Freshworks CEO Dennis Woodside: We can be the ServiceNow alternative

Dennis Woodside has helped several iconic Silicon Valley companies scale into major power players, namely Google and Dropbox. A year into his tenure at Freshworks, a push to move beyond small and medium-size businesses and cater to bigger companies is paying off.

Shares of Freshworks rose nearly 10% last week after the customer-service and IT-management software company reported earnings that beat Wall Street estimates while raising its revenue guidance for the year. After joining the company as president in 2022, Woodside took over for founding CEO Girish Mathrubootham last May to help "the first India-born SaaS firm to trade on a US exchange" reach new customers.

Under Woodside, Freshworks is taking direct aim at ServiceNow, which has been the leading company across several aspects of IT management software for years. "Three years ago, our product wasn't anywhere near as enterprise-ready as it is today," he told Runtime in an interview last week, and he's betting that Freshworks can continue to win over larger customers: "They need an enterprise-grade solution that's AI-enabled, but a solution like ServiceNow, which is really more suited for J.P. Morgan, is not going to work particularly well for them."

Woodside discussed the company's product strategy, why pricing for agentic AI is likely to continue to evolve, and Freshworks' potential for expansion into other enterprise software markets like HR and asset management.

This interview has been edited and condensed for clarity.

Given that it's the anniversary of you becoming CEO, what needed to change when you came in and what have you managed to get changed?

Woodside: I've been in tech for 25 plus years, worked at Google, Dropbox, and helped scale those organizations as well. I think with Freshworks, before I came in we had a great set of products in markets that are must-have markets. If you have a customer-support team or an IT team, you have to automate their operations. You have to apply AI.

The business has been built so that the market that you can go after, and the use cases that you can solve, are very broad because we have a generalized workflow engine and AI that can apply across that workflow engine. Those things were really important to me.

I think what we've been able to accomplish, number one, is that the motion up-market is absolutely working. Our average revenue-per-customer is up 10% year over year. Two years ago, a $100,000 deal was a big cause for celebration. Now we do many, many of those every quarter. And I like that. I like that ability to succeed with larger accounts which have a bigger impact on the world.

I think the second big thing that we've done in the last year that's notable is we really have leaned in and delivered products that monetize on AI. AI Copilot was just introduced into GA a year ago February, and we already have 2,700 paying accounts. Those accounts are using the product intensively. We have over a million calls to the Copilot product every single day from the 2,700 accounts.

One thing I've noted as I've talked to a lot of companies that are thinking about implementing AI apps or agents is just how unprepared they were for how much work they needed to do to get their data ready for this kind of service. How do you work with that?

Our differentiating proposition for AI Agent in particular is that you can be up and running in minutes. You don't need a structured database, you don't need an answer center. You point it to whatever content you want the agent to know — it could be PDFs of product manuals, it could be an answer center if you have one, it can be previously asked questions to your customer support team and the answers —  so it learns all that that very quickly, and then it can answer any question pertaining to that content.

In a service operation, writing the knowledge-base articles is a pain in the ass, and people don't want to do it.

The other thing that our AI will do is where it identifies new or novel questions that have not been asked in the past, but where it's starting to see a pattern. It will suggest that you create a knowledge base entry, it will pre-populate the knowledge base entry, and at a click of a button, you can publish it. That also is really valuable because in a service operation, writing the knowledge-base articles is a pain in the ass, and people don't want to do it.

One thing I wanted to ask you about was the pricing: A lot of people I've talked to recently are considering different pricing options for agentic AI products pointing toward a consumption-based strategy, but it sounds like you are bringing these AI capabilities in to some customers as an additional per-seat charge?

There's three different pricing models. For AI Agent — that's the front-line agent — it's consumption based: You purchase packages of sessions and it's 1000 sessions for $100. That's how we price it today. Now that probably is going to evolve and change; as the value of the AI increases, that will probably increase the price, but that's what we charge today.

For Copilot, which is the agent assistant product, we charge $29 per month per seat. And for Insights, we're wrapping that into the enterprise plan, because that really is for the more sophisticated buyer.

If you can free up 10% of a $100,000 person's time, that's worth $10,000 to the customer. What's the fair share that we should capture versus the customer?

Different agents have different values. What we're in beta with now for AI Agent is a set of agents that are very specific for ecommerce and for travel that are deeply integrated into systems of record in those industries. We may decide that the value that we're creating there — because customers can change their order on the fly, they can add to an existing order on the fly, or change their travel plans on the fly — the value might be so great that we charge more.

I think the charging models are going to definitely change over time because, at the end of the day, the software is enabling the people to be more productive, and the people are very expensive. So if you can free up 10% of a $100,000 person's time, that's worth $10,000 to the customer.

What's the fair share that we should capture versus the customer? That's kind of how we think about setting pricing in the right way for an AI agent.

When it comes to developing your own AI models, is that something you think you need to do? How do you think about the models that you use to power some of these capabilities?

There's a ton of innovation happening at the LLM layer. So today, we have over 40 large and small language models that we rely on; we rely on OpenAI, or [Microsoft] Azure's version of OpenAI, for lots of use cases. But we also rely on Google for images, we look at Anthropic for coding, we have small language models that we use for things like ticket routing and we use [Meta's] Llama for a couple of uses, too. That gives us flexibility as prices change.

We're mostly concerned about performance today. We haven't built our own large-language model. I don't think we will, but there are cases where it makes sense to have a smaller model, or take an open-source model, modify it on [our] server, and serve it from that server because it's cheaper.

You mentioned ServiceNow and trying to go a little bit more upmarket into the enterprise world; As you think about that, are you also looking at other areas of SaaS, other things beyond the customer support and internal support things that you've been known for, taking on a Salesforce or a Workday?

One of the things I've intentionally done is focus the company. We actually have a sales-automation product for small business and a marketing-automation product for small business. But we're not big enough to support world-class development in that many fronts, and I needed to focus the engineering teams on customer support and on employee experience, or the IT business. Within each of those spaces, though, there's tons of adjacencies that we can build into.

An example: We're going to be launching a product for managed-service providers, and that's a big segment of the market. Lots of managed-service providers already use our product to serve their customers, but we don't have a layer that allows them to manage multiple workspaces or consolidate billing. And so we're going to launch that. Another example is we've done a lot of work outside of IT with ESM, and we have a generalized workflow engine that works in any department today.

We're going to go deeper into HR in particular, because there's a lot of requests that we get from customers to do things like onboarding and off-boarding in a seamless, automated way, using AI and workflow automation.

Device42 is our advanced IT Asset Management solution that I think is a source of data for future AI innovation that is relatively untapped. A customer of ours is New Balance. Device42 knows every asset in the New Balance workspace, whether it's in the cloud or [on premises], and all the relationships in between software and hardware. There's a lot of insight you can derive from that for things like compliance. If you're in Europe, you need to know the heat footprint and the thermodynamics of all your servers. You need to know the energy consumption. You need to report on that [and] you get audited on that. Well, we could build a service line around that.

We could go into spaces like Snow Software is in with software asset management; we know all the software, we just need to supplement that with things like the licensing provisions and so forth in order to compete in that space.

These spaces that we're in, there's a lot of adjacencies over in the [customer experience] side, things like workforce management, where we don't have a product today. We could build one, or potentially acquire one. That's kind of how we're thinking about innovation, and that I think is plenty of avenues to go and pick off the next workload that our customers are asking us to help.