Every year, ServiceNow's flagship Knowledge conference brings together thousands of customers, partners, and practitioners to discuss where work is going next.
This year, with agentic AI moving from concept to deployment across industries, the conversations had a particular urgency. They were less about whether to adopt AI and more about whether employees would actually use it.
Few people are better positioned to have those conversations than Amy Lokey, who leads ServiceNow's global experience team, and Sumeyya Khan, vice president of digital technology experience, strategy, and product adoption at ServiceNow.
Lokey's work shapes how AI gets built into ServiceNow products; Khan's work makes sure it lands and delivers value for ServiceNow employees (who are users). Together, the two sit at the intersection of design intent and real-world behavior, the heart of the hardest AI challenges.
At Knowledge, Khan and Lokey took turns asking each other questions about the agentic enterprise and the power of good design in AI experience. What follows is an edited version of that exchange.
SK: When you're making product decisions about AI experience design, what's the principle that matters most for earning and keeping user trust?
AL: The No. 1 factor is making sure the user knows what the AI did, how it reasoned through its decision-making, and what steps it took to get the job done.
To that end, it’s critical to keep a human in the loop. The human-in-the-loop method gives you a clear audit trail. If something goes amiss, a human can step in, review the actions that an AI agent took, and figure out what should have happened.
SK: Finish this sentence: An AI experience earns trust when...
AL: It's right about something you already know. I spoke with another head of design at an AI company, and their data proves it out: People tend to approach a new AI experience by asking a question they already know the answer to. If it gets that right, the first impression goes a long way. That initial moment of accuracy is doing a lot of work.
That dynamic—trust built in the first 30 seconds—shapes everything downstream. The harder question is: What happens after that first good impression, when AI has to prove itself in the daily grind of real work?
AL: You evaluate how well ServiceNow product experiences actually land. What separates an AI experience that sticks from one that doesn't?
SK: Our favorite problems to solve are the ones where the data shows users want a solution. We do the research, then try to build something that meets them where they are and adds real value to their daily work. Joy matters too. We like to look for solutions that enhance someone's life.
AI Control Tower is how we measure and govern all that. AI Control Tower, which we use internally, helps ensure we maintain strict controls over the way AI functions in the enterprise.
SK: What's the difference between AI adoption that shows up in metrics and AI adoption that changes how people work?
AL: Sometimes you see that people are using a certain tool, but that doesn't tell you whether behavior is changing. You need both quantitative and qualitative data to figure that out. The numbers might show you the flow and where someone dropped off. They don't tell you why.
That's when you go talk to people. Qualitative data is enormously important for understanding how, whether, and why someone is using a tool. The full picture requires both numbers and qualitative data.
SK: Yes! And while we look at monthly active usage internally, we track the daily active usage closely, because it tells you if people are really changing their behaviors and the experience is sticky, making them want to come back.
SK: Switching gears, you've published a strong set of human-centric AI guidelines for ServiceNow. When you're advising customers on building their own playbook, where do you tell them to start?
AL: Data readiness and content readiness are the name of the game. We have Workflow Data Fabric and solutions for getting data onto the platform, so we can help with that. But content matters just as much. An AI experience will surface everything you have on your site, so if your data and content hygiene are off, that shows up immediately. We point customers to products on the platform that can help them get there.
It's a point that tends to surprise people. The limiting factor is often the underlying content and data infrastructure the AI is drawing from, and getting that right is unglamorous work.
AL: What should any enterprise do during the next 90 days to accelerate AI adoption?
SK: Start by educating employees on why your AI vision matters now. Hold your leaders accountable for driving it and modeling the right behaviors. Empower your teams with tools and training that are real and easy to use.
The mindset is human centered and human led. And remember: progress over perfection. Keep taking feedback to make AI, and ourselves, even better.
Find out how ServiceNow can help you simplify AI experience.
