Zendesk’s AI Summit Puts Enterprise Customer Service in Focus

The next major change in customer service may happen before a customer ever visits a company's website.

Instead of searching for a support page, opening a chat window or calling a contact center, a customer could simply ask an AI assistant to resolve a problem. The assistant might locate the appropriate business, establish a connection with its support system and help complete the request.

For businesses, that possibility raises questions extending well beyond the performance of a chatbot. Who controls the interaction? How does an AI assistant establish that it is contacting the correct company? What happens when a request requires access to an account, a financial transaction or a human decision?

These questions form part of the backdrop to Zendesk's AI Summit 2026, scheduled for November 12. The online event will bring together executives and technology leaders to examine the next generation of AI-powered customer and employee service.

Zendesk plans to preview new AI capabilities designed to move beyond providing information toward completing work. The broader significance is a potential change in how companies organize service operations, measure performance and interact with their customers.

From answering questions to resolving problems

For years, customer service automation has focused heavily on handling repetitive inquiries: checking an order status, explaining a return policy or helping a customer find the right information.

More capable AI systems introduce a different possibility. Rather than merely explaining how a customer can change a reservation, an AI agent might help make the change. Instead of directing an employee to the appropriate internal system, it might initiate an authorized workflow on the employee's behalf.

That distinction matters.

An AI system that answers a question can be evaluated largely on the relevance and accuracy of its response. One that takes action must also be evaluated on whether it completes the correct task, follows company policies and handles exceptions appropriately.

Consider a customer requesting a replacement for a defective product. Successful automation could require verifying the purchase, checking warranty eligibility, obtaining approval when necessary, arranging shipment and confirming the outcome.

An accurate explanation of the replacement policy is useful. Completing the replacement correctly is a different operational achievement.

Zendesk's summit agenda reflects this shift. A keynote focused on the autonomous workforce is expected to introduce AI capabilities intended to take action, manage more complex customer journeys and improve outcomes over time.

The company has not yet presented those upcoming capabilities in full. Executives evaluating the announcements will need to distinguish between product demonstrations, generally available functionality and features still under development.

A new entry point for customer service

One of the more consequential developments has already been announced.

In October, Zendesk introduced its Open Service Directory, an initiative intended to help personal AI assistants identify and connect with verified customer-support destinations.

The underlying problem is straightforward: an AI assistant may understand what a customer wants but still lack a reliable way to locate the company's authorized support system.

Searching the web for a support address or attempting to navigate a company's website does not necessarily establish that the assistant has reached the correct service provider.

Zendesk's proposed directory would allow participating businesses to approve official support listings that compatible AI assistants can discover and use.

A customer asking a personal assistant to contact a business could potentially be connected to that business's authorized support experience without having to navigate the usual entry points.

The directory is designed for participation across service platforms, rather than exclusively for businesses using Zendesk.

It also has an important limitation: discovery is not the same as authorization. The directory does not itself authenticate customers, execute support actions or store service conversations. Those responsibilities remain with the systems handling the interaction.

In an October 27 product update, Zendesk invited eligible customers to express interest in a closed early-access program. Additional controls for identifying assistants and managing their permissions remain under development.

Zendesk has said it plans to discuss the directory and related controls further at the summit.

For enterprise leaders, the implications extend beyond a new technical integration.

If customers increasingly delegate service requests to personal AI assistants, companies may need to make their customer-service capabilities accessible through interfaces they do not own.

That could change how businesses think about digital customer experience. A well-designed website and a capable contact center would remain important, but the company's ability to serve customers through external AI systems could become another consideration.

The challenge will be preserving accountability, security and service quality as those interactions cross organizational and technical boundaries.

The economics of resolution

The promise of AI-powered service is frequently described in terms of efficiency: fewer repetitive tasks, shorter waiting times and less manual work.

Those benefits are attractive, particularly for large organizations managing substantial volumes of customer and employee requests.

Yet measuring the return on AI investments requires care.

A company can automate a large share of incoming conversations without necessarily achieving the same improvement in completed customer requests.

An automated response that leads to a second contact, a reopened ticket or an escalation may reduce work at one point in the process while creating additional work elsewhere.

For finance and operations executives, a more useful measurement framework examines the entire resolution process.

That includes the cost of completing a request, the frequency of repeated contacts, the number of handoffs, the quality of the outcome and the effort required to correct errors.

Customer satisfaction remains important, but it should be considered alongside these operating measures rather than treated as a substitute for them.

The economics can become more complicated when AI systems interact with multiple applications. A task that appears simple to the customer may involve retrieving information, checking permissions, updating a record and communicating the result.

The financial benefit depends on whether the complete process becomes more efficient—not merely whether the first conversation becomes shorter.

For organizations considering substantial AI investments, this suggests a practical approach: select a defined category of requests, establish baseline costs and outcomes, and measure what changes when automation is introduced.

The most valuable results may come from identifying workflows in which AI consistently completes work, rather than maximizing the percentage of conversations handled without a human.

Human expertise is changing, not disappearing

Greater automation also raises questions about the future organization of support teams.

As AI systems handle more routine tasks, human specialists may spend a larger portion of their time dealing with exceptions, sensitive situations and problems requiring judgment.

This could alter staffing, training and management requirements.

An employee who previously handled a steady stream of straightforward requests may increasingly work on cases involving incomplete information, conflicting policies or customers whose circumstances do not fit a standard process.

The organization may need fewer manual steps for some workflows while requiring stronger expertise in escalation management, quality assurance and the supervision of automated systems.

Another consideration is what happens during a handoff.

A customer who has already explained a problem to an AI assistant should not necessarily have to repeat the entire interaction when a human becomes involved.

Preserving useful context—including what was requested, what actions were attempted and why the case was escalated—can help support teams deliver a more coherent experience.

These capabilities are operationally important because automation is most valuable when it works as part of the service organization rather than as an isolated layer in front of it.

Trust becomes an operational requirement

Allowing AI agents to take action creates responsibilities that are less prominent when software merely supplies information.

An incorrect answer can confuse a customer. An unauthorized change to an account, an inappropriate refund or an action based on mistaken identity can create more serious consequences.

Companies therefore need clear policies governing what an AI system may do independently, when additional approval is required and which decisions should remain with people.

They also need records sufficient to understand what occurred when an interaction goes wrong.

These requirements do not make autonomous service impractical. They are part of what makes it possible to deploy such systems responsibly at enterprise scale.

Zendesk's work on verified support discovery and assistant-specific controls illustrates how the industry's attention is expanding beyond conversational performance to the infrastructure around AI interactions.

For business leaders, the objective should be to enable automation while retaining appropriate control over customer information, company policies and the actions performed on their behalf.

What executives should watch at the summit

The November 12 program includes a conversation moderated by technology journalist Casey Newton, featuring Colin Fleming, Chief Marketing Officer, Business at OpenAI, and Elena Verna, Head of Growth at Lovable.

Zendesk executives and other technology and business leaders are also scheduled to participate, with product previews and technical discussions forming part of the program.

The event's value for enterprise decision-makers will depend less on the number of new features announced than on the operational questions those features help answer.

Can AI complete complex work reliably? How easily can businesses connect it to existing systems? What controls govern its actions? How does the organization measure financial and customer outcomes? And how can human teams remain effective as their responsibilities evolve?

These are questions that extend across the customer-service industry, regardless of which platform a company uses.

Zendesk's summit offers an opportunity to examine how one major service technology provider is approaching them, alongside perspectives from the broader AI ecosystem.

The larger transition is becoming clearer. Customer service is moving toward systems that do more than respond to requests. Increasingly, the ambition is to coordinate information, software and human expertise to complete the work behind those requests.

Whether that ambition produces meaningful business value will depend on execution.

For executives, the important question is no longer simply how much customer service AI can automate. It is whether that automation makes the overall service experience more reliable, more efficient and better for the people who depend on it.

---

Zendesk AI Summit 2026 takes place online November 12, beginning at 11 a.m. Pacific. Registration is free through the official event website.