On GenAI (Data Analytics) Users, Use Cases and Concerns
Sec. Research • Survey • Strategic Concept Testing • UX Research
Overview
A quick disclaimer: Please note that ServiceNow research is held to strict confidentiality standards, so this page focuses on approach and impact, not specific findings, data, or the product itself. A lot of the materials that, in other projects I was able to share, even if heavily blurred, I CANNOT do the same here. I'm happy to speak to more detail in person.
Throughout 2024, ServiceNow began shipping Generative AI functionalities into Platform Analytics, bundled into one of the existing license packs. The push was real, but the validation wasn't. At the time, we knew almost nothing about who these users actually were, what they'd realistically use GenAI for, or what would stop them from trusting it.
This project is one of several AI-focused studies I've led at the Platform Analytics team in ServiceNow. Generative and predictive AI are reshaping nearly every part of the data analytics experience, and this study became one of the foundational references that later AI-in-analytics work continued to build on. This is less about validating one feature, and more about understanding how people think about AI in this space at all.
Team
• 1 Staff UX Researcher
• 2 Product Managers (supporting)
• 2 Product Designers (supporting)
My Contributions
• Led User Research end-to-end:
- Framing the research goals and questions
- Running secondary research alongside primary interviews
- Synthesising findings into mental models, use cases, and adoption requirements
- Sharing results and driving their use in roadmap decisions
Timeline
• late Q3-early Q4 2024
The Process
I followed the same process I use for other comparable foundational work — Kickoff, Planning, Preparing, Running, Analysing, and Sharing — compressed here into a single study, since this project didn't require a follow-up quantitative stage.
Kickoff
The starting point was a gap, not a brief: a functionality was rolling out to a license base where only a small fraction of users held the entitlement. As such, the kickoff stage focused on:
Scope: mental models, use cases, and concerns, rather than a single feature's usability
Stakeholders: Research, Product, and Design across Platform Analytics, and potentially other BUs.
What we already knew: very little, beyond industry-wide predictions about GenAI in analytics; no ServiceNow-specific validation existed yet
Why this mattered: building GenAI features without understanding the people behind the use cases is a fast way to build the wrong thing
Planning
Based on the information gathered during kickoff, I developed a Research Plan, structured around two key goals:
Identify potential use cases for GenAI among Platform Analytics users;
Assess how well the ongoing products' direction could address the needs of the at the time addressed user segment(s).
Research Questions
The plan organised around questions like:
What are users' current mental models of GenAI in the context of data analytics — their tools, triggers, and reasons for use?
What do they expect from ServiceNow's GenAI, specifically in an analytics context?
What concerns do they raise, and how do these compare to known concerns about AI more broadly?
Methodology
This was framed as exploratory, mixed-methods research, run in two connected stages:
Stage 1: Grounding (what did we know already?)
Secondary research reviewing third-party industry reports (Gartner, Forrester, and others) alongside a substantial body of existing ServiceNow AI research, to avoid re-discovering what was already understood. This wasn't background reading done for its own sake, it surfaced critical findings that I carried directly into the rest of the project (details ommitted due to NDA constraints), and which fed directly into Stage 2 design.
Stage 2: Assessing it with users
In-depth, semi-structured interviews with ServiceNow users who create or consume analytics content, to test and discuss industry-level assumptions against real, ServiceNow-specific behaviour.
Preparing
For the interviews, I designed a session script moving through: warm-up and role context, current GenAI habits and tools, two short ideation exercises exploring different AI concepts, and a discussion of how AI could support each stage of the analytics creation journey. It's important to mention that during these discussions, we anchored these topics to participants' current tasks and pain-points. This was NOT a standard concept demo or validation study.
The secondary research directly shaped that script, rather than sitting separately from it. The lifecycle-based use case categories the analyst reports pointed to, became a structural backbone for our discussions, so I could ask about each stage on its own terms instead of about "AI" as one vague topic. Moreover, because the secondary research flagged a potential shift in customer types and behaviors, I made sure recruitment (and the script's warm-up questions) captured those nuances.
Running
I ran a round of moderated interviews, UXR-led with Product and Design support, with ServiceNow customers who create (and/or consume) analytical content as part of their role. Participants spanned both technical positions (e.g. data engineers, technical consultants) and business-facing ones (e.g. product managers, service owners, IT leadership). This ensured a relevant coverage of the full spread of people involved in building or relying on analytics.
Sessions were conducted remotely, with either Zoom, or MS Teams (depending on the participant's preferences or policies), and ideation exercises were conducted with support a whiteboarding application (Miro).
Analysing
Through thematic analysis of the interviews, I organised the data around three lenses:
Mental models: if/how participants already use AI in their day-to-day, and what they implicitly compare any new AI experience to.
Use cases: mapped by level of interest against the concerns each one raised, separating "high interest, low concern" ideas from ones that sounded good but triggered hesitation. As with other studies you'll find in this portfolio, we're not simply looking at anecdotal reports from participants - we're gathering evidence, based on the participants' own context and current experiences.
Adoption requirements: the recurring conditions participants attached to "trusting AI" at all.
When looking at this interview data, it was also important to go back to the secondary research, to check the interview data against it rather than treat the two as separate deliverables. Some of the predicted themes held up directly, with interest in some use cases echoing the broader industry pattern almost exactly. Others needed real customer voices to become credible, adding caveats that previous reports simply hadn't: trust, not skill, was the actual gating factor; a distinction only surfaced because we gathered user data to set against the industry "narrative", not the other way around.
Sharing
Findings were shared as a full report, with an executive summary to make the results easy to act on quickly. As always, continuous updates to stakeholders were provided throughout, so the team didn't have to wait for a final readout before starting to plan.
The secondary research review was also shared on its own, as a standalone document, not folded silently into the final report. Meaning that other people working on AI in Platform Analytics could go straight to a synthesized view of what the industry already knew, without needing to wait on (or duplicate) a full primary study of their own first. This was also very useful in our many follow-up or related AI research studies.
Outcomes & Reflections
This research fed directly into virtually every follow up study regarding the application of AI in data analytics in ServiceNow. This project gave also the team a shared, evidence-based starting point instead of assumptions. The mental models, the use-case framework, and the adoption requirements it surfaced weren't specific to one feature; they became reference points for how ServiceNow thinks about AI in data analytics more broadly, informing conversations well beyond the initial concepts themselves.
Again, this was one piece of a wider thread of AI research I've led at ServiceNow, but a piece that helped establish the vocabulary the rest of that thread now shares.
© Tiago Gonçalves, 2026