Gartner recently published The Future of NetOps Is Agentic, predicting a major change in how network and network security systems will be operated.
Today, most administrative activity starts in a graphical interface. Gartner expects that by 2030 AI agents and traditional automation will initiate the majority of network operations, while the share originating from GUI dashboards will fall substantially.
Whether the eventual number is 30%, 40% or 50% is probably less interesting than the direction.
We agree with the direction.
AI agents will perform a growing part of investigation and routine operational work. Natural language will replace a lot of navigation. Humans will increasingly describe intent instead of finding the correct page, tab and button.
But there is an important second question:
If AI becomes a primary way of interacting with operational systems, what is the GUI still for?
We don't think the answer is "nothing."
A GUI Does Several Different Jobs
It is easy to talk about "the GUI" as if it were a single thing.
It isn't.
A network management interface is used for very different purposes:
- navigating to information,
- understanding current state,
- investigating historical data,
- configuring systems,
- entering structured data,
- validating changes,
- and executing actions.
AI affects each of these differently.
Navigation is particularly vulnerable. If an administrator wants to know why a server is slow, there is little value in requiring them to remember which section contains the relevant data and then navigate through several screens.
They should be able to ask.
An AI system can identify the relevant nodes, metrics, alerts, events and configuration changes and bring the useful information together.
That is clearly better.
But presenting the answer is a different problem.
A Dashboard Is Not a Report
The monitoring industry has stretched the word dashboard until it means almost any page containing charts.
A screen with dozens of historical charts, tables and filters that scrolls through five browser heights is often called a dashboard.
We would call that an interactive report.
A dashboard has a different purpose:
A dashboard should summarize the current state well enough that a person can understand it at a glance.
Think about a car.
You would probably be happy to ask the car:
"Why is energy consumption unusually high?"
or:
"Find a charging station on my route."
You probably would not want to replace the speedometer, warning indicators and remaining range with a chat window.
Current state is exceptionally well suited to visual representation.
Humans process position, shape, relative magnitude and color very efficiently. After using the same dashboard repeatedly, we also learn where information is located. We no longer consciously search for it.
That spatial consistency is valuable.
A conversational interface is sequential. A good operational dashboard is perceived largely in parallel.
AI doesn't change that aspect of human perception.
Reports Will Change Much More
Historical analysis is different.
Traditional reports often exist because somebody had to decide in advance which data might eventually be useful.
So we build pages containing CPU charts, latency charts, traffic charts, events, averages, percentiles and tables — and then expect a human to search through them for something interesting.
AI is much better suited to this problem.
It can search much more data than a person reasonably can. It can compare time ranges, examine events around an incident, look for deviations from normal behavior and identify relationships that nobody thought to put next to each other on a report.
Instead of asking a human to inspect fifty charts, AI might conclude:
Response time started increasing shortly after storage latency changed, while CPU and memory remained normal.
That is useful.
But we would still want to see the relevant chart.
AI can identify the evidence. Visualization lets a human understand and verify it.
This suggests that the future of reporting may contain fewer permanently designed reports and more dynamically assembled evidence around a particular question.
Natural Language Will Become Navigation
This is probably one of the largest changes.
Today's software frequently exposes its internal organization directly to the user.
To perform a task, you must know where the designer placed it.
AI allows us to reverse this relationship.
The user expresses intent:
"Show me switches with increasing packet errors."
"Why did this service become unavailable yesterday?"
"Find Windows servers approaching end of support."
The system figures out where the information lives.
Traditional navigation will remain useful, but it no longer needs to be the only way into the product.
In many situations, it may not even be the primary one.
Configuration Is Different
Configuration is another area where claims about the disappearance of GUI can go too far.
AI can be excellent at translating intent into configuration.
For example:
"Alert me when CPU stays above 85% for ten minutes on production servers, but don't notify during maintenance."
An AI system can turn that sentence into a precise configuration.
But once it does, a structured interface becomes valuable again.
The administrator should be able to see:
- threshold: 85%,
- duration: 10 minutes,
- scope: production servers,
- maintenance behavior,
- notification action.
This allows the resulting state to be inspected, validated and changed precisely.
The likely future is therefore not conversation replacing forms.
It is:
natural language expresses intent → AI prepares the configuration → structured UI exposes the resulting state → deterministic software validates and applies it
For consequential changes, showing exactly what will happen is often better than another paragraph of generated text.
Automation Still Has a Major Role
One of the strongest points in Gartner's analysis is the distinction between AI agents and traditional automation.
Predictable, repeatable operations are usually better implemented as deterministic software.
If we know exactly how to perform a task reliably, there is little reason to ask a probabilistic model to rediscover the procedure every time.
AI becomes valuable where the problem itself is uncertain:
- investigating incidents,
- finding correlations,
- choosing what to examine,
- interpreting unfamiliar situations,
- or deciding between possible next steps.
Once the decision is made, ordinary software can perform the operation.
The future is not AI replacing automation.
It is AI deciding when and how to use automation.
The Interface Is Becoming a Combination of Tools
We therefore expect network operations interfaces to become less GUI-centric without becoming GUI-less.
Different tools will dominate different jobs:
AI for intent, investigation, correlation and discovery.
Automation for repeatable deterministic execution.
Dashboards for understanding current operational state.
Charts and visualizations for evidence and pattern recognition.
Structured interfaces for configuration, validation and precise control.
The major change is not that one of these replaces everything else.
It is that users will increasingly move between them without having to think about where information is stored or which subsystem provides it.
Where We Think NetCrunch Is Going
This distinction is important to how we think about the future of NetCrunch.
We don't see much value in adding AI simply because every software product is expected to contain an AI button.
AI becomes interesting when it understands the monitoring environment deeply enough to help investigate problems, find useful information, build the right view of the data and turn human intent into reliable operations.
At the same time, monitoring itself must remain deterministic, efficient and dependable.
And when AI reaches a conclusion, we still want the operator to be able to see the state, inspect the evidence and understand what the system intends to do.
So we agree that the traditional GUI will become less dominant.
But we don't think it is disappearing.
Its job is changing.
