From Go-Live to Get-Value: The Real Work of AI Transformation

kelliabbott · yesterday

From Go-Live to Get-Value: The Real Work of AI Transformation

Welcome back to our series from the AI Center of Excellence (CoE) team at ServiceNow. Through countless advisory and hands-on engagements, we've gathered valuable insights and practical guidance that we're excited to share with the broader ServiceNow community. My role on the team is AI Strategy inclusive of OCM, training, and of course planning beyond the latest and greatest but how to achieve value today.

Here's something I see repeatedly in customer conversations: the "go-live" celebration becomes the high point of an AI journey—and then things quietly stall. The technology works perfectly. The integrations are solid. But six months later, adoption hovers around 12%, and the promised ROI never materializes.

The uncomfortable pattern: 95% of GenAI pilots stall before production. Not because the technology fails—but because organizations treat deployment as the destination rather than the starting line.

Understanding the Deployment-Value Gap

Let me paint this picture more clearly. You've invested months in planning, thousands in licensing, and countless hours in configuration. Your ServiceNow instance is humming with AI capabilities. Now Assist is enabled. Virtual Agent is trained. Everything works beautifully in testing.

And then... crickets.

The gap between AI deployment and AI value turns out to be wider than most leaders expect. The numbers tell a sobering story: 72% of executives report they've yet to forge a data culture, despite accelerating AI investments. As ServiceNow CEO Bill McDermott puts it, the root cause of failed deployments is "a lack of integration and, fundamentally, culture."

Let me show you what this looks like in practice with two real trajectories:

The success path - Stellantis: When this global automotive manufacturer decided to transform their supplier management, they didn't just deploy technology. Executive leadership drove a unified vision from day one. They implemented ServiceNow AI Control Tower for immediate visibility into AI performance and adoption. The result? Live in just six weeks with a 50% reduction in supplier inquiries. But here's what matters more: their teams actually used the tools because the culture was ready.

The stall pattern - What we see more often: A large enterprise identifies 30 AI opportunities in a year. Technical teams get excited. Pilots launch. But only a "very small percentage" ever reach production. The barrier? It's never the platform capability. It's always "cultural hesitation despite technical readiness," as one customer told us.

So how do we bridge this gap? Through three interconnected leadership actions that address the human side of AI adoption.

Three Leadership Actions to Bridge the Gap

Frame AI as Your Team's Ally, Not Their Replacement

Here's where most AI initiatives go sideways right out of the gate.

Imagine being on a call where a CIO proudly presented their AI roadmap emphasizing "90% deflection rates" and "reduced headcount requirements." I imagine the room would go silent. You would be able to feel the tension. Domain experts immediately would translate this as "we're automating away your jobs" and the adoption would stop there.

The reframe that actually works: Stop talking about replacement. Start talking about liberation from the work nobody wants to do anyway.

Think about it this way: your service desk agents didn't get into IT because they love resetting passwords 50 times a day. Your HR team didn't dream of answering "where's my W-2?" for the thousandth time. These tasks are necessary but soul-crushing. AI can handle them.

Let me share a story that illustrates this beautifully. GE Aviation's engine monitoring division uses AI to monitor thousands of jet engines worldwide. During one flight from Sydney to Hawaii, their system detected a minute temperature anomaly. Here's what didn't happen: the AI didn't "replace" the mechanics or "take over" the operation. Here's what did happen: the AI flagged the issue instantly, enabling the ground team to preposition a replacement engine in Hawaii before the plane even landed. Seamless engine swap. No delays. No lost revenue.

This is the model: AI provides the early warning. Humans make the smart decisions and execute the solution. Neither works without the other.

How this translates to ServiceNow: When you implement Now Assist, it can deflect 90% of Tier 1 requests—password resets, access provisioning, basic how-to questions. But here's the key: this doesn't eliminate your agents. It elevates them. Suddenly, they're spending their time on the complex, interesting problems that require judgment, empathy, and architectural thinking. The work they actually trained for. The work that makes the job worthwhile. Think of the new areas of the business they could focus on.

Quick proof point: Southern Glazer's Wine & Spirits saved 490 hours in just five months through AI-generated case summaries and resolution notes. But the real win? Their leadership framed it as "training AI to help you" rather than "mandatory compliance." That single reframe turned potential resistance into enthusiastic adoption.

Your action this week: Schedule a 30-minute conversation with your team. Address the elephant in the room: "Yes, AI will change your job. Here's how." Then be specific about what AI will handle versus where human judgment remains essential. Use examples like GE Aviation. Make it real. Make it honest. Make it hopeful. Ask questions.

Build Your Data Foundation Before It Becomes Your Bottleneck

Now let's talk about why so many AI pilots work brilliantly in demos but faceplant in production.

The dirty secret? Your data probably isn't ready. And I don't mean "sort of needs cleanup" not ready. I mean "AI is trying to learn from garbage" not ready.

Here's the reality check: Gartner predicts 60% of AI projects will be abandoned through 2026—not because the AI failed, but because organizations lack AI-ready data and the governance to support it. Think about that. Billions invested. Sixty percent failure rate. Not because of bad technology, but bad data culture.

So what does "AI-ready" actually mean? I'm going to give you a practical framework that cuts through the buzzwords.

The Five P's Framework gives you a checklist for data excellence:

Provenance (Where did this data come from?): Your AI needs to trust its sources. ServiceNow's Workflow Data Fabric lets you connect to external platforms like Snowflake or Databricks without copying data—maintaining a single source of truth with clear lineage back to the original system. No stale copies. No "which version is right?" debates.
Purpose (Why do we have this data?): Data collected for one purpose shouldn't be repurposed without context. The Common Service Data Model (CSDM) solves this by mapping technical assets to business services. Your AI doesn't just see "Server CPU at 90%"—it understands "Payroll processing at risk."
Protection (How do we secure it?): Here's where many pilots fail security review. AI Control Tower provides centralized governance, ensuring every automated action adheres to your security policies at the transaction level. Not perimeter defense—transaction-by-transaction guarantee.
Privacy (Who can see this?): ServiceNow Vault and Machine Identity Console ensure AI agents access only what they're explicitly authorized to see. No accidental over-sharing. No "oops, the chatbot leaked PII" incidents.
Preparation (Is it actually usable?): This is the killer. RaptorDB is built specifically for the massive transaction volumes agentic AI generates while maintaining the data integrity your analytics depend on. It's the difference between AI that scales and AI that chokes under load.

Your data readiness diagnostic (do this before your next pilot):

Sample 100 recent cases or incidents
Ask: Is our data trapped in silos or flowing through a unified fabric?
Check: Do we have standardized frameworks like CSDM, or is our CMDB an unverified mess?
Assess: Are we treating data as an application byproduct or a strategic asset with owners?

The reset action: Don't wait for perfect data to start. But don't start without governance. Assign "Product Owners" to key data domains—someone who treats internal data consumers as valued customers. Use ServiceNow's CSDM framework to give your data business context. Then launch focused experiments in governed sandboxes.

Escape "Pilot Purgatory" with Governance That Accelerates

Here's a pattern I see constantly, and it's painful to watch.

An organization gets excited about AI. They launch pilots. Lots of pilots. Marketing has a chatbot. IT has a ticket classifier. HR has a resume screener. Finance has an expense analyzer. Six months later, they have 30 disconnected experiments, zero production deployments, and a leadership team asking "where's the value?"

Welcome to what the industry calls "pilot purgatory" or "pilot sprawl." Deutsche Bank Research notes that 2025 was characterized by this exact problem—organizations with dozens of disparate AI tools where monetization fell short because projects never transitioned to production.

Why this happens: Without centralized orchestration, you get chaos. Teams can't find each other's work to build on it. Security can't track what's deployed where. Finance can't measure aggregate value. As one customer put it, "machines simply can't govern themselves."

The maturity model most organizations follow:

Stage 1 (Ad Hoc): Individual experimentation—"let a thousand flowers bloom"
Stage 2 (Projects): Team-based tools—"we built a thing for our department"
Stage 3 (Programmatic): Coordinated functions—"our AI initiatives align cross-functionally"
Stage 4 (Enterprise): Embedded ecosystems—"AI is the fabric of how we work"

The common sticking point: Getting from Stage 2 to Stage 3. This is where most organizations are stuck right now.

Here's the counterintuitive insight that changes everything: governance accelerates innovation. I know, I know—everyone thinks governance means bureaucracy and delays. But here's what actually happens: bad governance (or no governance) means delays when things blow up. Good governance means you move fast with confidence.

ServiceNow AI Control Tower demonstrates this perfectly. It gives you a single pane of glass to discover, map, and manage all AI assets—whether you built them in-house or sourced them externally. Instead of discovering "shadow AI" six months after deployment (usually during a security audit), you have real-time visibility from day one.

Your 90-day escape plan:

Week 1-2: Audit your current AI portfolio. Make a spreadsheet. Every pilot, every experiment, every "proof of concept" someone's running.
Week 3-4: Identify the "shadow AI"—tools running on local machines, unauthorized cloud accounts, or external platforms you don't control.
Week 5-8: Migrate everything into governed sandboxes with AI Control Tower providing visibility.
Week 9-12: Establish "kill criteria"—if a pilot can't demonstrate clear value path within 90 days, shut it down and reallocate resources.

What success looks like at Stage 4:

Let me show you what the other side looks like with three quick examples:

Siemens automated operations across 11 global locations, saving over 1 million hours annually. Not through 30 disconnected pilots—through one coherent ecosystem.
Griffith University achieved an 87% increase in self-service by unifying ITSM and Customer Service Management under one AI strategy.
KPMG built fully integrated agentic AI across finance, HR, and procurement that functions as cohesive enterprise intelligence, not a collection of disparate tools.

Notice the pattern? They didn't get there through more pilots. They got there through ruthless focus and unified execution.

Your Path Forward: Start This Week

I know what you're thinking: "This sounds like a lot." It is. But here's the thing—you don't have to do it all at once. In fact, you shouldn't. Pick one action and do it well.

Choose your starting point this week:

Option 1 - Reset the narrative: Schedule that 30-minute team conversation. Address AI fears head-on. Share the GE Aviation story. Define specifically what AI handles versus where humans have authority. Make it a conversation, not a presentation.
Option 2 - Assess data readiness: Sample 100 cases. Run them through the Five P's Framework. Identify your weakest P. Launch a 30-day sprint with one team documenting to AI-ready standards. Measure the difference.
Option 3 - Audit your pilots: Map every AI initiative against the maturity model. Be honest about which stage you're in. Implement AI Control Tower for visibility. Set that 90-day kill criteria or whatever time period works for your culture.

Then build momentum this quarter:

Launch your "Feed the Brain" campaign—reframe data quality as AI training, not compliance
Establish a cross-functional steering committee (or partner with ServiceNow Impact) that meets biweekly
Host monthly AI showcases where users share their stories—not executives presenting metrics
Adjust performance metrics temporarily: reward AI engagement and data quality during the adoption period, not just speed
Invest in training. You can even get started with the free trainings on ServiceNow University.

Scale it this year:

Prove value in one focused area before expanding (follow the Stellantis trajectory)
Build learning systems that persist: weekly standups, monthly showcases, quarterly strategy reviews
Integrate AI adoption into onboarding, performance reviews, and career advancement paths
Document everything: "what we tried," "what happened," "what worked," "what's next"

The Leadership Challenge (And Opportunity)

Let me bring this home with something important.

The organizations that successfully bridge the Deployment-Value Gap aren't the ones with the biggest AI budgets or the most sophisticated models. They're the ones whose leaders master three interconnected responsibilities:

Framing: They position AI as augmentation through narratives that appeal to expertise rather than threaten it. "Infinite intern," not "replacement."
Structuring: They build data culture through frameworks like the Five P's, treating data excellence as strategic practice, not IT overhead.
Evaluating: They guide the maturity journey from ad hoc experiments to enterprise ecosystems with governance that accelerates rather than inhibits.

But here's what matters most: they recognize this is cultural transformation, not technical implementation. They prioritize psychological safety and continuous learning over go-live dates and deployment metrics.

So here's my question for you: Will your organization develop the cultural foundation necessary to make AI adoption meaningful?

Because the technology is ready. The platform works. The question is whether you're ready to do the harder work of organizational change.

The future belongs to leaders who architect the human and organizational systems that allow technology to create lasting value. Not leaders who deploy the most tools—but leaders who create the conditions where those tools actually deliver on their promise.

The choice, and the opportunity, is yours.

Resources to Get Started:

Need Guidance? Contact your ServiceNow account team or explore ServiceNow Impact services for tailored support.

Views expressed are my own and do not represent ServiceNow, my team, partners, or customers.