Listen to this Post
Introduction: Security Is Entering a New and Unforgiving Era
The enterprise security landscape is changing faster than many organizations can redesign their infrastructure. For years, companies have tried to bring networking and security closer together through Secure Access Service Edge, or SASE. But the arrival of agentic AI is forcing that strategy to evolve again.
Artificial intelligence agents are no longer limited to generating text, summarizing documents, or answering questions. Increasingly, they can interact with applications, retrieve information, make decisions, execute workflows, and operate across cloud and corporate environments. In other words, AI systems are beginning to behave less like software features and more like a new category of digital worker.
That creates a difficult question: If an AI agent can access the same systems as an employee, who is responsible for controlling what that agent is allowed to do?
Cisco believes its SASE architecture can provide part of the answer. The company says it has been recognized as a Leader in the 2026 IDC MarketScape: Worldwide Secure Access Service Edge Vendor Assessment, where 12 vendors were evaluated.
The recognition highlights several areas Cisco considers central to its strategy, including architectural breadth, flexibility, artificial intelligence capabilities, identity integration, threat visibility, and research and development.
But the award itself is only one part of the story.
The more important development is
Cisco’s SASE Strategy Gets a New Test
Cisco says the IDC MarketScape assessment recognized its ability to bring networking, security, identity, endpoint, browser, firewall, switching, and routing technologies into a broader architecture.
The underlying idea is straightforward: instead of forcing security teams to manage a collection of disconnected products, organizations should be able to enforce security policies across different parts of their infrastructure through a more unified architecture.
That sounds simple, but the practical problem is enormous.
Modern businesses rarely operate from a single data center. Employees connect from offices, homes, mobile devices, cloud environments, branch locations, SaaS platforms, and third-party services. Applications communicate across multiple networks, while data constantly moves between systems.
Now add AI agents to that environment.
The number of identities, connections, transactions, and automated decisions can increase dramatically.
Why Agentic AI Changes the SASE Equation
Traditional security models were largely designed around identifiable users, devices, applications, and services.
An employee signs in.
A device establishes a connection.
An application requests access.
A security system evaluates the request.
Agentic AI introduces a much more complicated chain of activity.
An agent might receive a task from a human, communicate with another model, call an API, retrieve information from a database, access a cloud service, execute an operation, and then trigger another automated process.
The challenge is not simply determining whether the agent is allowed to connect.
The real challenge is determining what the agent is allowed to do after it connects.
This distinction could become one of the defining security problems of the AI era.
AI Agents Are Becoming a New Digital Workforce
Cisco describes AI agents as a new workforce because they can act, make decisions, and access systems at machine speed.
That comparison is useful, but it also exposes a major weakness in existing enterprise security architecture.
A human employee usually has an identifiable account, manager, department, device, authentication method, and set of permissions.
An AI agent may instead rely on API keys, service accounts, tokens, application credentials, or other machine identities.
Those mechanisms can prove that something has access.
They do not always provide enough context to explain why the action is happening, who authorized it, what the agent is supposed to accomplish, or whether its behavior has drifted beyond its intended purpose.
That is where identity becomes much more important.
Speed Without Trust Can Become Automated Risk
Cisco cites research suggesting agentic workflows can generate dramatically more traffic per task than human-driven activity, while only a minority of organizations report having sufficient control and monitoring over what their agents actually do.
The broader lesson is more important than any individual statistic.
Automation removes the natural friction that exists in human workflows.
A person may pause before opening a sensitive file.
An automated agent may process thousands of records in seconds.
A human may notice that an unusual request looks suspicious.
An agent can potentially execute the request exactly as instructed unless another control stops it.
That means security cannot simply monitor whether an AI system is connected.
Security has to understand what the AI system is doing.
Identity Becomes the Center of AI Security
Cisco’s proposed approach places identity at the center of its SASE strategy.
The goal is to give every agent a verifiable identity connected to its human owner, host environment, and security posture.
This is an important evolution from conventional machine credentials.
Instead of treating an API key as the identity of an automated process, organizations could theoretically build a richer security context around the agent.
Who owns it?
What task is it performing?
Where is it running?
What data should it access?
How long should that access remain valid?
What actions is it authorized to perform?
What happens if its behavior suddenly changes?
These questions move security away from simple network access and toward continuous authorization.
Short-Lived Credentials Could Reduce AI Exposure
One of
This follows a broader security principle: the less time a credential remains valid, the smaller the window of opportunity for misuse.
A long-lived secret can become extremely dangerous if it leaks.
A temporary credential with tightly defined permissions has a much smaller blast radius.
Cisco describes this model in terms of just-in-time and just-enough access.
The concept is particularly relevant for AI because agents may need powerful capabilities for very short periods.
An agent that needs to update a database for one task should not necessarily possess permanent access to that database.
An AI assistant that needs to retrieve a document should not automatically receive broad access to the entire corporate file system.
Least privilege becomes much more meaningful when permissions can change dynamically with the task.
From Destination-Based Security to Action-Based Security
Traditional network security frequently asks a relatively simple question:
Can this identity access that destination?
AI security increasingly needs another question:
Can this identity perform this specific action?
Those are very different security problems.
An agent might legitimately access a CRM system but should not be allowed to export an entire customer database.
It might access source code but should not modify production infrastructure.
It might read financial information but should not automatically transfer money.
This is why action-level authorization could become one of the most important areas of enterprise AI security.
Cisco’s Agent Gateway Concept
Cisco says its SASE-native Agent Gateway is designed to broker credentials and authorize actions rather than merely authorize destinations.
The broader architectural concept is compelling.
Instead of allowing agents to directly collect and store powerful credentials, a security layer can potentially stand between the agent and sensitive services.
The gateway becomes a policy enforcement point.
It can determine whether an action is permitted, issue temporary credentials where appropriate, collect behavioral signals, and feed those signals back into security policy.
That creates a feedback loop:
Identity → Authorization → Action → Monitoring → Risk evaluation → Policy adjustment
This is much closer to the security model enterprises will likely need as autonomous systems become more capable.
Cisco ThousandEyes and Visibility Across the Digital Experience
Security is only half of the challenge.
Organizations also need to know when AI-driven workloads are slowing down or failing.
Cisco points to ThousandEyes and AI Canvas as technologies intended to provide broader visibility into digital experiences and infrastructure performance.
That matters because enterprise AI systems increasingly depend on long chains of services.
An AI agent may rely on a model provider, API gateway, DNS infrastructure, cloud service, identity platform, database, internal application, and network connection.
If one component becomes slow, the agent may appear to be malfunctioning even though the model itself is operating normally.
Observability therefore becomes part of AI reliability.
Cisco cites a Forrester Total Economic Impact study that reported faster mean time to identify issues and substantial reductions in employee hours spent per incident by the third year.
The larger point is that visibility can translate directly into operational efficiency.
Resilience Must Extend Beyond the Network
Cisco is also connecting SASE with resilience and future cryptographic requirements.
Quantum computing remains a longer-term concern, but organizations are already discussing post-quantum cryptography because replacing cryptographic infrastructure across a global enterprise is not something that can realistically happen overnight.
The strategy is therefore shifting from:
Protect
to:
Build infrastructure that can adapt to
Cisco describes its edge and WAN foundation as incorporating quantum-safe protection while also pointing to technologies designed to reduce exposure before traditional patch cycles.
Whether every organization needs the same level of preparation today is debatable.
But the strategic direction is clear: security architecture is increasingly expected to survive changes in technology rather than simply respond to today’s threats.
Deep Analysis: What Cisco’s SASE Strategy Really Means
SASE Is Becoming More Than Network Security
SASE was originally associated with the convergence of networking and security functions delivered through cloud-oriented architecture.
That model is now expanding.
Modern SASE increasingly touches identity, endpoint security, browser security, application access, threat intelligence, observability, data protection, and AI governance.
Cisco’s strategy reflects that expansion.
The company is effectively arguing that networking and security cannot be separated when workloads themselves are becoming autonomous.
The Security Boundary Is Disappearing
The traditional corporate perimeter has already weakened dramatically.
Employees work remotely.
Applications live in multiple clouds.
Data crosses organizational boundaries.
Third-party SaaS platforms process sensitive information.
AI agents introduce another layer of abstraction.
The result is an environment where the question “Is this inside the network?” is becoming almost meaningless.
The better question is:
Should this specific identity perform this specific operation under these specific conditions?
That is the foundation of zero-trust security.
Commands for Basic SASE and Network Visibility
Security teams can use ordinary network tools to establish a baseline before deploying sophisticated policy controls.
For example, DNS resolution can be checked with:
nslookup example.com
Or, on Linux systems:
dig example.com
Basic connectivity can be tested with:
ping example.com
A route can be inspected using:
traceroute example.com
On Windows:
tracert example.com
HTTP behavior can be examined with:
curl -I https://example.com
For TLS certificate inspection:
openssl s_client -connect example.com:443 -servername example.com
These commands do not replace SASE platforms, identity controls, or continuous monitoring.
They provide something equally important: a baseline.
Before an organization can identify unusual behavior, it needs to understand normal behavior.
Logging Is Critical for AI Agents
AI agents should not be treated as invisible background processes.
Security teams need logs showing authentication events, authorization decisions, API calls, resource access, unusual destinations, privilege changes, and failed operations.
A useful investigation should ideally answer:
Who initiated the task?
Which agent executed it?
What identity did the agent use?
What credentials were issued?
Which resources were accessed?
What actions were performed?
What policy allowed those actions?
What changed during execution?
Without this information, an organization may know that something happened without knowing why it happened.
Least Privilege Becomes Dynamic
Traditional least privilege often means giving a user or application a carefully limited set of permissions.
AI agents require something more dynamic.
A permission should potentially exist only while it is needed.
A credential should expire when the task ends.
A high-risk action should trigger additional verification.
A sudden change in behavior should reduce or terminate access.
This is where identity-aware SASE architectures could become particularly valuable.
The Human Owner Still Matters
One of the most important ideas in
That creates accountability.
An enterprise should not have thousands of anonymous autonomous processes operating with unclear ownership.
If an agent can make a consequential decision, security teams should be able to determine who deployed it, what application controls it, what task it was designed for, and who remains responsible for its operation.
AI autonomy does not eliminate human accountability.
It makes accountability more important.
SASE Could Become the Control Plane for AI
If AI agents continue moving between applications, networks, clouds, and data sources, SASE platforms could increasingly become a central enforcement layer.
The platform would not necessarily need to understand every detail of the AI model itself.
Instead, it could control the environment around the model.
Identity.
Network access.
Credentials.
Application access.
Threat detection.
Device posture.
Data movement.
Behavior.
Policy.
That is a powerful position in the enterprise security stack.
But Consolidation Has Its Own Risks
Cisco’s unified architecture also raises an important question.
Does consolidating more security capabilities into one ecosystem reduce complexity, or does it create greater dependency on a single vendor?
The answer will depend on implementation.
A unified platform can reduce administrative overhead, simplify policy management, and improve visibility.
But organizations must still consider interoperability, portability, resilience, pricing, migration costs, and the consequences of a major platform outage.
Consolidation is valuable only when it does not become another form of lock-in.
Microservices Could Make Convergence More Flexible
Cisco says it is moving toward deeper convergence through microservices and micro-applications.
That direction is significant because modern security platforms cannot remain rigid monolithic systems.
Organizations want individual capabilities to evolve quickly without rebuilding the entire architecture.
Microservice-based designs can potentially make it easier to introduce new security functions, AI capabilities, analytics, and policy controls.
The challenge is ensuring that greater modularity does not create another layer of operational complexity.
Cisco Cloud Control and Unified Management
Cisco also points toward Cloud Control as a way of bringing SD-WAN, SSE, and firewall management into a more unified operating experience.
This could be particularly useful for large organizations where networking and security teams historically operated separate systems.
A shared management plane can improve visibility and reduce duplicated policy work.
But organizational processes matter just as much as technology.
A unified dashboard cannot automatically solve disagreements between networking, security, cloud, and identity teams.
Technology can remove friction.
It cannot eliminate organizational responsibility.
What Undercode Say:
AI Is Turning Identity Into the New Security Perimeter
The most interesting part of
It is the recognition that AI agents are creating a new identity problem.
Machines Are Becoming Decision Makers
For decades, machines mostly followed instructions created by humans.
Agentic AI changes that model.
An agent can interpret an objective, determine a sequence of actions, and interact with multiple systems.
That makes machine identity much more important.
API Keys Were Never Designed for Autonomous Employees
API keys are useful because they are simple.
They are also dangerous when they become permanent credentials for increasingly autonomous systems.
If an agent has a powerful key, compromising the agent could mean compromising everything accessible through that key.
Temporary Access Is the Better Direction
Short-lived credentials can significantly reduce the exposure created by stolen secrets.
They do not solve every problem.
But they make it harder for an attacker to turn one compromised credential into a permanent foothold.
Authorization Must Follow the Task
The security system should understand not only who is acting but why the action is occurring.
An AI agent performing a legitimate task should receive the minimum permissions necessary to complete it.
Behavioral Monitoring Will Become Essential
Identity alone is not enough.
An authenticated agent can still behave maliciously or unexpectedly.
Security systems must therefore evaluate behavior continuously.
AI Creates New Attack Paths
Attackers will eventually target agents because agents can have access to valuable systems.
A compromised AI workflow could potentially become a stepping stone into databases, source repositories, cloud platforms, and internal applications.
SASE Could Become an AI Security Layer
This is where
SASE already sits between users, applications, networks, and security controls.
Adding AI identity and agent authorization could turn it into an important enforcement layer for autonomous workloads.
Visibility Will Matter as Much as Prevention
Organizations cannot secure what they cannot see.
Agent activity must be observable across clouds, networks, applications, and endpoints.
AI Traffic Will Be Different
Human traffic has predictable patterns.
Agents can generate bursts of automated activity that are difficult to interpret using traditional assumptions.
Security analytics will have to adapt.
Zero Trust Becomes More Important
The zero-trust principle of never automatically trusting a connection becomes especially relevant when the connection originates from an autonomous system.
Every request needs context.
The Agent Should Not Become the Administrator
One of the biggest risks is excessive privilege.
An agent created to summarize reports should never casually receive administrator-level access to unrelated infrastructure.
Security Policies Need Granularity
Future policies may need to specify not simply “allow access to this application.”
They may need to specify:
Allow this agent to perform this action against this resource for this period under these conditions.
Human Accountability Must Survive Automation
Companies should know who owns every important autonomous workflow.
Otherwise, responsibility becomes impossible to establish after an incident.
AI Governance and Cybersecurity Are Converging
AI governance is often discussed as an ethical or regulatory issue.
In practice, it is also becoming a cybersecurity issue.
The question of what an agent is allowed to do is fundamentally a security question.
Network Security Alone Is Not Enough
An agent can operate through legitimate channels.
That means firewalls alone cannot determine whether its behavior is appropriate.
Identity Security Alone Is Not Enough
A valid identity can still perform a dangerous action.
Continuous behavioral analysis is therefore necessary.
Observability Creates Context
Tools such as network and application monitoring can help security teams understand where an automated workflow is failing or behaving unusually.
Performance Is Part of Security
An unreliable security layer can cause teams to bypass controls.
Performance therefore matters.
Security that constantly disrupts legitimate business operations will eventually face pressure to be weakened.
Complexity Is a Genuine Security Threat
Every additional console, policy engine, connector, and credential system creates another opportunity for configuration mistakes.
Reducing unnecessary complexity can improve security.
Consolidation Needs Discipline
However, putting everything into one platform is not automatically safer.
Organizations should maintain strong architecture, redundancy, and exit strategies.
Quantum Security Is a Long-Term Engineering Problem
Post-quantum migration should not be treated as a marketing checkbox.
Cryptographic systems are deeply embedded in enterprise infrastructure.
Preparation takes time.
Cisco’s Broader Strategy Is Clear
Cisco appears to be positioning itself as more than a networking company.
It wants to become an integrated security, identity, observability, and AI infrastructure provider.
The Competitive Battlefield Is Changing
SASE vendors are increasingly competing on more than secure connectivity.
They are competing on identity, AI security, endpoint protection, analytics, cloud management, and automation.
AI Could Accelerate SASE Adoption
Organizations that previously viewed SASE as an architectural improvement may now see it as a requirement for controlling autonomous workloads.
Agent Security Could Become a Major Market
The number of enterprise AI agents is likely to grow rapidly.
That means agent identity, authentication, authorization, monitoring, and governance could become major cybersecurity categories.
The Winner Will Need More Than a Good Firewall
Future enterprise security platforms will need to understand people, devices, applications, workloads, and autonomous agents.
The firewall remains important.
But it is no longer enough.
Trust Must Become Continuous
Security cannot simply establish trust at login.
Trust should evolve throughout the entire session and workflow.
Every Action Creates a Signal
An
That data can feed risk engines and policy decisions.
The AI Era Rewards Context
A simple allow-or-deny model is increasingly inadequate.
Security needs context.
Who is acting?
Why?
From where?
Using what identity?
With what permissions?
Against which resource?
SASE Could Become the Bridge
The architecture sits at an intersection between connectivity and security.
That makes it a logical place to enforce these decisions.
Cisco Still Has to Prove the Vision at Scale
Architecture diagrams are easy.
Running them across thousands of employees, applications, devices, agents, and cloud environments is much harder.
Customer deployment results will ultimately matter more than awards.
The Real Test Is Operational Simplicity
If Cisco can genuinely reduce the number of tools, policies, credentials, and workflows security teams must manage, its strategy becomes significantly more compelling.
AI Security Cannot Be Bolted On Later
Organizations deploying agents today should think about identity and authorization before those agents become deeply integrated into business processes.
Retrofitting security later will be considerably harder.
The Enterprise Security Model Is Being Rewritten
The traditional model was built around people using computers.
The emerging model includes machines acting on behalf of people.
That is a fundamental architectural change.
Cisco’s Recognition Is a Signal
The IDC MarketScape recognition should be viewed less as an endpoint and more as a signal of where enterprise security is heading.
The Next Security Perimeter May Be an Agent
The most important security identity in
It may belong to an autonomous software process operating on that employee’s behalf.
Trust Will Determine
AI can dramatically increase productivity.
But productivity without control can become an operational and security liability.
The Core Principle Is Simple
Let AI move fast—but never let it move without identity, authorization, visibility, and accountability.
That may ultimately be the most important promise behind SASE in the AI era.
✅ Cisco Was Named a Leader in the 2026 IDC MarketScape Assessment
The supplied article states that Cisco was recognized as a Leader in the 2026 IDC MarketScape: Worldwide Secure Access Service Edge Vendor Assessment.
It also states that the evaluation covered 12 SASE vendors and highlighted Cisco’s architecture, AI capabilities, identity integration, threat visibility, and R&D.
The recognition is therefore presented as a third-party industry assessment rather than simply an internal Cisco award.
✅ Cisco Is Expanding SASE Toward AI and Agent Security
Cisco’s announcement explicitly connects its SASE strategy with the rise of agentic AI, identity, least privilege, observability, and automated workflows.
The company is positioning SASE as an architecture capable of governing both human and machine-driven access.
This is consistent with the broader evolution of zero-trust and identity-centric security.
✅ Short-Lived Credentials Align With Security Best Practices
Using temporary credentials and just-in-time access is consistent with least-privilege security principles.
Reducing credential lifetime can limit the damage caused by stolen secrets.
However, short-lived credentials alone cannot prevent an authorized but malicious or compromised agent from performing harmful actions.
⚠️ AI Traffic Statistics Require Context
The article cites figures concerning substantially higher traffic generated by agentic workflows and the percentage of organizations able to control agent behavior.
Those figures should be interpreted according to the methodology, sample size, definition of an “agentic workflow,” and environment used by the underlying research.
They are useful indicators, but they should not automatically be treated as universal measurements across every enterprise.
⚠️ Quantum-Safe Readiness Is a Long-Term Transition
Post-quantum security is becoming increasingly important, but the urgency differs depending on an organization’s systems, data lifetime, cryptographic dependencies, and threat model.
A company should therefore treat quantum readiness as an architectural migration program rather than a single product feature.
✅ Cisco’s Strategy Extends Beyond Traditional Networking
The supplied announcement describes
This represents a significant expansion of the role Cisco wants its infrastructure to play inside modern enterprises.
Prediction
(+1) Agentic AI Will Accelerate the Adoption of Identity-Centric SASE
As enterprises deploy more autonomous AI systems, organizations will increasingly discover that traditional API keys, static service accounts, and broad permissions are difficult to govern safely.
This should increase demand for security platforms capable of assigning identities to agents, issuing temporary credentials, enforcing least privilege, and monitoring actions continuously.
(+1) Action-Level Authorization Will Become a Major Enterprise Security Requirement
The industry is likely to move from simply controlling where an AI agent can connect toward controlling what it can actually do.
This distinction could become one of the defining features of enterprise AI security platforms.
(+1) SASE Vendors Will Compete Directly on AI Security
Cisco is unlikely to be alone.
Competitors will increasingly integrate agent identity, AI-aware policy enforcement, application security, endpoint controls, observability, and automated threat detection into SASE offerings.
The SASE market could therefore evolve into a much broader AI infrastructure security market.
(+1) Temporary Credentials Will Become Standard for High-Privilege Agents
Permanent secrets are poorly suited to autonomous systems with rapidly changing tasks.
Short-lived credentials, dynamic authorization, and just-in-time access are likely to become increasingly common for sensitive AI workflows.
(-1) Poorly Governed AI Agents Could Become a Major Attack Surface
If companies deploy autonomous agents without strong identity and authorization controls, attackers may eventually target those agents as pathways into corporate systems.
A compromised agent could potentially provide access to data and services far beyond the original AI application.
(+1) Cisco’s Biggest Opportunity May Be Architectural Simplification
If Cisco can genuinely unify networking, security, identity, observability, and AI governance without creating another layer of complexity, its strategy could become highly attractive to large enterprises.
The real competitive advantage will not be the number of features.
It will be the ability to make a complicated AI-driven environment easier to secure.
Final Outlook: The AI Era Will Reward Trusted Automation
The most important lesson from Cisco’s latest SASE positioning is that enterprise security is moving toward a world where machines will need the same level of identity, accountability, and policy enforcement traditionally associated with human users.
AI agents will not simply consume information.
They will increasingly act.
They will access applications, invoke APIs, manipulate data, trigger workflows, and potentially make decisions with real business consequences.
That makes trust the central problem.
SASE can provide connectivity and security controls, but the next generation of SASE will need to understand identity, intent, behavior, privilege, and risk at a much deeper level.
Cisco’s recognition by IDC is therefore interesting not only because of the company’s position in the SASE market, but because it reflects a much larger shift taking place across enterprise technology.
The network is becoming software-defined.
Security is becoming identity-centric.
Applications are becoming distributed.
And AI agents are becoming autonomous.
In that environment, the organizations that succeed will not necessarily be those that deploy the most AI.
They will be the ones that can give AI enough freedom to create value while maintaining enough control to keep that freedom from becoming a security disaster.
That is the real challenge for Cisco, its SASE competitors, and every enterprise entering the agentic AI era.
🕵️📝Let’s dive deep and fact‑check.
🎓 Live Courses & Certifications:
Join Undercode Academy for Verified Certifications
🚀 Request a Custom Project:
Secure, high-velocity infrastructure and disruptive technological engineering. Contact our engineering team for high-tier development and proprietary systems:
[email protected]
💎 Smart Architecture | 🛡️ Secure by Design | ⭐ Trusted by Thousands
References:
Reported By: blogs.cisco.com
Extra Source Hub (Possible Sources for article):
https://stackoverflow.com
Wikipedia
OpenAi & Undercode AI
Image Source:
Unsplash
Undercode AI DI v2
🔐JOIN OUR CYBER WORLD [ CVE News • HackMonitor • UndercodeNews ]
📢 Follow UndercodeNews & Stay Tuned:
𝕏 formerly Twitter 🐦 | @ Threads | 🔗 Linkedin | 🦋BlueSky | 🐘Mastodon | 📺Youtube
