Main content

Can generative AI revolutionize Customer Support? Up to a point, says Freshworks President Dennis Woodside

Stuart Lauchlan Profile picture for user slauchlan May 25, 2023
Generative AI might well change the way organizations deliver customer service and support, but it's important to think through how this might work in practice.


Last week, telco BT dropped the bombshell that it would be axing 10,000 jobs in the next few years, including in the area of customer support, replacing them with AI. This will, apparently, result in a better customer experience from a company not exactly noted for its track record on that front. 

It’s perhaps one of the highest profile examples to date of a customer-facing organization getting overly-excited at the potential for AI - and particularly now, generative AI - in engaging with its most important assets - the customers themselves. 

By co-incidence, prior to the BT announcement, I’d had a chance to discuss how far AI can and will impact on the way customer service can be delivered in the future with Dennis Woodside, President of Freshworks, a firm that specializes in customer support and which has some major announcements on the role of AI in its product offerings coming up next month. 

Woodside’s starting point was one that chimed with my own foundational view on this topic: 

At the end of the day, empathy matters. If you're a consumer brand in particular, or if you're a B2B brand, that connection and that understanding of what my problem is and the willingness to do something about my problem, is really important. That's what you build a brand on. I don't think many brands are going to outsource that to a Large Language Model. 

But what they may do, and what we are working on, is they may empower their agents with the benefit of that Large Language Model so that the agents can do their job better and bring the right attitude, the right empathy to the customer. 

He cited an example of having a ‘suggest’ feature within Freshworks Customer Support offering, something that can come up with possible answers to an agent in response to complex queries that might not be answered in a standard FAQ today: 

One of our customers in the US is Guitar Center. They sell guitars. Well, 'How do I string a guitar?' might be a question that could come in through a Customer Support line, because you just bought a guitar from this company. That might not be something that is in a normal FAQ, but you can quickly get an answer on it using a Large Language Model. That is perfectly appropriate for the customer. We want an agent to moderate that, to see it and decide, 'Yes, I'm going to say it', because we don't think that generative AI is ready for that kind of primetime interaction with the end customer. That's how we're thinking about it as a starting point, at least. 

Another thing that we're doing is a summarized function where, if I've had a back and forth with a Customer Support rep over multiple channels and multiple days, snap your fingers and you can summarize that interaction in a paragraph, which is very useful. If that escalates to my manager or my manager wants to get involved, the first question that manager is going to ask is, 'What happened, what's happening, what's going on?'. 

So we think that there's a lot of applications, but we also want to be smart about how we're rolling it out. We have a number of features in beta today with about 380 customers that are using that kind of beta functionality, and it's all in furtherance of making the occasion more productive.

Buyer beware or buyer be wary? 

The most important thing to keep in mind with new tech should be the Informed Buyer. While it’s now de rigeur for tech vendors to talk up their generative AI story in public at every opportunity, this evangelical fervor on the sell side needs to be balanced against the receptiveness of the buy side end user it’s being preached at. Woodside reckoned that this may not be as high as the current hype levels suggest: 

If you talk to customers, generative AI is interesting to them. Is it the number one criteria by which they're making their decisions today? It's not at all. Are they are gung ho, 'We're gonna light this up for our customers?'. Hell no, because they've played around with it too and they understand that, at least the current technology, by admission of [OpenAI’s own] CEO, is not reliable enough to provide truthful answers all the time, and you need to do that as a brand. [Generative AI] is now table stakes, but there's a lot more that people are making their buying decisions on.

Freshworks clearly intends to proceed with due caution. Woodside said:

What has made us successful yesterday is still going to make us successful tomorrow. We'll just incorporate AI into our product in the right way over time...We built a generalized platform that has a lot of new technologies that have come along that we plugged in.

At first Freshdesk was a ticketing system, and bots came along and bots became the thing. So now we incorporate bots into the product, right? It's another feature that that we offer. That's going to be true with AI. Some of that AI will be based on Large Language Models, some of that AI we already offer today that's based on a smaller subset of data. 

It’s also going to be important not to assume that AI is some universally-applicable ‘silver bullet’. Woodside noted: 

There will be business departments for which it's not appropriate. There will be use cases for which it's actually quite useful, quite helpful.

That said, there are already clear cases where generative AI tech has been a disruptive force. Woodside points to Chegg, a US education tech company that provides a service to students to help them with their homework, in the form of automated answers to questions: 

They said that their business basically got blown up in March, when ChatGPT started becoming more and more popular. So there are businesses that are going to be directly impacted. But I think what we do is much broader. It's more about a platform for managing workflow and interactions  between requesters and responders. That's a much more sophisticated solution than what Chegg was doing, which was based on Q&A. 


There are also still a lot of unknowns to be addressed, Woodside observed:

I do think it is the technology of the day…We do see it as something that's going to change how customer support is done, but we still have yet to see how that plays out in terms of, is it a complete different pricing model? Are we going to have half the number of agents as we have today? We think some of those statements might be a little overblown, although we don't yet have the answer either.

And there are also, inevitably, wider societal concerns about the impact on trust: 

I find myself, even when I'm looking at video content today, being very distrustful unless I know I'm on the Wall Street Journal site or something like that, because there's so much fake stuff. You're starting to see Deep Fake material just pop up in my Instagram feed. So I think that's really worrying. But I think it potentially creates more opportunity for trusted brands to continue to build businesses on top of the trust that they've earned. If that doesn't happen, think about a world in which we can't trust what we're reading. That's a challenge.

My take

On the last Freshworks earnings call, there were a lot of questions about AI from investors and analysts, as there now are on every such occasion with almost every vendor. CEO Girish Mathrubootham shared some thoughts on the tech in response, stating: 

Where we are excited and actually investing right now, which is a priority for us, like in this quarter and the coming quarters, is how do we actually use generative AI to think about improving our ability to help businesses deliver better customer experiences and better employee experience?  So we are thinking along the lines of, how can generative AI help customers self-serve themselves? How can it help customer service agents solve customer problems or leaders actually get better insight? And we're doing this across not just our CX products, but also across our IT and sales and marketing products.

This is a journey that is not new, but definitely the question is being asked louder after ChatGPT. But in 2016, we embarked on this journey where we foresaw a future where the industry is moving more towards bots and automation. I think the reality of, let's say, our biggest market, which is customer service and support…the reality is level one support always will have a lot of deflection and automation, and the higher order support will actually always require some amount of human touch personalization and subject matter expertise.

This was just an appetizer for a more detailed roadmap event on 22 June, where the company intends to share developments that are already in beta, as Woodside referenced, and its thinking on what might be to come. We’ll take another look at what that holds after that. 

A grey colored placeholder image