Leveraging AI to make data accessible across Freshworks

Tarush Aggarwal (00:00)
Welcome to another episode of People of Data, where we get to highlight the wonderful people of data. These are our data leaders from various industries, geographies, and disciplines. Discover how companies are using data, their achievements, and the challenges they face. Today on the show, we have Sachin Mishra.

who is a leader of data science and AI at Freshworks. He leads strategic decision science to drive AI guided sales and marketing. He's got 18 years of experience, worked at companies like Wipro, Amazon, Cognizant. Sachin describes himself as a true practitioner and loves solving real world problems with an application of scientific knowledge. He's also stayed away from a lot of the buzz-worthy.

type stuff in the AI world which is one of reasons we're super excited to have him on the show today. Sachin, welcome to the show. Thank you, Tarsh. It's my pleasure to be with you today. I'm looking forward to our conversation. Let's do it. Are you ready to get started? Yes. Peace. So, you know, I just had a high level. What is Freshworks?

‍

Sachin Mishra (00:50)
Thank you, Tarush. It's my pleasure to be with you today and looking forward to our conversation.

Yes, please.

‍

Tarush Aggarwal (01:05)
How does it make money and how does data play a role in helping Freshworks with what it wants to do? Yeah, that is so Freshworks basically delivers modern AI guided customer and by experience solutions. think today around 68,000 companies in 120 countries, I think choose Freshworks for uncomplicated solutions.

‍

Sachin Mishra (01:12)
Yeah, that is so Freshworks basically delivers modern AI guided customer and employee experience solutions. I think today around 68,000 companies in 120 countries. I think choose Freshworks for uncomplicated solutions which deliver rapid time to value. American Express, Bridgestone, Databricks, Sony.

‍

Tarush Aggarwal (01:33)
which deliver rapid time to value. Express based on Databricks, Sony, are some of the things using today's Freshworks AI customer and by experience solutions. Our primary product are around IT service management. think IT assist management and as well as.

‍

Sachin Mishra (01:40)
are some of the things which are using today Freshworks and its AI powered customer and employee experience solutions. Our primary product lines are around IT service management, think IT assist management and related categories, as well as customer experience, omni-channel products and sales and marketing related solutions.

‍

Tarush Aggarwal (01:58)
customer omnichannel and sales and solutions. Now, Freshworks is as a market partner in San Mateo, California as a global looking help our customers across the

‍

Sachin Mishra (02:06)
Now, Freshworks is, as of now, headquartered in San Mateo, California. I think we operate as a global team. We are looking to help our customers across the globe. I think we do offer our entire suite of AI-assisted,

‍

Tarush Aggarwal (02:19)
I think do entire suite of AI agents and related to make it easy for companies at any scale, right? customer experience to their customer and employee experience to customers and frequently their employees itself. Got it.

‍

Sachin Mishra (02:24)
agents and related solutions to make it easy for companies at any scale. They provide the best possible customer experience to their customer and employee experience to respective customers and subsequently their employees itself.

‍

Tarush Aggarwal (02:39)
What's the culture like? Obviously, big company, traditionally, a lot of it was built in India, past or part of the built in India, built for the world playbook. What's the culture? But at the same time, in the last few years, expanded, having people actually globally. What's the culture like at a company level? How does this map into the culture you're creating inside the data org?

So, you if you have people who are listening into the show who are potentially considering careers at Freshworks and especially the data team, how should, you know, how would they, how should they be thinking about this? that's a question, Tarush. I think Freshworks obviously has roots or from Chennai and India. over the period of we have

‍

Sachin Mishra (03:12)
that's a great question, Tarush. I think Freshworks obviously has strong roots or origins from Chennai in India. over a period of time, we have truly embraced our cultural philosophy, is Kutumbua. So think we treat the entire world, including our employees, as part of extended family. think we are practicing that.

‍

Tarush Aggarwal (03:23)
truly embraced our culture, which is Kutumbua. So I think we entire world, including our employees, as part of extended family. we are practicing that day in and day out in our culture as As basically we have come out of initial stages of, software as a service platform.

‍

Sachin Mishra (03:36)
day in and day out in our culture as well. As basically we have come out of initial stages of modern, software as a service platform, we have paid basically attention to the details while you go and scale softwares at this scale. think our strength still remains in very uncomplicated.

‍

Tarush Aggarwal (03:48)
we have attention to the details go and scale this scale. think our strength still remains in very uncomplicated, simple to use solutions. we a lot of attention as well as take a lot of pride in building sophisticated solutions, which are really simple and easy to use.

‍

Sachin Mishra (04:00)
simple to use solutions. So we pay a lot of attention as well as take lot of pride in building modern, sophisticated solutions, which are really simple and easy to use. it's much easier said than done. Building a simple software is way more complicated and challenging than it is given credit.

‍

Tarush Aggarwal (04:13)
much easier said than done. Building a simple software is way more complicated and challenging it is given to. Yeah. And what's the culture like on the data team? know simplicity and simplifying the experience for your customers is, as you mentioned, paramount importance. But what's the culture like in the data team? And, you know, we've always seen that

individual teams have their own quirks and have their own sort of ways of doing things. We've had some great examples on the show before. What's one example of something which you see sort of is present in the culture of the data team which might be different from the culture of the global organization?

‍

Sachin Mishra (04:52)
I'll simply say that I think as a data team, we practice something which is inherently as part of Freshworks overall culture, which is agility with accountability. In this kind of environment where there is no shortage of, I think, data for

‍

Tarush Aggarwal (04:52)
I'll simply say that I think as a data team, we practice something which inherently as part of thresholds overall culture, which is agility with In this kind environment there is no shortage think, data footprint across your employee and customer lifecycle journeys at we all have been guilty of getting caught up into analysis paralysis and

‍

Sachin Mishra (05:10)
print across your employee and customer lifecycle journeys at times. we all have been guilty of getting caught up into analysis paralysis. And you can go and dig as much as you want, but is it really, allowing your stakeholders having access to the right information at the right point of time for them to be able to make informed decisions? at the data team,

‍

Tarush Aggarwal (05:18)
think you can go and dig as much as you want, but is it really, allowing your stakeholders having access to the right information at the right point of to be able to make informed decisions? especially on the decision science function that I mean, I think we somehow try to balance that thing here.

‍

Sachin Mishra (05:33)
especially on the design decision science function that I lead. think we really some try to balance that thing here and practice agility with accountability where I think are we able to deliver the immediate value at the time of decision making without getting caught up into the complexities of the board.

‍

Tarush Aggarwal (05:41)
and practice agility with accountability where I think I'll be able to deliver that immediate value at the time of decision without getting part of the complexities of the work. Yeah, that makes sense. you know, talking about the data team, what's the footprint like? know, how big is Freshworks in terms of number of employees, in terms of the data group, how big is it? What's the split?

between data engineering, science, analytics, whatever functions you have. Yeah, so Thresholds as of now has around 5,000 plus employees across the globe. In terms of data footprint and data it's a usual structure where you have embedded data related functions and teams as part of your product.

‍

Sachin Mishra (06:10)
so thresholds as of now has around 5000 plus employees across the globe in terms of data footprint and data teams. it's the usual structure where you have embedded data related functions and teams as part of your product engineering and product analytics groups. Then we have central.

‍

Tarush Aggarwal (06:32)
engineering and product analytics groups. we have platform data teams, are more heavy on engineering or data engineering function side. you have, in a way, at the edge kind of teams, edge teams the team that I run, which is focused on helping fresh works,

‍

Sachin Mishra (06:37)
platform data teams, which are more heavy on engineering or data engineering function side. Then you have innovate at the edge kind of teams, edge teams like the team that I run, which is primarily focused on helping Freshworks truly enter the enterprise market globally by allowing enterprise grade decisions.

‍

Tarush Aggarwal (06:54)
enter the enterprise market globally by enterprise-grade decisions prioritizations and other be we enter the market and grow. Yeah, in terms of size, you know, out of the 5,000-odd employees, how many people are employed in the data domain? collectively, you know, right, I think data word has evolved and

‍

Sachin Mishra (07:01)
and prioritizations and other stuff to be made as we enter new markets and grow our scale.

collectively, as you know, right, think data word has evolved and there we have added more and more nuances and variations into the role. So if I combine, right from analysts, data product managers, think certain other teams, think we may have, few hundred people

‍

Tarush Aggarwal (07:18)
there we have added more more nuances and variations into the role. So if I right data product managers, think other teams, think we may a few hundred I think we still have ecosystem.

‍

Sachin Mishra (07:31)
I think we still have very distributed ecosystem, which is aligned more for speed and agility as per the teams, et cetera. So I think it's a very hub and spoke model where we have some sent live teams as well as teams which are at the edge to support their respective business functions.

‍

Tarush Aggarwal (07:35)
aligned speed and agility as per the teams. So I think it's a very hub and spoke model where we have some kind of live teams as well teams which are at the support their respective business functions. Yeah, that makes sense, right? Like very typically, just very, very, you know, sort of table cloth math, we see typically technology functions inside organizations being between 20 to 25%.

And out of that data and technology function being about 20% of data. So very typically, we're 4% of a company size. At a company size of sort of fresh works, that would ballpark, put it at about 200 employees out of 5,000. You mentioned a few hundred. So probably as we analyze this in line with what we expect of a 5,000 person org, what do you think is like, what's the breakup between, know, and...

We see like larger businesses go through this all the time. What is centralized versus what is decentralized, right? The more stuff which is centralized, you can have higher quality by centralizing it, but it becomes a little bit slower and then decentralized, you know, the fragmentation of it. So how do you So, Parish, unfortunately there is no magic pill for it. I've seen it. I've data organizations from scratch to a few hundred people.

‍

Sachin Mishra (08:41)
Yeah, yeah, so that is unfortunately there is no magic pill for it. I've seen it. I built data organizations from scratch to a few hundred people. I think with AWS in the last 10 years with Amazon. So I feel like it shifts or changes as per as your organization grows.

‍

Tarush Aggarwal (08:53)
with AWS in the last 10 years, with Amazon. So I feel like it shifts or changes as your organization grows and how you are basically aligning yourself as per your product lines, strategy goals and market thing that has worked over the years is staying flexible.

‍

Sachin Mishra (09:04)
and how you are basically aligning yourself as per your product lines, strategy goals and market conditions. For me, the thing that has worked over the years is staying flexible and not over committing to either of these models because as your business needs and your overall landscape shifts, just like any other organization, right?

‍

Tarush Aggarwal (09:16)
and not over committing to either of these models as your business and your overall landscape shifts, just like any other organization, data teams also need to evolve. In some the data teams staying closer to the business for certain works well in early days as you scale up.

‍

Sachin Mishra (09:28)
Data teams also need to evolve in some cases. The data teams staying closer to the business for certain functions, works well in early days as you scale up, I think you start getting into operational excellence and other challenges. And that is the right time to think about some of the centralization and other pieces. I am far believer in that, Tarush I myself have not found an ideal solution

‍

Tarush Aggarwal (09:39)
start getting into operational excellence and other challenges and that is the right time to think about some the centralization and other pieces. I am top believer in that, I myself have not found an ideal solution. Neither I tried to recommend getting into this debate of centralization versus distributed teams. I love that. I love the flexible approach.

‍

Sachin Mishra (09:53)
Neither I try or recommend anybody getting into this debate of centralization versus distributed teams.

‍

Tarush Aggarwal (10:05)
over there and you know being open to fundamentally solving the business problems as opposed to being tied up to one of these approaches. I think it's a great answer. What is you know you think about your technology stack particularly your data technology stack you know what is the stack today and then as you can guess my follow-up question would be like what is the flexibility the on-edge teams have?

‍

Sachin Mishra (10:20)
Mm-hmm. Mm-hmm.

‍

Tarush Aggarwal (10:27)
as opposed to using some of the central data stack versus potential niche tools which might be more relevant for you. Yeah, so just like any other global organization, I'll say that it's a multi multi-platform there is flexibility for respective right from some of the

‍

Sachin Mishra (10:34)
Yeah, so just like any other global organization, I'll say that it's a multi cloud, multi platform driven, ecosystem. there is flexibility available for respective team. right from some of the

‍

Tarush Aggarwal (10:51)
customer facing teams such as marketing teams, et cetera, are using more market properties. I think that then there is central product and engineering where I think you need care about scale and volume of the data. There are some streaming specific solutions, et cetera, because ultimately if you are enterprise grade customer and employee experience

‍

Sachin Mishra (10:52)
customer facing teams such as marketing teams, etc. are using more market-property tools. Then there is central product and engineering where you need to care about scale and volume of the data. There are some streaming-specific solutions, etc. Because ultimately, if you are providing a scaled enterprise-grade customer and employee experience solutions with AI as a

‍

Tarush Aggarwal (11:18)
primary you need to take care of some of those scaling challenges for your product as well, et it's a very, I'll say multi multi some of the central data lake or related functions are getting streamlined and nowadays with emphasis on more modern.

‍

Sachin Mishra (11:18)
primary feature, you need to take care of some of those scaling challenges for your product as well, et cetera. So it's a very, I'll say that multi cloud, multi platform ecosystem. some of the central data lake or data warehouse related functions are getting streamlined. And nowadays, with emphasis on more modern data capture.

‍

Tarush Aggarwal (11:42)
data capture approaches, et cetera. However, edge teams more or flexibility to some of the options, how they are and serving their own customers. What do you use? What's your data warehouse? What do you guys use for ingestion? What are some of the BI tools? What does the de facto data platform look like today?

‍

Sachin Mishra (11:43)
approaches, etc. However, edge teams more or less have flexibility to choose some of the options, how they are consuming and serving their own customers.

Yeah, so over the period of time, the overall data platform has evolved around Databricks as the primary tool. I think we have had a snowflake. there is a lot of AWS and GCP behind the core as well as for any, data platform. At the same time, address some of the scaling challenges, we do have

‍

Tarush Aggarwal (12:04)
Yeah, over the period of the data platform has evolved Databricks as the primary tool. think we have had a snowflake. there is a lot of AWS and GCP behind the core as well as platform. At the same time, address some of the scaling we do

‍

Sachin Mishra (12:27)
some of the other open source options as well, lot of investments and around some of the, vector DBs, et cetera. as I said, it's a multi-cloud, very distributed ecosystem with all possible situations.

‍

Tarush Aggarwal (12:27)
some of the other open source options as well, lot of investments and around some of vector DBs, et as I said, it's a multi cloud, very distributed ecosystem with all situations. And in terms of BI in particular, do you have a central BI tool across the organization or do you also have multiple offerings?

‍

Sachin Mishra (12:49)
I think primarily we have been on here with the scale and I think obviously for the considerations of operational efficiencies. In terms of BI, if I define as end user reporting or dashboarding capabilities, it's primarily Power BI off late that I have seen. At the same time, there are other

‍

Tarush Aggarwal (12:52)
I think primarily we have been on here with the and I think obviously for the considerations of operational terms of BI, if I define as end user reporting or dashboarding capabilities, it's primarily power that I've seen. At the same there are other heap analytics and other users.

‍

Sachin Mishra (13:15)
heap analytics and other user facing solutions, which are also available. we do have our own products having native reporting capabilities itself, et cetera.

‍

Tarush Aggarwal (13:17)
facing solutions which are also we do have our own products having made the capabilities itself And now what you said about you know at an organization of your size it's no longer about do you use Databricks or Snowflake, it's very often both right? How do you,

do you really sort of differentiate between them and is the idea ever to have a consolidated one?

‍

Sachin Mishra (13:40)
Yeah, that's right.

Again, I'll say that there is no right or wrong answer here. my simple take is as any organization evolves, especially in modern software as a service world, AI monetization had been something that has been there.

‍

Tarush Aggarwal (13:43)
I'll say that there is no right or my simple take any organization evolves, in modern software as a service word, AI monetization had that has been there.

‍

Sachin Mishra (13:57)
for last eight to 10 months, as you go down the food chain. It's easy for chip manufacturers, et cetera, to make money. But ultimately, for end user applications, some of the products that we have, think AI monetization has been seen as something which could be make or break for industry as well, after all the hype and everything. So I think for us,

‍

Tarush Aggarwal (13:57)
for last eight to 10 as you go down the food chain, right? It's easy for chip manufacturers, et cetera, to make ultimately for end user some of the products that we have, I think AI monetization has been seen as something that which could be make or break for industry as well after all the hype and everything. So I think for

‍

Sachin Mishra (14:23)
how relevant and how easy and how efficient we could be around our end-to-end data, which ultimately, is the food for some of our features and other stuff that we have been building for. So it's more or less how efficiently, how we can basically optimize the overall cost and scale with a lot of open source flexibility, et cetera.

‍

Tarush Aggarwal (14:24)
how relevant and how easy and how efficient we could be around end-to-end which is food for some of our features and other stuff that we have been building So it's more or less how efficiently, how we can basically optimize the overall

scale with a lot of open source flexibility, et cetera.

What is one aspect which you're really proud of? Like what is one thing which the team has done, which has had a, you know, which your data group has done, which has had a huge impact in the business? it's a relatively new function here my goal was how I can basically move some of these to be informed more in

‍

Sachin Mishra (14:54)
It's a relatively newer function here where my goal was how I can basically move some of these decisions to be informed more in proactive way. right from consuming data after the fact for reactive decision making, hey, can I bring?

‍

Tarush Aggarwal (15:07)
right from consuming data after the fact for reactive decision Can I bring some of these things earlier in the overall decision making life cycle, et cetera? So it's a cultural transformation that we have gone of the things that, one of the first things earlier in the year, as I was talking most of the tech

‍

Sachin Mishra (15:16)
some of these things earlier in the overall decision-making life cycle, etc. So it's a cultural transformation that we have gone through. One of the first things earlier in the year, as I was talking about, most of the tech, you can say word, was setting expectations for markets and other stuff in terms of AI monetization.

‍

Tarush Aggarwal (15:33)
you can say was setting markets and other stuff in terms of AI We were ahead of the curve, so we had roughly 60,000 customers in beta trials of our AI Already, think, since Q4, Q3, and Q4 up last

‍

Sachin Mishra (15:40)
We were ahead of the curve. So we had roughly 60,000 customers in beta trials of our AI solutions. Already, think, since Q3 and Q4 of last year. So I think for us, the decision was, while we are setting some of the market guidance, et cetera, how should we basically have a very thoughtful approach?

‍

Tarush Aggarwal (15:57)
I think for us the decision was, while we are setting some of the market guidance, et cetera, how should we basically have a very thoughtful approach of value? And which are these customers who are the right customers where we can optimize the both thinking from their perspective as well and it being a cold start problem where you can mimic a of purchase behaviors.

‍

Sachin Mishra (16:06)
to proof of value and which are these customers who are the right customers where we can optimize the value, both thinking from their perspective as well as ourselves. And it being a cold start problem where obviously you can mimic lot of purchase behaviors, but identifying especially in the current macroeconomic situation who will be the real customers who are ready to pay for some of the talk about AI solutions could be challenged.

‍

Tarush Aggarwal (16:22)
But especially in the current macroeconomic situation, who will be the real customers who are ready to pay some of the talk about AI solutions could be a challenge. Because the overall macroeconomic landscape, I think everybody was going for more selective and investment strategies. People wanted to results.

‍

Sachin Mishra (16:33)
the overall macroeconomic landscape, think everybody was going for more selective pricing, investment strategies, people wanted to see results of investment or returns on investment first before making long-term commitments, et cetera. So identifying those things, the right customers at the right time for having those monetization conversations.

‍

Tarush Aggarwal (16:43)
group of investment or returns on investment first before making long term commitments, et cetera. So identifying those things, the right customers at the right time for having those monetization conversations to be able to convert them trial to paying and setting the tone for the future as well as setting the right around the AI-powered features

‍

Sachin Mishra (16:58)
to be able to convert them from trial to paying customers and setting the tone for the future roadmap, as well as setting the right expectations around AI-powered features was one of the most meaningful things we did just within few days against all odds without having historical purchase behavior data or any third party intent.

‍

Tarush Aggarwal (17:10)
was one of the most meaningful things we just within few days against all odds without behavior or party. We need to reinvent some of the go ahead and chase apply base as well as I think some of based propensity approaches.

‍

Sachin Mishra (17:22)
I think we had to reinvent some of the views, go ahead and chase and apply some hybrid heuristic base as well as I think some of the ML based propensity modeling approaches to finally, reduce the subset and we were able to guide our go-to-market teams to have that customer cohort base for prioritized discussions, which led to more meaningful

‍

Tarush Aggarwal (17:34)
to reduce the be able to go to market teams to that customer cohort base for prioritized discussions, which led to a meaningful proof of value framework and later on, I think, to selective decision making. I love that your solution kind of added a component to it, which is also sort of cultural. Very often on the show, we hear like, you

‍

Sachin Mishra (17:46)
proof of value, selective decision making.

‍

Tarush Aggarwal (17:59)
very concrete examples of here's one feature which we delivered which had XYZ impact so it's really good to see combining both technology as well as people in process so that's awesome. What's one of the, on the flip side of this, what's one of the challenges which you've faced, again you mentioned it's a new role, what's one of the challenges with your face which you are working through today?

‍

Sachin Mishra (18:20)
Yeah, again, we are dealing with some of the perennial problems as well. For example, sales forecasting is not a new problem. However, given the volatility in the technology market as well as overall macroeconomic situation, especially when you don't have stable time series or historical revenue or bookkeeping information with

‍

Tarush Aggarwal (18:21)
we are dealing with some of the perennial problems as well. For example, sales forecasting is not a new However, given in the technology market as well as overall macroeconomic situation, especially when you don't time series or historical revenue or bookkeeping information, but

‍

Sachin Mishra (18:43)
in early stages of your product market maturity or S-curve, think organizations go through changes. modern, basically technology or SaaS word could be volatile as well. So that's one of the things we have been trying how we can bring in some of the macroeconomic indicators and really predict the course of our action.

‍

Tarush Aggarwal (18:44)
In the early stages of your product market maturity or S curve, think organizations go through changes and modern basically technology or SaaS could be volatile as well. So that's one of the things we have been trying how we can bring in some of the macroeconomic and really predict the course of our action

‍

Sachin Mishra (19:07)
Apart from that, as the market and tools and our modern generation or customer base is getting savvy, right? Where not everybody really wants to interact with sales and marketing related, maybe historical notions or business processes, how to build a more fully.

‍

Tarush Aggarwal (19:07)
Apart from as the and tools and generation of customer base is getting savvy, Where not everybody really wants to interact sales and marketing maybe historical notions or business processes, how to build a more sales and marketing business processes.

‍

Sachin Mishra (19:26)
sales and marketing business processes, where I think we can minimize the interactions or touch points. think deliver very personalized buying and selling experience and how to not only sell our customers, AI and related features and solutions, how to really use it and demonstrate the value.

‍

Tarush Aggarwal (19:28)
where I think we can minimize the interactions or touch points. I think deliver very buying and selling and how to not only sell our AI and related features and solutions, how to really use it and demonstrate the

That makes a ton of sense. Sachin, thank you so much for being in the show and have a wonderful rest of your day. Thank you, Tarush. It was pleasure and I look forward to future such conversations. Thank you. Bye-bye. We're just going to wait a few

‍

Sachin Mishra (19:53)
Thank you, Tarush. It was pleasure and I look forward to future such conversations. Thank you. Bye bye.