ServiceNow's AI Strategy: How the Enterprise Workflow Leader Listens to Customers in 2026

TL;DR#

ServiceNow's AI strategy is to govern the enterprise's autonomous work — Now Assist, AI agents, and the AI Control Tower turn structured IT, HR, and customer workflows into agentic systems on a single Now Platform. The bet is paying off: ServiceNow closed fiscal 2025 with $13.278 billion in total revenue (up 21% year over year) and Now Assist annual contract value crossed $600 million, with management targeting $1.5 billion by the end of 2026. CEO Bill McDermott frames AI as "civilization's opportunity of this century" and is pushing the company past IT into a $600 billion total addressable market spanning CRM, industry workflows, data, and security. But there is a structural irony in ServiceNow's enterprise workflow leadership: the company that automates intake forms, case fields, and CSAT surveys still listens to its own customers and end users mostly through those same structured artifacts — fields that capture what happened but rarely why. That gap is exactly where conversational AI interviews capture the reasoning that tickets and surveys flatten. This article maps ServiceNow's real 2024–2026 AI moves, shows where form-and-ticket-based listening bottlenecks genuine product and customer discovery, and explains why AI-first research cannot start with a web form — even (especially) at the company that perfected the form.

What is ServiceNow's AI strategy?#

ServiceNow's AI strategy is to make agentic AI and enterprise workflows "harmonious and synonymous" on one platform — embedding generative AI (Now Assist) and autonomous AI agents directly into the IT, HR, customer, and CRM workflows that already run on the Now Platform, then governing all of it through a central AI Control Tower. Rather than selling AI as a standalone product, ServiceNow positions itself as the system of action that orchestrates AI agents — its own and third parties' — so that intelligence becomes governed, accountable, and tied to real business execution.

The strategy rests on a simple thesis from CEO Bill McDermott: as raw model intelligence commoditizes, the durable advantage shifts to whoever orchestrates, governs, and operationalizes that intelligence inside real work. At Knowledge 2026 he put it bluntly — "AI intelligence is commoditizing, but chaos is coming" — and pitched ServiceNow as the platform that turns enterprise AI chaos into governed, autonomous work. It is the same logic Perspective AI applies to research: the model is not the moat; what you do with the conversation is. (For the broader market backdrop, see the 2026 state of customer research.)

ServiceNow by the numbers: scale that makes the AI bet credible#

ServiceNow is one of the largest enterprise software companies in the world, and its scale is what makes its AI ambitions believable rather than aspirational. The company reported total revenue of $13.278 billion in fiscal 2025, up 21% year over year, with subscription revenue of $12.883 billion, according to its fiscal 2025 financial results filed with the SEC. For fiscal 2026, ServiceNow guided to subscription revenue of $15.53 billion to $15.57 billion.

Five data points define the footprint behind the AI strategy:

• $13.278 billion in fiscal 2025 total revenue, growing 21% year over year — a rare combination of scale and growth rate at this size.
• Now Assist ACV crossed $600 million in 2025 and reached roughly $750 million by the first quarter of 2026, with management projecting more than $1.5 billion by year-end and over 30% of total ACV from AI by 2030.
• ~35 customers spent more than $1 million in annual contract value on Now Assist by the end of 2025 — a category that grew more than 130% year over year.
• 2,109 customers carried more than $1 million in total ACV (as of December 2024, up 12%), and 553 carried more than $5 million (up 18%) — the enterprise install base agentic AI now upsells into.
• 91% of net-new ACV in 2025 came from customers buying five or more products, a platform-consolidation pattern McDermott describes as AI and ServiceNow "devouring" point SaaS vendors.

That last point is the strategic core. ServiceNow is not just adding AI features; it is using AI to absorb adjacent categories — CRM, customer service, HR service delivery, security operations — and its total addressable market has grown from roughly $90 billion a few years ago to about $600 billion today. The same dynamic plays out across enterprise software: see how Salesforce approaches conversational discovery as it defends CRM, and how Workday listens across HR and finance cloud workflows.

Where ServiceNow uses AI today: Now Assist, AI agents, and the Control Tower#

ServiceNow's AI surfaces today through three layers — Now Assist (generative assistance), AI agents (autonomous task execution), and the AI Control Tower (governance) — all embedded in existing workflows rather than bolted on as a separate product.

Now Assist brings generative AI into IT service management, customer service, HR, and developer workflows: summarizing long case histories, drafting agent responses, generating code, and turning natural-language requests into structured actions. It is the front door for most customers' first AI purchase, which is why its ACV trajectory is the metric analysts watch most closely.

AI agents push from assistance to autonomy — software agents that resolve routine IT incidents, triage customer cases, and execute multi-step processes without a human in the loop for each step. McDermott's framing at Knowledge 2025 and 2026 is that agentic AI "puts IT back in control," because agents run on governed workflows rather than ungoverned chat. ServiceNow reinforced this with its acquisition of Moveworks (an AI assistant and enterprise-search company) and a partnership with Microsoft to pair Microsoft Copilot with ServiceNow AI agents in the front office.

The AI Control Tower is ServiceNow's answer to the governance problem — a centralized command center to govern, manage, and secure any AI agent, model, or workflow, including third-party ones, on a single unified platform. McDermott's pitch is visceral: he has warned that an ungoverned AI agent could "delete everything in nine seconds", positioning ServiceNow as the enterprise "kill switch." This governance-first posture is what differentiates ServiceNow from companies selling raw model access — and it rhymes with how the rest of the enterprise stack is maturing, from Snowflake's data-cloud product discovery to Okta's identity-layer AI moves.

The structural irony: a workflow giant that listens through forms#

ServiceNow's core competency — turning messy human requests into structured, automatable workflows — is also the source of its biggest blind spot when it comes to understanding customers. ServiceNow runs on tickets, case fields, intake forms, and CSAT surveys. Its Customer Service Management product is built around a self-service portal where customers submit cases through structured forms, and its CSAT surveys are configured in a Survey Designer that triggers a fixed set of questions based on case priority or type. That machinery is brilliant at capturing what happened and how fast it was resolved. It is structurally incapable of capturing why.

This is the irony worth sitting with. The enterprise workflow leader automates the form — and then, when it wants to understand whether its own roadmap is right or why a buyer churned or what an end user actually needed, it reaches for the same instrument it sells: a survey, a ticket field, a closed-loop case form. Forms flatten people into dropdowns. A CSAT score of 3 out of 5 tells you a customer was lukewarm; it does not tell you the workflow was technically resolved but the customer felt unheard, or that they only opened the case because a different, unmentioned process was broken upstream. That reasoning — the constraint, the "it depends," the unspoken alternative the customer considered — never enters a ticket field. We argue this point in depth in why AI-first research cannot start with a web form.

The bottleneck compounds at ServiceNow's scale. When you have thousands of enterprise accounts and millions of end users, the temptation is to standardize listening into structured fields because they aggregate cleanly into a dashboard. But aggregating clean fields produces confident charts about the wrong questions. ServiceNow's own product analytics can tell you which workflows are used; they cannot tell you which workflows enterprise buyers wish existed but never asked for because they assumed the answer was no.

Forms vs. conversations: what structured intake misses#

The difference between form-based and conversation-based listening is the difference between capturing fields and capturing reasoning. The table below maps where each approach wins for an enterprise like ServiceNow.

The conversational gap matters most at the exact moments enterprise revenue turns on understanding: a buyer evaluating whether to expand from Now Assist into AI agents, a champion deciding whether to renew at 3x ACV (the average expansion ServiceNow's renewing Now Assist customers achieved in 2025), or an end user quietly routing around a workflow that technically "passed" its CSAT survey. Surveys catch none of these inflection points with enough depth to act on. Conversations do — and the broader case for them is laid out in why conversations beat surveys for real customer research and the survey-alternative playbook.

How conversational AI interviews capture what ServiceNow's intake can't#

Conversational AI interviews capture the reasoning structured intake misses by letting customers and end users speak in their own words while an AI interviewer follows up, probes vague answers, and surfaces the "why now" behind a decision — then synthesizes hundreds of those conversations into structured insight automatically. This is the model behind Perspective AI's interviewer agent: instead of forcing a customer to translate their experience into a dropdown, it asks, listens, and digs into the parts that don't fit a schema.

For a company shaped like ServiceNow, conversational research closes three specific gaps:

1. Product discovery beyond the roadmap survey. Instead of a feature-request form (which only surfaces things customers already know to ask for), an AI interviewer explores the jobs enterprise admins are trying to get done — including the workarounds they've built because a workflow didn't exist. That's the difference between validating your roadmap and discovering it, the same discipline behind the 2026 guide to product-market-fit research.

2. Churn and renewal reasoning, not just scores. When a major account hesitates at renewal, a CSAT trend line is a lagging indicator. A conversation conducted by an AI agent at the moment of friction captures the constraint or competitive pressure driving the hesitation — exactly the kind of voice-of-customer signal detailed in the 2026 voice-of-customer program guide and the build-from-scratch VoC playbook.

3. Intake that interviews instead of interrogates. ServiceNow's intake forms front-load effort before value. A concierge agent flips that: it greets the customer conversationally, captures context the way a great human intake specialist would, and routes intelligently — which is precisely the intelligent intake pattern that companies like Chime used to replace onboarding forms and Hims & Hers used for patient intake.

Crucially, the objection ServiceNow would raise — "conversations don't scale, that's why we use forms" — is the one AI dissolves. The entire premise of conversational research at scale is conducting hundreds or thousands of in-depth interviews simultaneously, then auto-synthesizing them. The trade-off between depth and scale, which forced enterprises into surveys for decades, no longer holds. Teams comparing approaches can map the landscape in the modern customer research stack and the enterprise AI customer-insight platform ranking.

Who should care: product, CX, and research leaders at scaled enterprises#

This analysis is for product managers, CX leaders, and research teams inside large enterprises — especially those that run on the same structured-workflow DNA ServiceNow does. If your organization captures customer signal primarily through tickets, NPS, and intake forms, you are likely measuring satisfaction with high precision and understanding motivation with almost none.

The pattern generalizes well beyond ServiceNow. Enterprise platforms across categories are confronting the same listening gap as they layer AI onto form-heavy workflows — from Zendesk's support-team listening to DocuSign replacing forms with conversations and Twilio's developer-engagement strategy. The common thread: agentic AI is excellent at executing structured workflows and weak at understanding unstructured human reasoning — and the second problem requires a different tool than the first. CX and product leaders can see how the discovery-to-renewal arc fits together in the complete guide to AI-powered customer experience, and teams structuring this work can start from a purpose-built research workspace.

Frequently Asked Questions#

What is ServiceNow's Now Assist?#

Now Assist is ServiceNow's generative AI capability embedded across IT, HR, customer service, and developer workflows on the Now Platform. It summarizes cases, drafts responses, generates code, and turns natural-language requests into structured actions. Now Assist annual contract value crossed $600 million in 2025 and reached roughly $750 million by the first quarter of 2026, with ServiceNow targeting more than $1.5 billion by year-end.

How does ServiceNow use AI agents?#

ServiceNow uses AI agents as autonomous software workers that execute multi-step tasks — resolving routine IT incidents, triaging customer cases, and running processes without a human approving each step. The agents run on governed workflows rather than ungoverned chat, and ServiceNow manages them through its AI Control Tower, a centralized command center for any AI agent, model, or workflow, including third-party ones. CEO Bill McDermott frames this as agentic AI "putting IT back in control."

Why does form-based intake limit customer understanding?#

Form-based intake limits customer understanding because forms capture structured data — what happened, when, and a satisfaction score — but not the reasoning, constraints, and unstated alternatives behind a customer's behavior. A CSAT score tells you a customer was lukewarm; it cannot tell you why, what they considered instead, or which unmentioned problem actually drove the interaction. That "why" is where product and churn decisions are really won or lost.

How do conversational AI interviews differ from CSAT surveys?#

Conversational AI interviews differ from CSAT surveys by replacing a fixed set of rating questions with an open dialogue in which an AI interviewer follows up, probes vague answers, and captures decision drivers in the customer's own words. Surveys produce a score and a quadrant; interviews produce the reasoning. AI also removes the historical depth-versus-scale trade-off by running hundreds of interviews simultaneously and synthesizing them automatically.

Can conversational research scale to enterprise volumes?#

Yes — conversational research scales to enterprise volumes precisely because AI conducts and synthesizes the interviews. The reason enterprises historically defaulted to forms was that human interviews could not scale; AI interviewer agents conduct hundreds or thousands of in-depth conversations at once and auto-generate structured insights, so depth and scale are no longer in tension. This is what makes conversational discovery viable for companies with millions of end users.

Conclusion: the workflow leader's next listening upgrade#

ServiceNow's AI strategy is one of the most credible in enterprise software — $13.278 billion in revenue, Now Assist ACV racing toward $1.5 billion, and a governance-first agentic platform that McDermott rightly argues becomes more valuable as raw intelligence commoditizes. ServiceNow has earned its position as the enterprise workflow leader by turning chaos into structured, automatable work.

But the same structured-intake DNA that makes ServiceNow great at execution makes it, like most scaled enterprises, weakest at understanding the why behind customer behavior. Tickets, case fields, and CSAT surveys capture what happened; they cannot capture reasoning. AI-first research cannot start with a web form — and that is doubly true for the company that perfected the form. The next listening upgrade is conversational: letting customers and end users speak in their own words, at scale, with an AI interviewer that follows up and synthesizes.

If your team runs on structured workflows and wants to capture the reasoning your forms and surveys miss, start a research study with Perspective AI, explore how it's built for product teams and CX teams, or see how the platform compares to survey-based tools.