As AI agents take on tasks inside enterprise workflows, monday.com security lead Ryan Nigro argues that controlling their permissions is not simply a compliance constraint. Carefully limited access, he said, can give organizations the confidence to deploy agents more quickly while reducing the risk that automated systems inherit excessive privileges.
Nigro, security trust and enablement team lead at monday.com Ltd., discussed the approach at the company’s Elevate event on Oct. 29, 2026. In a conversation with Rebecca Knight, he described permission management as a practical foundation for governing AI agents as monday.com develops its work platform with native agents.
His central recommendation is to apply least privilege: give each user or tool only the access necessary to perform its assigned work. Rather than enabling every feature and granting broad access across an organization at once, Nigro said, teams should introduce AI incrementally and widen permissions as they establish that a use case is working as intended.
Agents inherit the risks of existing access
The concern is not unique to AI. Agents generally perform work on behalf of people, Nigro said, so weaknesses in how organizations manage human access can become weaknesses in automated workflows as well. An employee who moves from finance to legal, for example, may retain permissions from the previous role while gaining new ones if access is not reviewed and adjusted. An agent operating with those entitlements could act within an unnecessarily broad set of permissions.
That makes lifecycle management an important part of agent governance. Organizations need to ensure that access remains appropriate as employees change responsibilities, and that the tools acting for them do not receive more authority than their tasks require. Nigro’s advice is to avoid treating broad access as the default and instead scope permissions to each role and use case.
He pointed to an IBM Corp. study estimating that 80% to 90% of breaches stem from human error associated with over-permissioning. The figure underscores a longstanding identity and access management challenge that becomes more consequential when software can use granted privileges to perform actions at machine speed.
In practical terms, access boundaries should reflect the separation of responsibilities inside a company. Nigro used the example of a sales representative who should not have the ability to change a production environment. Applying that principle to AI agents means assessing what data, systems and actions are necessary for a particular workflow, rather than granting expansive access for convenience.
Guardrails as an enabler of deployment
Organizations often worry that additional governance will slow AI adoption. Nigro presented least privilege differently: controls are there to support confident use, not to force teams to stop. He compared the role of security safeguards to brakes on a car, saying they allow a driver to move faster with confidence rather than requiring constant slowdowns.
That framing reflects a broader operational challenge for enterprises adopting agents. The technology’s usefulness depends in part on its ability to interact with business tools and information, but those connections also make permission design consequential. A staged rollout can help teams limit exposure while they establish which functions an agent needs and whether those permissions remain appropriate.
For security and IT teams, the approach places familiar identity controls directly into AI planning. Access decisions should be made before agents are broadly enabled, then revisited as workflows and employee responsibilities change. It also suggests that organizations should avoid equating successful deployment with maximum access: an agent’s permissions should be tied to its defined purpose.
Monday.com has positioned its platform as an AI work platform with native agents. Nigro’s comments focused on the security principles that should accompany that direction, rather than detailing particular technical controls or product capabilities. The interview took place during an event segment sponsored by monday.com.
As enterprise use of agents expands, least privilege offers a way to connect adoption with established access-management discipline. The operational goal, in Nigro’s account, is not to eliminate risk by preventing agents from acting, but to make sure their authority is narrow enough that organizations can use them with greater confidence.
