost all vendors are built on top of some API’s. Whether we use for chatbot based system or use a mix of text and images. We have also done some use cases where there’s a text and image. We also have use cases which are avatars based on Gen AI. So these are the three different cases that we have already done.

How would a company understand on how to engage with a vendor is that they should be able to get a long-term strategy from the vendor. You cannot have only a short-term strategy. It will not help you, it might give you small wins, but it will not really help you in the long term, so you need to look for vendors who have a long term perspective. That’s one of the things that I would try to understand.

So this long-term strategy requires the vendors to have a bit of the capabilities which are not just based on the existing API’s on OpenAI or Azure or AWS or Bard etc. People who are just built on top of an API as the core are not the right kind of vendors.They should look for vendors who have at least some bit of the technology that was pre existing in some way or develop on their own, at least a part of it and I would say that most of the vendors who are working in this may have this technology already. They may have this technology, but if they don’t have this technology, which many vendors are also having, definitely those are not long-term. These are people who would basically be able to implement a certain API for a limited use case.

The second thing is about having a layer which is at an enterprise level. Every enterprise has a very complex ecosystem of software as of today. There is the CRM, the ERP and this is just to name a few but then there is the document management and there would be a bunch of things which are often not from a single entity. It’s not like Microsoft provides all the solutions. There’s no single system which actually provides all solutions for the entire Enterprise.

So it’s important that any of these point solutions that they want to integrate using Gen AI, should be integrable with all the other things.

There should be a layer which is independent or agnostic of a Microsoft Azure or AWS or Google and that’s of cloud but it’s also of which API they are going to use. Are they going to use an open source Llama type of tool or are they going to use Bard or ChatGPT or they’re going to use Midjourney or whatever it is.

That independent layer which is not very tightly integrated with any of the large vendors is also one of things that they need to have a look at. It should not be so dependent on the offering of these vendors that in the future it’s impossible for them to move on to other areas of their business without having to change their backbone. This is also another important criteria. There should not be over dependence on a single API, a cloud or some service which prevents them from extending it through the organization which means that there’s another block for them to move from short term to long term.

The third thing that I would look at, when I try to get a vendor is to be able to partner with these large service providers. For example, as of today, Microsoft is the biggest player in the game in terms of the large service providers of Gen AI. They are one of the best and they’re more successful as of today in this field, but using their services also has a lot of glitches and roadblocks, especially if you’re not already onboarded with them. So if you want to create a new relationship with Microsoft in this particular domain there could be a lot of roadblocks. So it’s important to find vendors who are able to work with partners, but not completely dependent. So it’s a very thin line. While I would still say that it’s more important to have somebody independent and not too dependent but it should not be in a way that they are not able to leverage what is available in the right way that is good for the organization. It’s important that we have a mix of these three and it’s a subtle balance that the organization needs to look for.

AIM: Have you ever declined a client request based on fundamental capabilities, such as banking on certain parameters to build a solution? If so, why? 

Ananth Krishnan: One of the biggest reasons why we say no to any of the companies that we work with is their understanding of timelines. Their short term is sometimes so short that it’s impossible, and it doesn’t make sense to go that route. Many of the enterprises want to show something about Gen AI often to their management just for the sake of it. This is the only goal that they have; they don’t really have a strategy for what they want to achieve. It’s more like ‘I’ve done something with Gen AI and I want to show this to my top management.’

The other kind of use cases, where we usually easily say, ‘Okay, we can do something very small, but we’re not going to engage with you on this line of thinking; let’s think about the longest-term strategy. Let’s think of how this would work in the future. Let’s have a plan. Let’s have a roadmap. Then we can engage with you.’ This is what we usually say. Sometimes we are successful, sometimes there’s pressure from sales; we also need to close deals, make new deals, and there’s a bit of both.

For people who just want to do it for the sake of a buzzword, these are the deals which we try to stay away from. We have a clear defined boundary to what we do. We work on communication and conversations, and then we work on generation of images and 3D. If it is not aligned with our focus, and also within sales and marketing, we don’t go for use cases even for use cases which are allied, such as call center use cases. We have bots, but we never did bots for service, for example for somebody raising complaints etc. So we still have those same boundaries; we still work within those similar boundaries. Since we are a bit mature in that sense, because we have been in this field for five or six years.

So we know that we shouldn’t take up certain types of deals. That part is already there. This is an advice that I would give to any earlier stage founders, startups, that a definition of what you want to do, which are the verticals that you want to tackle and what are the problems that you want to solve even if there is potential for you to solve you should define it quite early. The way you define it is by having conversations with customers and really trying to find out which are the customers who are willing to invest in. Any customers who would say that you do this for me, but it’s only about a middle management presenting to their board. If it’s about that, it maybe difficult to identify, but after some time you’ll get an idea about that.

So initially, you need to not fall into the trap of trying to do everything but try to limit yourself to certain things. But nowfor us that problem doesn’t exist because we are already old enough, I would say, to avoid that kind of trap. We have our boundaries and we stick to that.

AIM: What characteristics do you think Chief Data Officers or data leaders from consumer industries need to look for in vendors, especially in the realm of AI for sales and marketing? Given the rapidly evolving landscape of AI, what trends do you envision emerging in the next few years for identifying the right vendor in the realm of Generative AI?

Ananth Krishnan:  So, sales and marketing, I would say, have three to four categories where Gen AI can have a significant impact. One of them is automating sales and marketing processes, wherein a large portion of the conversation about selling a product occurs before human interaction. This trend is expected to grow, leading to increased outreach and sales conversations.

Then there’s the marketing aspect, which involves rapidly generating personalized content for campaigns. Personalization here means tailoring content to individual preferences and past behaviors, going beyond just product recommendations to presenting products in a manner most suited to each person’s likes and dislikes. For instance, we’re exploring sending customized content via direct communication channels like WhatsApp, leveraging customer data to predict content preferences.

Another area is reducing reliance on marketing teams or agencies. Many companies spend heavily on agencies for tasks like creating brochures or designing campaigns. By using Gen AI, they aim to cut costs by automating tasks such as product descriptions and generating campaign ideas, enabling executives to generate ideas in minutes or hours instead of months.

Businesses are also leveraging Gen AI to personalize marketing campaigns to a level never seen before. With access to vast amounts of customer data, companies can predict the type of content that would best engage their audience. This includes considering factors such as whether humanized content, feature-related content, or even cartoonish content would resonate better with individual customers. By tailoring marketing content to each customer’s preferences, businesses can enhance customer engagement and drive sales.

Moreover, companies are exploring how Gen AI can help reduce reliance on external agencies for creative marketing tasks. By automating processes such as generating product descriptions and brainstorming campaign ideas, businesses can cut costs and streamline their marketing efforts. Executives can now generate a plethora of ideas within minutes or hours, a task that previously took months of brainstorming sessions.

Finally, there’s evaluating campaign success, where modern automation improves data interpretation and allows for quick A/B testing. This means marketers can compare multiple strategies at a lower cost, leading to better ROI analysis and potentially uncovering more effective strategies. Overall, these are the key areas where Gen AI is poised to revolutionize sales and marketing practices.

In summary, Gen AI is revolutionizing sales and marketing by automating processes, personalizing campaigns, reducing reliance on external agencies, and improving campaign evaluation techniques.