Organizations should treat AI agents that serve customers or act on company systems like employees gaining responsibility: define the job and permissions in writing, test competence before launch, supervise performance, and expand authority only when evidence supports it. That is the governance approach Sushil Kumar, Cyara’s CEO, outlined in a Unite.AI interview published October 6, 2026.
Who is Sushil Kumar?
Kumar became Cyara’s CEO in December 2025, according to the company’s appointment announcement; Cyara’s current leadership page also lists him in the role. The interview describes more than 25 years of leadership spanning AI, DevOps, cloud infrastructure, product strategy, and software testing. Before Cyara, he co-founded and led RelicX.ai, which was acquired by Harness, and held roles at Broadcom, CA Technologies, and Oracle.
![Claude AI for Beginners Bible: [5 in 1] The Ultimate Guide to Automate Your Work, Save Hours Every Week, and Use AI for Real-World Results](https://m.media-amazon.com/images/I/415+fSJacsL._SL160_.jpg)
As an Amazon Associate I earn from qualifying purchases.
Cyara says its platform tests, validates, and monitors customer journeys across voice, digital, messaging, and conversational AI. In its December 2025 announcement, the company reported customers in more than 135 countries, more than 350 million customer journeys each year, and over a million AI-generated responses validated. These are Cyara’s own reported figures, not independently verified measurements.
Quick wins for a faster PC:
Clear out junk files and repair common Windows errorsFree Scan →Scan for outdated or missing drivers - takes under a minuteDriver Scan →Why does an AI agent need a defined role?
Kumar’s starting point is that an AI agent can produce a plausible but incorrect answer without the kind of obvious error signal that conventional deterministic software may provide. A technically fluent reply is not proof that the agent understood the customer, used the right information, or acted appropriately. He argues that organizations should define the agent’s role before allowing it into production.
#1 Best Overall
The role description should state:
- Purpose: what customer or business need the agent is there to handle.
- Authoritative knowledge: which sources it should rely on when answering.
- Customer data: what information it may access and use.
- Permitted decisions: what it may decide without human approval.
- Responsibility boundary: where its job ends and escalation begins.
As Kumar puts it, “If a company cannot write that down in a paragraph, the agent is not ready for a role. It is ready for a demo.” This is his guidance in the interview, not a formal industry standard.
How should organizations set permissions and handoff boundaries?
A job description alone does not define authority. Kumar recommends specifying permissions separately, especially whether an agent may only read information or also change it. Access to view a customer record does not automatically justify editing it, issuing a refund, changing an account, or making a commitment on the company’s behalf.
Before deployment, organizations should make clear:
- Which systems and data the agent can read, and which it can write to or change.
- What promises, approvals, or other commitments it is allowed to make.
- Which situations require a human handoff, including cases outside its role or authority.
The more consequential an action or commitment, the stronger the evidence should be before the agent is permitted to make it autonomously. Kumar’s framework therefore distinguishes not just between automated and human work, but between different levels of impact and permission.
What does an HR-like operating model for agents look like?
Kumar compares agent governance to workforce management: an organization defines a role, looks for evidence that the person can do the job, supervises performance, and grants more responsibility over time. For agents, he describes autonomy as a progression rather than a one-time launch decision: “My read is that autonomy is not a deployment decision. It is a series of promotions.”
- Define the role and authority. Write down the goal, knowledge sources, data access, allowed decisions, and limits before production use.
- Collect pre-launch evidence. Evaluate the agent in conditions resembling actual use, rather than relying only on a successful demonstration.
- Assign an accountable business owner. A named person should own the agent’s purpose, performance, and permission changes.
- Monitor production behavior. Review what the agent actually does after launch, including actions and outcomes, not just whether it returns a valid response.
- Set promotion gates. Expand permissions only after the agent meets defined evidence requirements; the proof should become more demanding as autonomy and potential impact rise.
- Turn failures into tests. When a production failure occurs, add a test that checks for the same problem before a later release.
“A pilot can run on an organization’s conviction. Scale requires evidence,” Kumar says. The interview does not prescribe universal score thresholds or a single evaluation method; organizations must set gates appropriate to their agents’ roles and risks.
Rank #4
What should companies measure after launch?
Kumar says evaluation should target the business or customer outcome the agent is meant to achieve, not only whether its response is technically valid. A fluent answer can still fail the task, make the wrong commitment, or leave the customer worse off. Production review should therefore connect the agent’s actions to the intended result and to its written authority.
For voice systems, Kumar also emphasizes that speech-recognition errors can distort downstream evaluation: an agent may respond incorrectly because it misheard the customer’s question. In his view, assessing the agent without accounting for what it actually heard can confuse a recognition problem with a reasoning or response problem.
Best Value
What is Kumar’s advice for enterprises?
Start with a narrow, written job and explicit permissions; gather evidence under realistic conditions; keep a business owner accountable; and treat greater autonomy as something earned through monitored performance. In Kumar’s framing, an AI agent is not ready for broader responsibility merely because it can produce convincing answers. Its authority should track demonstrated competence and the consequences of its actions.
The perspectives and recommendations in this article are Kumar’s as presented in the Unite.AI interview published October 6, 2026.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
Free tools Windows power users keep installed
One-click scans. No signup required.
