A new agentic AI for ServiceNow can take a business requirement and configure the platform without a developer in the loop. That's not a future roadmap item. It's shipping now.
The PocketOS incident is still fresh. A Claude agent running inside Cursor wiped a production database and its backups in nine seconds because the right approval gates were not enabled. It was a painful reminder that agentic AI needs human oversight baked in — not bolted on after the fact.
So it's worth paying attention to what Dyna Software is announcing this week at Knowledge 2026 in Las Vegas. Their new Platform Copilot is an agentic AI that connects directly to a customer's ServiceNow development instance and configures it based on natural language input. No developer required. And unlike the PocketOS situation, they've thought carefully about what happens when AI has write access to your environment.
Platform Copilot is not a code assistant. It doesn't suggest code for a developer to review and paste in. It connects to your ServiceNow development instance, understands your specific schema and configuration, and executes changes directly.
The workflow looks like this: a business analyst or admin describes what they need — a new service catalog item, a workflow, a form — and Platform Copilot interprets the requirement, figures out how to build it for that specific instance, and builds it. The result comes back as a preview within about a minute. When it looks right, you promote it forward.
The instance-awareness piece is what makes this different from using Claude or GPT directly. Generic LLMs can build ServiceNow configurations, but they build them generically. Every ServiceNow environment is different. What works cleanly in one instance may conflict with how another organization has configured theirs. Platform Copilot reads the target instance first and builds accordingly.
Image to Build — Not Just Text
One of the more interesting capabilities shown in early use: you can take a screenshot of a legacy form — say, an old catalog item from a system you're migrating off of — upload the image, and Platform Copilot will generate a ServiceNow equivalent within minutes.
An Australian partner working with a government health client used this to migrate legacy catalog items that would have taken six months the traditional way. With Platform Copilot, the same work was completed in about a week.
That's not a marketing claim — it reflects how much time normally disappears in the gap between a business request and a completed build: requirements gathering, back-and-forth with analysts, prototypes, reviews, revisions. Platform Copilot collapses that cycle.
The Scope Question
The honest answer to "what can't it do yet?" is: it's strongest in ITSM, which is where most ServiceNow implementations start. Applications like Integrated Risk Management, HR, and security modules involve more practitioner-specific design decisions — and that's where human judgment still matters most.
But Dyna Software's CEO was candid in our briefing: every time the tool has been tested outside its core testing boundaries, it's performed well. The architecture is designed to adapt as the underlying LLMs evolve, which is a meaningful distinction. Most competing tools were built for the LLMs of 12-18 months ago. Platform Copilot is built to take advantage of what's coming.
The primary sweet spot right now is what most ServiceNow teams know well: the long backlog of low-to-medium complexity requests that never get prioritized because developers are busy with bigger projects. Catalog items. Workflow adjustments. Field additions. The kind of work that accumulates for months and frustrates business stakeholders. That's where Platform Copilot is delivering the clearest results.
Where Does the Developer Fit?
This is the question every ServiceNow developer reading this will ask. The honest answer is that Platform Copilot is designed to handle the work that experienced developers often find least interesting — repetitive, straightforward configurations driven by business requests.
The work that still requires a developer: complex new applications, integrations with external systems, anything where deep architectural decisions need to be made. AI at this stage is good at connecting the dots within a known environment. It's not yet good at designing the environment from scratch or knowing how two systems should talk to each other.
What shifts is the expectation of who sits in the middle of a business request. Right now, most ServiceNow shops have a developer or an architect translating business needs into platform configuration. Platform Copilot moves that translation to the AI layer — with a business analyst or admin validating the output before it's promoted.
That's a meaningful shift. It doesn't eliminate the need for ServiceNow expertise. But it does change where that expertise is most valuable.
The Governance Layer Matters Here
This brings us back to PocketOS. Platform Copilot has write access to your development instance. That's the point — but it also means the governance question is real.
This is where Dyna Software's other product, GuardRails, becomes relevant. GuardRails provides continuous governance for the ServiceNow platform: automated code review, approval gates for deployment, technical debt management, and source code management with gated human control. That last piece was just added, and it's directly relevant to the regulated-industry customers — finance, healthcare, life sciences — who need to demonstrate that AI-generated configurations went through proper human review before going anywhere near production.
The two products are designed to work together. Platform Copilot builds. GuardRails governs. That pairing addresses the concern that the PocketOS incident put front and center: agentic AI doing real work in real environments needs real guardrails.
Pricing and Availability
Platform Copilot uses a credit-based consumption model — pay-as-you-go, no long-term commitment. It's accessible to ServiceNow customers without requiring an enterprise contract to get started, which is a deliberate choice to lower the barrier to entry.
The next generation of Platform Copilot is due in Q2 2026, with significant capability advances.
Dyna Software is a ServiceNow Elite Partner. GuardRails is available now on the ServiceNow Store. More information at dynasoftwareinc.com.