Your browser doesn’t support HTML5 audio
Zendesk last week closed one of its biggest-ever acquisitions in a bid to stay ahead of the move to agentic AI in the Customer Experience (CX) market. Buying Forethought brings one of the leading pioneers of agentic CX into its fold. Launched in 2018, it has built an impressive roster of enterprise customers, many in the digital technology field, including Airtable, Cohere, Cotopaxi, Datadog, Grammarly, Lime, UPS, Upwork, Fiverr and WordPress.
Unlike traditional CX platforms, which rely on pre-built decision trees and workflows, Forethought is trained on a CX organization's historical data and autonomously identifies knowledge and generates workflows to resolve customer issues. It co-ordinates the actions of various agents to automatically discover insights, resolve queries, hand off to human agents when needed, support them as they help customers, and perform QA oversight. It works across multiple channels, including voice, and has API connections to fetch information and complete actions in a wide range of other enterprise systems.
While Zendesk has already been investing in AI that is able to improve autonomously over time, the company says the acquisition will accelerate the progress of its product roadmap by over a year. Adrian McDermott, Zendesk's Chief Technology Officer, tells me:
They have just great technology that we can add into our roadmap, in terms of just the ability to have easy deployment [of a] continuously learning AI agent... They're already deployed across a bunch of different channels, like email and voice. They've already built what we call procedure generation, or autoflows, which is the ability to basically go in to a company's conversation history or customer service ticket history, and figure out what are the things that should be automated, and how should they be automated, and pre-build all of those for our customers, saving them a ton of time, and also a ton of learning, and bringing them on.
As automation lowers CX cost, demand rises
The acquisition is the latest and boldest in a series that have sought to boost Zendesk’s AI capabilities. December saw the acquisition of Unleash, an AI-powered enterprise search platform that will support Zendesk’s employee service offering. Other acquisitions include last year’s purchase of cloud-native CCaaS (Contact Centre as a Service) platform Local Measure, and 2024’s acquisitions of Ultimate, an AI-powered customer support automation platform and quality management platform Klaus.
These moves reflect the speed at which CX is evolving in the AI era towards ever higher levels of automation. McDermott reflects:
Twenty years ago, customer service was an oral tradition. People all sat in the same room, and when something was going wrong, I would write it on the whiteboard, like a cave painting, and we would all have a conversation, and it was lovely.
If your AI agents are off running the show... the only way to solve a problem in AI is with more AI. You need all these mechanisms, basically, of inspection and learning and introspection almost of your system, to understand what is going on — and how to make it better.
But he maintains the view — one that we’ve heard in earlier conversations with Zendesk — that all this automation isn’t leading to companies cutting back their customer service functions. Instead, he says they’re taking the opportunity to extend the support they offer:
I think no one says, ‘I have enough customer service.’ We are seeing this characteristic where our customers are paying down their service debt. We ourselves at Zendesk basically added 24-hour coverage, for more lines of business and more products.
You know, 74% of consumers in the recent Zendesk trends said they expected customer service to be 24/7, and 85% of CX leaders say one unresolved issue is enough to lose a customer. And so the first thing we did with these tools is really hugely increase our coverage and the amount of time we could spend with customers.
He argues that as AI automation lowers the cost of delivering customer service, a Jevons Paradox effect will come into play, increasing demand for customer service capabilities rather than reducing the amount spent on them. He goes on:
We introduced AI and the marginal cost to serve drops precipitously, or drops tremendously, as a result of the efficiencies you can get through automation and through generative search. Suddenly you reach a point where you're like, 'Okay, well, what would I do if it was free, or nearly free?' And the answer is, 'I would make sure that everyone could get rich contact and information and possibly a resolution, anywhere, at any time...
We see Jevons Paradox in service, where everyone is like, 'Well, now it is free to be available at any time for our customers and chat with them as long as they want to chat. I'm going to do more of that.' Because the only thing worse than talking to your customers is not talking to your customers.
My take
This latest is the most consequential so far of Zendesk's AI acquisitions, bringing Forethought's fully AI-native approach to how its agents function. Instead of simply speeding up what Zendesk has always done, Forethought's agents do it differently, performing discovery of how customer issues have been resolved in the past and directly implementing that knowledge and those processes into the creation of workflows, and adding observability and QA to provide oversight. This is in line with how a new breed of AI-native startups are approaching enterprise applications, helping Zendesk become competitive against the likes of newcomers like Sierra and others. Forethought's slightly longer history — note that the company was founded a few years before the advent of ChatGPT — gives it the added advantage of a substantive track record of understanding how to bring this new approach to CX teams.
It will take some time to blend Forethought's technology and capabilities into Zendesk's existing infrastructure. To that extent the vendor still has more of a challenge than those AI-native rivals, in that it has to support both approaches across its customer base and then help customers transition from one to the other. But this pattern of acquiring and then adopting newer technology approaches has been one that has served vendors well in previous transitions, such as from client-server to cloud, so it will be interesting to follow Zendesk's progress here.
It's also interesting to hear Zendesk talk about enterprises using these new technologies to increase what they're able to do on the CX front rather than simply reducing costs through employing fewer people to deliver the same level of service. It's a line we've heard from other CX vendors too. Perhaps there's an element here of, 'They would say that, wouldn't they?' as it helps them grow if their customers to do more with the tech rather than cutting back. But it's in contrast to all the stories we're hearing at the moment about AI leading to headcount reductions across the tech industry. Given that CX has been one of the earliest functions to adopt this latest AI technology, that may suggest that those who are pre-empting anticipated efficiency gains by cutting back capacity have got it wrong. (Or that they are choosing to obscure their real reasons for cutting staff).
On the other hand, the explanation may be more nuanced. Perhaps CX is a field where accurate, up-to-date knowledge has been so hard access in the past that it benefits from the increased Knowledge Velocity that agentic AI supports. Whereas in other areas, such as software development, perhaps it has been the collaboration burden that has held back productivity, and agentic AI's ability to slice through that overhead is freeing up capacity that is no longer required. These are questions to dig into urgently — we'll be returning to this topic again — so that enterprises can face these decisions with accurate insight into their consequences.