Generative AI · Product discovery & change
What happens when the technology is moving faster than the people using it?
I joined a generative AI initiative supporting legal discovery workflows, working with attorneys, paralegals, subject-matter experts, product leaders, and technology teams to understand where the technology could actually improve the work.
Generative AI created an opportunity to rethink parts of the legal discovery process, but introducing the technology didn’t automatically mean people knew where it would be useful, how to work with it, or what problems it should solve. Attorneys, paralegals, and subject-matter experts brought different workflows, needs, expectations, and levels of familiarity with the technology. At the same time, the product team needed enough evidence to make decisions about where to invest, what capabilities mattered, and how the experience needed to support the people doing the work.
The challenge was to separate what the technology could do from what people actually needed it to do. The question wasn’t “What can we do with AI?” It was “Where can AI actually make the work better?”
What I found
Before we could design the right AI experience, we had to understand both the work and how people would work differently with AI.
This opportunity wasn’t simply to automate an existing legal workflow. We first needed to understand where people were spending time, where the friction actually existed, and where generative AI could meaningfully improve their work. Legal discovery looks different depending on who’s doing it, the jurisdiction, what they’re trying to accomplish, and where they sit in the process. So rather than designing around a single perspective, I used research to understand the workflow from multiple angles and identify where needs, friction, and opportunities consistently appeared.
- AttorneysHow legal professionals approached discovery work, made decisions, and where existing workflows created friction.
- ParalegalsThe detailed, hands-on work behind discovery requests, and where time, manual effort, or unclear processes affected the experience.
- Subject-matter experts & partnersThe business, process, technology, and organizational considerations that shaped what was possible.
Research happened in two phases: pre-funding discovery, where I interviewed 21+ stakeholders across 11 sessions, and then 30+ additional user interviews once the approved product moved into buildout. Across both phases, needs varied depending on the work being performed and people’s familiarity with AI. Some opportunities were obvious, while others required us to challenge assumptions about where the technology would actually be of value. There was also another challenge underneath the workflow itself: users needed a way to understand how to work with generative AI. Effective use depended not only on the tool, but on how users communicated context, intent, and expectations to it.
Synthesize
Connect patterns across interviews and workflows.
Reframe
Turn findings into clearer problems, needs, and opportunities rather than jumping directly to AI features.
Decide
Bring evidence into conversations about where generative AI could create meaningful value and what deserved further exploration.
The research wasn’t about proving where AI belonged. It was about discovering whether, where, and how it could make the work better.
What I changed
The work wasn’t just about designing for AI. It was helping people make better decisions about AI.
I worked across design, product strategy, and change to help translate an emerging technology into something grounded in real legal workflows. Partnering with the UX designer, Product Owner, product analyst, and change lead, I conducted discovery, brought user needs into product decisions, and helped build a shared understanding of both the opportunity and what it would take to move it forward. Through discovery, facilitation, and product coaching, I helped the team move beyond assumptions about what AI could do and toward a better understanding of what users needed and how their work actually happened.
Before the product was funded, I co-led pre-funding discovery and a three-week cross-functional workshop sequence to define the problem, assess build, buy, or modify options, and separate traditional automation opportunities from generative AI use cases. That work helped shape the business case behind the approval.
01 · Ground the technology in the work
Start with the workflow, not the AI.
I led discovery interviews and working sessions to understand how attorneys, paralegals, and subject-matter experts actually worked, where friction existed, and which problems were meaningful enough to explore before determining where generative AI belonged.
02 · Make AI understandable
Translate the technology into something people could use.
When traditional explanations weren’t connecting, I changed how I communicated the technology. I helped users think about generative AI like working with someone on their first day: it needed context, clear direction, and enough information to understand what was being asked of it.
03 · Strengthen the team around the work
Build capability, not dependency.
As product leadership changed, I helped newer team members build context around the product, users, and discovery work while coaching the Product Owners and change partner on problem framing, prioritization, stakeholder communication, and connecting research findings to the decisions ahead.
What changed because of it
The conversation shifted from what the technology was capable of to what the people using it actually needed.
The discovery work gave the team a stronger foundation for deciding where generative AI could create meaningful value and what users would need for it to work in practice.
01 · A clearer view of the opportunity
The team had evidence behind the problems it chose to pursue.
Research across legal roles helped distinguish meaningful workflow needs from assumptions about where generative AI might be useful, giving the team a stronger basis for product decisions and future investment.
02 · User needs shaped the product
The workflow mattered as much as the technology.
Understanding how attorneys and paralegals actually approached the work helped the team consider the context, guidance, and experience users would need alongside the AI capabilities themselves.
03 · A more usable mental model
People knew how to work with it, and knowledge stayed connected as roles changed.
Reframing generative AI as something that needed context and clear direction made prompting more understandable for legal users who were still learning. Carrying research context forward and coaching newer partners helped preserve understanding of the users, problems, and decisions behind the work as the team evolved.
What this taught me
Introducing powerful technology isn’t the same as making work better.
This work reinforced something I think matters even more as AI becomes part of more products and workflows. If people have to fight the tool, learn around it, or absorb more complexity just to use it, we’ve added to the noise rather than reduced it.
It also reminded me that understanding the user sometimes means changing how I work, not asking them to meet me where I am. When my initial explanations of generative AI weren’t connecting, changing the way I communicated gave people a more practical way to understand the technology and how to work with it.
Introducing AI into real work?
Let’s figure out where it actually helps.
The best AI question starts with the work, not the technology.
Let’s talk →or email sbinger@stephaniebinger.com
Better questions lead to brighter ideas.