ServiceNow executive: AI agents need data, workflows and clear limits

AI agents are gaining more freedom to make decisions and perform tasks directly in CRM systems. Daniel Wilks, who leads ServiceNow’s CRM sales in EMEA, says successful deployment requires more than a powerful AI model. After more than 20 years in the technology industry, he identifies three foundations: complete data, clearly defined workflows and limits on autonomy.

Wilks says an agent needs a complete view of the customer. If information is fragmented between CRM, ERP and other systems, the agent works from an incomplete picture. Companies must also define which systems do what, which rules apply and how employees work with the agent’s decisions.

For European companies, governance is also shaped by the EU AI Act. Most CRM agents that access customer data and act on it will fall under the regulation, and some uses may be classified as high-risk AI. The rules require human oversight, documentation and cybersecurity controls. High-risk AI rules enter into force in December 2027. GDPR and, in financial services, DORA may add further requirements.

Agents can handle bounded tasks such as creating and pricing a quote, resolving simple service cases or scheduling technicians. Wilks says people should remain involved when decisions involve exceptions, major financial consequences, irreversible outcomes or regulated approvals. He also warns that fragmented systems and undefined workflows can make broad projects fail, and recommends starting with one clearly bounded use case.