ServiceNow Project Arc: The Autonomous Desktop Agent Explained
NVIDIA and ServiceNow unveiled Project Arc at Knowledge 2026 — an autonomous desktop agent for knowledge workers. Here is what it does and who it is for.
The Knowledge 2026 Keynote
On May 5–6 2026 at ServiceNow Knowledge 2026, Jensen Huang and Bill McDermott took the keynote stage to announce Project Arc — an enterprise autonomous desktop agent for knowledge workers. The announcement is one of the most concrete commitments to "agents that actually do work" we have seen from a major enterprise software vendor.
Project Arc is in early preview. The supporting governance stack — ServiceNow AI Control Tower and the NVIDIA Enterprise AI Factory validated design — is generally available now.
What Project Arc Actually Is
Project Arc is positioned as a long-running, self-evolving desktop agent for three primary user types:
- Developers — write code, run tests, manage dependencies, file PRs
- IT professionals — diagnose incidents, change configs, manage tickets
- Administrators — orchestrate workflows, reconcile data, generate reports
It thinks, writes code, executes commands, and adapts over time. The "self-evolving" framing means the agent learns from outcomes within governance boundaries.
What it is not:
- A consumer assistant
- A point chatbot
- An ephemeral one-shot tool
It is closer in shape to Anthropic's Managed Agents or Google's Project Mariner than to a copilot sidebar.
Hear it before you finish reading
Talk to a live CallSphere AI voice agent in your browser — 60 seconds, no signup.
The Three-Layer Stack
Project Arc sits on top of three named components, all introduced at Knowledge 2026:
- NVIDIA OpenShell — the open-source secure runtime. Sandboxed, policy-governed, gives the agent a controlled environment to execute code and run commands.
- ServiceNow AI Control Tower — the governance layer. Defines policies, monitors execution, logs every file read, every command, every API called.
- ServiceNow Action Fabric — provides workflow context. Lets the agent understand the enterprise's existing processes, not just files and APIs in isolation.
Project Arc is the agent. OpenShell is where it runs. AI Control Tower is what governs it. Action Fabric is what it knows about your business.
Who Should Care
Three buyer profiles will move on Project Arc first:
- CIOs of large enterprises with heavy ServiceNow ITSM/ITOM investment
- VP of Engineering organizations with developer-productivity mandates
- Shared services and BPO operations that want to compress repetitive admin work
Smaller enterprises and SMBs will get value from Project Arc more slowly — the operational overhead of governance and integration is real.
What Project Arc Does Not Cover
Project Arc is a back-office and engineering agent. It is not a customer-facing agent. The flagship use cases are: a developer asking it to refactor a service, an IT analyst asking it to triage a P1, an admin asking it to reconcile two systems.
What it does not do, by design:
- Answer your customer's phone call
- Reply to a customer's chat at 3 AM
- Handle a multilingual inbound voice conversation
- Book an appointment for a salon, dentist, or property tour
That is a different category of agent.
Still reading? Stop comparing — try CallSphere live.
CallSphere ships complete AI voice agents per industry — 14 tools for healthcare, 10 agents for real estate, 4 specialists for salons. See how it actually handles a call before you book a demo.
Where CallSphere Fits in a Project Arc Stack
Enterprises adopting Project Arc still need a customer-facing voice and chat layer. CallSphere is that layer. The stacks are complementary:
- Project Arc handles internal knowledge work — developers, IT, admins.
- CallSphere handles external customer conversations — voice, chat, SMS, WhatsApp.
- Both produce structured logs and CRM updates that the other can consume.
CallSphere ships with 6 prebuilt verticals (healthcare, real estate, sales, salon/beauty, IT helpdesk, after-hours escalation), 57+ languages, and ~14 function tools out of the box. Standard deploy time is 3–5 business days, which makes it the fastest external-comms layer to put alongside a multi-quarter Project Arc rollout.
See pricing — Starter is $149/mo (2K minutes), Growth is $499/mo (10K), Scale is $1,499/mo (50K).
What to Watch
Three things to track as Project Arc moves from preview to GA:
- The first developer-productivity case study. Watch for a public number: "X% reduction in MTTR" or "Y PRs merged per week per dev." That will set the market.
- Pricing. ServiceNow and NVIDIA have not announced public list pricing. The pricing model — per seat, per task, per token — will shape adoption.
- Governance maturity. AI Control Tower is GA, but enterprise security teams will stress-test the audit log model.
Frequently Asked Questions
Q: When will Project Arc be generally available? A: It is in early preview. The supporting governance stack (AI Control Tower + NVIDIA Enterprise AI Factory validated design) is generally available now.
Q: Does Project Arc replace ServiceNow Now Assist? A: No. Now Assist is a copilot inside the ServiceNow product. Project Arc is a long-running autonomous agent that uses Now Assist and other ServiceNow capabilities through Action Fabric.
Q: Can CallSphere integrate with ServiceNow today? A: Yes. CallSphere agents can create, update, and look up ServiceNow tickets through standard REST APIs, so a customer-facing call can open a ticket Project Arc later picks up.
Try CallSphere AI Voice Agents
See how AI voice agents work for your industry. Live demo available -- no signup required.