Itential

09/02/2026

Most teams don’t struggle to build an agent. They struggle deciding how much access to give it, what it should be allowed to change, and where humans still need to step in.

These patterns keep showing up when agents move toward production successfully:

• Smaller, specialized agents with scoped access are more reliable than one agent with every tool
• Governance has to cover who can build, what agents can access, and where humans stay in the loop
• Trust can be earned gradually, moving from read-only to write as the right controls prove themselves
• Reasoning has a cost, so knowing when to turn a proven agent into deterministic automation matters

In this session, Holly Holcomb and Karan Munalingal break down what teams need to move agents beyond the POC and into real infrastructure operations, plus demo what those patterns look like in FlowAI.

📺 Watch on-demand: https://bit.ly/4i8oUMO

09/01/2026

Scripts aren’t the problem. We just kept asking them to do a job they weren’t built for.

For years, network teams have added more conditionals, error handling, retries, and edge cases trying to make deterministic automation account for every possible scenario.

Reasoning agents change that equation:

- Scripts can stay focused on what they do best: fast, predictable, repeatable ex*****on.
- Agents can handle the exceptions, changing conditions, and decisions that require context.
- Your existing automation (Python scripts, Ansible playbooks, APIs, and workflows) become the tools agents call when it’s time to act.

Karan Munalingal breaks down what this hybrid operating model looks like for network and infrastructure teams. 👇

Deterministic scripts and AI agents aren't rivals, they're for different jobs. Learn the hybrid automation model every infrastructure team needs: deterministic where work is known, agentic where it's not

09/01/2026

Live tomorrow❗
Join us at 12PM ET for a live webinar on how to stay audit-ready with FlowAgents.

Compliance checks and certificate rotation are already continuous operational jobs. And with public certificate lifetimes moving toward 47 days, the window for manual processes is only getting tighter.

Dan Sullivan and Joksan Flores are going live to show how FlowAgents can help infrastructure teams keep up without giving up control.

The goal isn’t another periodic check. It’s continuously verified, governed infrastructure that’s ready when the audit comes.

Join us live tomorrow. Last chance to save your seat.

Register: https://bit.ly/4iGZl5r

08/29/2026

Only 13% of enterprises know how to define what sovereign AI is. And as AI moves deeper into critical infrastructure, that knowledge gap starts to matter.

Because sovereignty isn’t just about where your data lives. It’s about control over the models you use, the infrastructure they run on, who and what they can access, and what happens when an external dependency changes.

For network and infrastructure teams, those questions become even more important as AI moves from generating answers to taking action.

The architecture behind AI matters just as much as the intelligence itself.

Fierce Network digs into what sovereign AI actually means and why enterprises are paying attention. 👇

Sovereign AI awareness remains low despite rising concerns over data sovereignty, model control and infrastructur | The term sovereign AI is suddenly everywhere — but most enterprises still can’t define it, even as they worry about data leakage and privacy concerns around generative and agentic ...

08/28/2026

Your network automation work doesn’t have to start over just because agents are entering the picture.

In this demo, John Capobianco takes the pyATS tooling network teams already know and turns it into governed, reusable agent capabilities using PyATS MCP + Itential Builder Skills for Claude Code.

From a natural-language request, Builder Skills:
✅ Builds + connects the pyATS MCP server
✅ Exposes 26 pyATS tools to the Itential Platform
✅ Creates dedicated Interface Health + Overall Health FlowAgents
✅ Tests them against live network infrastructure
✅ Applies RBAC and limits each agent to only the tools it needs

The result: two working, read-only network health agents built from existing network expertise and governed from day one.

See the full build 👇

See how the Itential Builder Skills connects a pyATS MCP server to build two governed network health FlowAgents in under 15 minutes.

08/28/2026

One CVE ID. That's the only input an agent needs.

It searches NVD, verifies the resource model, fetches show command reference files, builds verification regex, and delivers a full impact summary with affected software versions and commands. In 55 seconds.

That's faster than reading the CVE description. The full demo goes on to scan the fleet, assess device impact, and generate a report with remediation steps for human approval.

See how it works in the full demo with John Capobianco and Joksan Flores 👇
https://bit.ly/4wTNQew

08/27/2026

It’s one thing to try and take your network back from outsourcing to save on cost. But doing it without adding the headcount to match can feel impossible.

That’s the challenge one major airline set out to solve as it began taking thousands of devices back from an outsourced provider.

The old math didn’t work: more devices meant more manual changes, more overnight windows, and eventually, a lot more people.

Today, they’re changing that equation with a governed ex*****on model that supports both deterministic automation and AI agents under the same guardrails.

Here’s the outcome:
✅ 100x more devices under internal management
✅ 95% faster routine changes
✅ 80% fewer after-hours change windows
✅ $3.2M+ in projected annual cost avoidance

See how they’re changing the economics of network operations. 👇
https://bit.ly/4xWfV5G

08/26/2026

Agents fetch results. Dogs fetch tennis balls. We’re big fans of both. 🐾

Today, we’re celebrating the four-legged Itentialites who keep our days a little lighter and our camera rolls a lot fuller.

Happy National Dog Day from our pack to yours. 🐶

08/25/2026

Modular AI agents outperform monolithic ones in production. Not because they're smarter, but because they're scoped.

Karan Munalingal's latest blog breaks down what happened when a customer built a single OS Upgrade Agent with 46 tools and a full runbook. More tools and more instructions crammed into one agent doesn't mean more capability. It means more chances for the agent to guess wrong.

The fix wasn't fewer tools. It was the same 46 tools, reorganized around intent. One monolithic agent became 8 specialized, scoped agents. Each one handles one or two steps. Each one sees only the tools that step actually requires.

Reusable components emerged along the way: a file transfer agent, an alarm suppression agent, a validation agent, now usable across firmware upgrades and config pushes for other vendors.

It's the same discipline that already governs orchestration at scale, applied to agents.

Read his complete breakdown + watch the session that shows how it works:

See why modular AI agent architecture delivers 100% consistency and creates reusable production capabilities compared to monolithic architecture.

08/25/2026

Agent roadmaps stall in a compliance review on one question about where the model runs and what it can see.

Today, John Capobianco and Joksan Flores are running a live demo to show how Itential FlowAgents run on any model, with governance built-in:

• Open weight models like Nemotron, Kimi K2, or Llama, hosted inside infrastructure you control
• Proprietary models like Claude, ChatGPT, or Gemini
• Your own model that the internal team built and fine-tuned

Every agent executes through the same governed path. Same RBAC, approvals, and audit trail, no matter which model did the reasoning.

🎟️ Join us here: https://bit.ly/4cCj3eU