iTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
An AI agent can successfully complete a transaction without being authorized to make it. A dashboard showing normal operation does not establish that the action followed company policy—or that the company can later explain why it was allowed. Enterprises deploying agents in business applications need to define their authority, control consequential actions and preserve evidence. That is a governance responsibility, not a blanket conclusion about legal liability, which depends on the applicable law, contracts and sector.
What does it mean for a company to own the risk?
It means the organization must decide what business actions an agent is allowed to take, under what conditions, and who is accountable for setting those rules. A vendor may provide and secure the software, but it cannot make the company’s business-policy decisions on its behalf. As CIO contributor Richard Ewing put it, “The vendor can provide the software, but the enterprise still owns the business rules.”
In a September 21, 2026 opinion article, Ewing described an architecture review in which a customer-support agent issued an unapproved account credit while technical dashboards showed normal operation. This is Ewing’s reported example, not an independently verified case study. It illustrates the distinction that matters: the system may have operated as configured while the transaction still fell outside the organization’s intended authority.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Technical capability is not business authorization. An agent might have permission to write to a customer record or initiate a payment because of the account access it has inherited. That access alone does not answer Ewing’s useful question: “What decisions is the software allowed to make with that access?”
Four questions to ask about an agent action
1. Did the system operate?
Did the agent run, connect to the application and complete the requested operation? Logs and dashboards can help answer this technical question.
2. Can the action be reconstructed?
Can the organization recover the relevant request, data, context, system state and sequence of events? A record that merely says a credit was issued may not explain how the agent reached that result.
Rank #2
3. Was the action permitted?
Was the action inside an explicit policy boundary—for example, an approved refund limit or a rule requiring a person to review a proposed contract commitment? Evidence of successful execution is not evidence that the action was authorized.
Do these 3 things before closing this tab:
1Repair Windows errors before they cause bigger problems2Scan for outdated or missing drivers - takes under a minute3Clear out junk files and repair common Windows errors4. Who owns the consequence?
Which business owner is responsible for the rule, the control and the response if the action causes harm? The answer should be established inside the organization rather than left implicit in a vendor’s configuration.
Rank #3
How to set an agent’s authority before deployment
The following governance steps are practical recommendations, not a universal legal mandate. The right controls depend on the agent’s role and the consequences of its actions.
- Inventory agents that can change live systems. Include vendor-supplied and internally built agents that can edit business records, alter contracts or initiate financial transactions. Separate read-only access from write access.
- Document effective access. Record which user permissions an agent inherits, plus any service-account or system permissions it receives. Identify the actions available through each connected application.
- Name a business owner for the rules. Assign a leader who can define allowed actions, exceptions and limits, and who is responsible for reviewing those rules as the process or agent changes.
- Classify actions by consequence and reversibility. A meeting summary is not equivalent to an account credit, a contract commitment or a change to sensitive data. Consider financial, contractual and data impacts, as well as how difficult an action would be to reverse.
- Put policy checks before consequential changes. Require applicable policy, approval or financial-limit checks before a high-impact record is changed. For particularly consequential actions, use a separate check rather than relying only on the agent’s own decision.
- Keep evidence of both context and authorization. Preserve enough information to explain what the agent was asked to do, what information it used, which rule applied and why the action was allowed. Where feasible, retain this evidence independently of agent behavior that a vendor or administrator can later change.
- Review behavior after deployment. Monitor outcomes and investigate exceptions. Also determine whether vendor updates can change the agent’s behavior or effective authority, and whether existing records will remain interpretable afterward.
Match review and governance to the consequences
Requiring a person to approve every agent action may be unworkable at scale; approving too broadly can leave consequential actions without an effective boundary. Ewing argues for business controls rather than treating blanket manual approval as the answer. That is an opinion argument, not a controlled study of approval models. A practical approach is risk-tiered review: use lighter checks for low-impact, reversible work and stronger policy gates or human review for actions with meaningful financial, contractual or sensitive-data consequences.
Gartner’s May 26, 2026 release warns that applying uniform governance across AI agents can lead to failure. The implication is not that every agent needs a different rulebook, but that controls should reflect what each agent can do and what could follow from an error.
PC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11Outdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware matchWhat NIST can—and cannot—provide
NIST’s Artificial Intelligence Risk Management Framework (AI RMF 1.0, published in January 2023, offers a voluntary structure organized around four functions:
- Govern: establish responsibilities, policies and oversight.
- Map: understand the system’s context, intended use and risks.
- Measure: assess and track relevant risks.
- Manage: prioritize and address risks over time.
NIST describes the framework as “voluntary, rights-preserving, non-sector-specific, and use-case agnostic.” It can organize an organization’s work; it is not a certification, legal opinion or guarantee of safety or compliance.
Monitoring remains important, but the operational methods are not settled. NIST’s March 6, 2026 report, Challenges to the monitoring of deployed AI systems, says “Post-deployment monitoring is crucial” for validating real-world reliability, identifying unforeseen outputs and seeing unexpected consequences. The same report says practices, validated methods and shared terminology remain nascent and scattered. A monitoring dashboard can help surface problems; it does not by itself establish that an action was authorized.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Why adoption forecasts do not prove safe deployment
Gartner forecast on August 26, 2025, that 40% of enterprise applications would feature task-specific AI agents by the end of 2026, up from less than 5% at the time of the forecast. That is a forecast about applications, not the share of companies, and it is not an observed outcome or evidence that those deployments are governed effectively. Gartner’s separate May 2026 warning about uniform governance reinforces the need to tailor controls, not to assume that adoption alone demonstrates readiness.
Free tools Windows power users keep installed
One-click scans. No signup required.
Quick Recap
Questions leadership should be able to answer
- Which agents can change business records, contracts or financial transactions?
- What user permissions, service accounts and connected-system access does each agent have?
- Which named business leader owns the rules governing each agent’s actions?
- How do policy checks and review change as an action becomes more consequential or harder to reverse?
- Can the organization reconstruct the context and authorization for a past action?
- Could a vendor update change agent behavior or its authority boundary, and how will that change be reviewed?
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.
