Zendesk is transforming artificial intelligence into one of the main growth drivers of its business. By the end of 2025, the company projects to reach approximately US$200 million in annual recurring revenue (ARR) generated exclusively from AI products, a figure that already represents more than 10% of the total ARR of the customer service software company.
Global CEO Tom Eggemeier explains to NeoFeed how the company has been deliberately separating revenue generated by new AI products from that generated by traditional solutions, in an attempt to accurately measure the real economic impact of AI on the business.
“Some companies are taking an existing revenue base, launching a few features, and calling that AI revenue. That’s not what we’re doing,” says Eggemeier.
He emphasizes that the company's numbers reflect new products launched in the last two years, such as AI Agents, AI Copilot, and AI QA, platforms that go beyond basic automated assistants, taking on real operational tasks with machine learning.
Throughout 2025, Zendesk saw these products gain accelerated traction: more than 20,000 customers already use its AI solutions, and about two-thirds of the 3,000 largest customers in its base have adopted some AI functionality.
“We are seeing clients who already automate 70%, 80%, or even 90% of their interactions,” says the executive, highlighting that agent-driven agents—those who not only answer questions but also execute actions—are at the heart of the operational revolution.
One of the key elements of this strategy has been a series of acquisitions focused on expanding AI and automation capabilities. In recent years, Zendesk has acquired companies such as Tymeshift, HyperArc, and Local Measure to strengthen analytics, quality, and voice technology in its products.
On Friday, December 19th, the company announced the acquisition of the enterprise search platform Unleash, which incorporates generative intelligence and enterprise search with RAG (Retrieval Augmented Generation) to connect internal knowledge and accelerate employee support within tools such as Slack and Microsoft Teams.
The plan is to announce at least two more acquisitions in 2026 that will help build a portfolio focused on increasing the automatic resolution rate of customer service requests.
“The first phase of this revolution was simply about becoming aware and answering questions. Now our agents are able to execute actions, change tasks and procedures, and act on behalf of clients or employees,” says Eggemeier.
With the expansion of AI usage, Zendesk chose to revise its commercial policy. Billing ceased to be based on the volume of interactions and began to depend exclusively on the effective resolution of the customer's problem.
The contract-by-resolution model, which was launched globally and is already in operation in Brazil (currently Zendesk's third-largest market in the world), creates, however, a known side effect in the sector. By reducing friction for the customer, satisfaction increases, but the volume of interactions grows, also raising the cost of customer service—something the global CEO says he is not worried about.
“The volume of interactions has increased, and that’s a positive thing. What has changed is that our clients have started to focus much more on how to automate resolutions, and not just on handling more contacts. Now, our priority is to solve more problems more efficiently,” he says.
Over the next three years, the CEO predicts that AI's role in customer service will expand even further, with consumers no longer interacting directly with companies, delegating decisions and negotiations to personal agents trained on data such as consumption habits, purchase history, and preferences for time and price.
According to him, direct interaction between customers and companies should also decrease, as companies are able to use artificial intelligence to prevent problems.
"In many cases, the future of customer service will no longer be about the company reacting to a complaint, but about anticipating issues even before the customer realizes that something has gone wrong," predicts Eggemeier.
