ServiceNow's Autonomous Workforce is here and it's impressive - are enterprises ready for it?

ServiceNow today launched Autonomous Workforce and ServiceNow EmployeeWorks, bringing together AI specialists and the Moveworks acquisition into a platform play designed to shift enterprise AI from experimentation to execution.

ServiceNow today announced two significant additions to its AI platform: Autonomous Workforce, a framework for deploying AI specialists that execute work end-to-end rather than assisting with individual tasks, and ServiceNow EmployeeWorks, which brings Moveworks' conversational AI and enterprise search into the ServiceNow platform as a unified front door for employees. 

I've been writing about ServiceNow's platform strategy for close to a decade, and this feels like the moment where a lot of the groundwork comes together. The "platform of platforms" approach, which I’ve argued is a competitive edge and is something that I've outlined time and time again, is once again proving itself valuable in these enterprise AI advancements, where ServiceNow is betting that whoever controls the execution and governance layer in an agentic AI future captures most of the long-term value. The question, as is always the case, is whether customers can actually get there.

What's being launched

The headline announcement today is the Autonomous Workforce, and specifically the first AI specialist available out of the box - an L1 Service Desk AI Specialist (apt, given ServiceNow’s ITSM history). And the vendor is keen to not minimize the capabilities of these new AI workers. Nenshad Bardoliwalla, Group Vice President of Product Management, said during the press briefing: 

It is not a bot. Bots were built to follow scripts. You define every branch, you define every condition... Our specialists are fundamentally different. They are not following a script. They are designed to actually do the job. They understand context, they can reason across various systems, they handle exceptions and they get better over time.

The taxonomy Bardoliwalla introduced is worth paying attention to too. He explained that agents handle task automation, agentic workflows mix deterministic and probabilistic elements, and then there is role automation - the Autonomous Workforce level - which introduces a virtualized employee representative that follows organizational policies, regulations, and auditability requirements. I want to be really clear that this is essentially a category claim, not just a product description. 

The practical difference, as demonstrated in the briefing, is that the AI specialist doesn't hand off to a human when it resolves an IT incident - it diagnoses, executes, documents, notifies the affected employee, and updates the knowledge base. Start to finish. John Aisien, SVP of Central Product Management, said it right during the press briefing:

Answers are not business outcomes.

In other words, a bot synthesizing a bunch of information is not really changing anything at a fundamental level. Whereas, an Autonomous Workforce actually gets work done. The L1 Service Desk AI Specialist is expected to be generally available in Q2 2026, but ServiceNow has already adopted it internally and is using the results of its own internal deployment as the reference point. The vendor claims that its Autonomous Workforce is handling more than 90% of employee IT requests, with the L1 specialist resolving assigned cases 99% faster than human agents. That’s quite something.

ServiceNow EmployeeWorks, which is the culmination of its Moveworks acquisition, by contrast, is generally available today. It combines Moveworks' conversational AI and enterprise search with ServiceNow's unified portal and autonomous workflows, and is available across Teams, Slack, browser, etc.

Bhavin Shah, SVP and General Manager of Moveworks and AI, who joined ServiceNow as founder and CEO of Moveworks, described the core proposition: 

ServiceNow EmployeeWorks is one of the first AI front doors that doesn't just summarize, it completes the work.

Moveworks was already used by over 5.5 million employees worldwide at the point of acquisition, which gives EmployeeWorks an unusually credible production foundation from day one.

The hub-and-spokes argument

The strategic framing that ran through the briefing is what Aisien described as the hub-and-spokes architecture. I think this is important too. ServiceNow, to hold on to its value proposition, is declaring itself as the enterprise hub - a governed, deterministic execution layer. Important in enterprise contexts.

AI models from OpenAI, Anthropic, Google, and now Moveworks are the intelligence spokes. In Aisien's words: 

Think of Moveworks as a spoke - our own spoke with differentiated access to our hub - but we will continue to develop and bring additional spokes to market: spokes powered by OpenAI, spokes powered by Anthropic, spokes powered by Google, and so on. This hub is really where work truly gets done.

It's a strong position and one which aims to protect ServiceNow from any advancements in AI models (see: SaaSpocalypse narrative keeping everyone entertained/distracted). I've written previously about how ServiceNow's two decades of workflow data, CMDB infrastructure, and cross-functional integration give it a foundation that departmental application vendors genuinely can't replicate quickly. The same capabilities that let ServiceNow orchestrate workflows across fragmented systems are now being used to orchestrate AI agents across those same systems.

The main risk in the hub-and-spokes framing, which is worth acknowledging, is whether the spokes stay spokes. OpenAI is building enterprise products directly. Anthropic is extending into workflow contexts. If a model provider decides the execution and governance layer is where they want to play, ServiceNow’s defenses will be tested. The vendor’s counter argument is that twenty years of workflow patterns, a unified data model, and enterprise context accumulated through the CMDB is not something you replicate quickly - and that's probably right, but it's a position that should receive regular scrutiny. I’d argue, more than anything, where ServiceNow is now positioning itself - where buyers can audit and effectively govern their AI deployments; AI deployments that make use of both probabilistic and deterministic tooling - is one that will be harder for the ‘spokes’ companies to replicate. 

CVS Health and the governance question

One of the most interesting parts of the virtual pre-briefing came from Alan Rosa, Chief Information Security Officer and SVP of Infrastructure and Operations at CVS Health - notably, a pre-existing customer of both ServiceNow and Moveworks before the acquisition. The CVS Health reference is strong precisely because it's not a pilot and because of the highly regulated industry the company operates in. Rosa was clear about what makes AI implementation work in a regulated, high-stakes environment: 

Boring is beautiful. Focus on value, not novelty. Don't chase butterflies. Focus on gritty, unsexy operational use cases - the ones with real ROI and a real impact on people's lives.

He said that every AI use case at CVS Health goes through a clinical, legal, privacy, and security review before touching production, with dynamic testing on top because, as he put it, a static review doesn't cut it when AI is learning and adapting. This is the sort of thing that resonates with serious enterprise buyers. On the relationship between security and delivery speed, his framing resonated: 

When security is not a gate but a partner, velocity goes up and risk goes down.

That will be music to ServiceNow’s ears. And he continued: 

Lose trust, and you lose the right to scale. The consistent themes: boring, value, trust, guardrails, repeatability. 

That’s the sort of disciplined AI implementation most enterprises will be pursuing, and it maps directly onto the governance-as-bottleneck argument that ServiceNow has been making since the AI Control Tower positioning took hold. 

What the CIO data says

The diginomica CIO network data from the past few months doesn't make for comfortable reading, however. Our January 2026 AI Projects micro-pulse, drawing on 124 respondents, found 40% negative sentiment, 35% neutral, and 25% positive when CIOs reflected on their AI projects in 2025. The February SaaS vs AI pulse found "Future SaaS Disruption and Evolution" scoring 10 out of 10 on impact - the highest of any theme - with CIOs increasingly questioning legacy SaaS spend. But they're not necessarily replacing it with platforms; they're questioning whether the economic model holds at all.

Our November 2025 research report with 35 CIOs and CTOs found that only 21.4% report AI success rates above 80% for their initiatives. Governance and risk management scored highest on both urgency and impact as barriers - which is precisely the bottleneck ServiceNow claims to solve. But the primary blockers weren't technical: poor data quality, disconnected systems, and difficulty aligning AI with business priorities consistently came up. Those are organizational problems, and no amount of platform capability solves them directly.

Shah made a comment in the briefing that I think is genuinely useful here, where he outlined the difference between what he called weak ROI and strong ROI. He said: 

A lot of the fast-moving AI tools out there are focused on improving employee productivity at a fairly superficial level. Those productivity gains come and go with the employee. When that employee moves on, the gains go with them.

Strong ROI, in Shah's view, is business process transformation that persists regardless of individual employee turnover - a mission-critical workflow that runs, delivers value, and compounds over time. The harder question is how many customers are genuinely reaching it, and what the realistic implementation journey looks like to get there.

My take

The product news today is genuinely impressive. What I keep coming back to is the gap between the disciplined implementation that makes Autonomous Workforce valuable and the organizational reality most enterprises are actually operating in. The platform requires a level of governance maturity, clean data, and aligned leadership that our CIO research consistently shows is the exception rather than the rule. ServiceNow's answer - the AI Control Tower governance layer, the EmployeeWorks front door that reduces friction - is a solid response to that gap. Whether it's sufficient will be tested over the next 12 months with customers - and we look forward to hearing those stories.

The hub-and-spokes architecture is a smart strategy if the execution and governance layer remain the battleground. I think it probably will be, for longer than many market observers suggest. ServiceNow’s core pitch is that the era of AI experimentation is over and we are now getting serious about AI implementation, where we offload serious work and tasks to autonomous agents - that’s a compelling prospect, but one which many organizations will feel ill prepared for. We’ve got a long road ahead of us and what I’d push ServiceNow to consider is how the platform itself can help manage the change. As we see time and time again, technology is an important enabler, but without people being brought along on the journey, returns will be minimal. ServiceNow’s annual user event in Vegas in a couple of months is shaping up to be a very interesting event indeed, and we will be on the ground testing these claims with execs and customers alike.