Hadley Ferguson has been with Uber for twelve years, where she currently serves as Director of Global Content for Community Operations. She leads a global team of more than 200 people responsible for creating and maintaining the knowledge resources that power support across Uber. As an early employee who grew with the company, Ferguson understands both the scrappy innovation of small teams and the complex needs of a global enterprise.
In our conversation, Ferguson shares practical strategies for rolling out GenAI successfully — including evaluating vendors, developing in-house tools, and driving team adoption. Her insights offer a valuable playbook for executives implementing GenAI at scale.
This conversation is part of InnoLead’s new “Early Adopters” series highlighting business leaders who are driving digital and AI transformation at major companies. For initiatives to succeed, there must be strong partnerships between innovation teams and business executives. Through these interviews, we’ll share perspectives from functional leaders who are putting emerging technology into practice.
• • •
What does your team do at Uber?
My Global Content team sits within Uber’s Support organization, Community Operations. Our core mission is building knowledge resources for everyone who needs support with Uber. This includes content for our Support agents, self-serve help articles for customers and earners, and content optimized for our AI-assisted Support experiences and systems. Whether you’re looking at Uber Help in the app or on the web, or getting assistance from an agent, you’re interacting with content our team created.
How is GenAI transforming customer experience at Uber?
While efficiency is important, we’re really focused on using GenAI to help create personalized, contextualized, high-touch experiences. It’s about meeting users where they’re at.
We’ve developed two key approaches to AI-powered Customer Support: First is “zero touch,” where AI resolves Customer Support issues, like a simple payment or account concern, without human intervention. Second is “one touch,” where AI does the fact-finding and prep work before customer interactions. Our Support teams can immediately jump in with solutions— instead of asking a million questions — because the AI puts all the right information at their fingertips.
This approach is particularly powerful given both our scale and complexity. We’re a high-volume company with lots of rides, deliveries and transactions happening all the time.
What GenAI tools are you using?
Uber has access to Writer, OpenAI, and Gemini. My team’s primary focus has been on Writer, which we’re using alongside our Marketing and Legal teams. This shared backend infrastructure is valuable because it gives us access to consistent legal guidelines and nomenclature across teams.
Our teams have also built our own custom tools, such as a Product Requirement Document (PRD) Scraper using GPT. The tool automatically scans PRDs related to new Uber product launches. These are super dense documents, and the tool allows us to identify Support-related actions our team needs to take for launches. It’s still in pilot, but it’s already transforming our work by letting us proactively find critical Support information instead of hoping it gets flagged to us.
How do you drive GenAI adoption across your team?
It’s critical to take a friction-free approach that builds trust. The team needs to see we are adapting the technology to their needs, not the other way around.
The team needs to see we are adapting the technology to their needs, not the other way around.
So, we’ve focused on integrating AI tools directly into our team’s flow of work. We integrated our AI — in this instance Writer— with Google Suite since we are already drafting content in Google Docs and Sheets. There is no need to go out of our day-to-day work.
Interestingly, this mirrors how we think about Customer Experience supported by GenAI. It’s a really similar philosophy. Can this tool meet Support team members where they’re at, and personalize and contextualize for their specific roles?
You mentioned your team built the PRD Scraper on your own. How have you fostered a culture of continuous learning and technical engagement among the team?
The team that built our PRD Scraper has been focused on innovating and finding areas where we can eliminate manual effort with AI. To accomplish these objectives, their leader introduced “Feed Your Mind Fridays.” These were days where the team members (spread out across the globe) came together to take a dedicated course and learn how to best utilize these new tools. It was powerful because they had a shared accountability with check-ins and clear milestones for themselves. The output from this team, outside of this PRD Scraper example, has been massive.
The key was giving people who were super excited about GenAI the accountability to create real impact and learn a new skill while they were doing it.
How did you evaluate startup solutions for your GenAI needs?
Speed to market was key for us. We needed to evaluate, select, and implement a GenAI solution before Q2 2024. So we partnered with Accenture to help us run a large-scale industry evaluation of the entire GenAI landscape. We focused on three key things: integration speed, the security approval process, and product flexibility.
For startup vendors specifically, I really lean into talking to their CEO and CTO. I want to understand what makes their tech stack unique and dynamic. But I’m also super upfront about our complexity; we’re big and we’ve grown to where we are for many different reasons. That means we’re a global, 24/7 operation where we can’t afford system outages or downtime. We need enterprise-grade reliability, security and overall stability.
Another crucial step: I always ask for client references and the ability to do an intro and discussion with them. With Writer, we had some really great conversations with clients that showed both immediate use cases for us and valuable learnings. Writer was also genuinely interested in our unique Support use case, which we saw as a positive for our future partnership as it benefited both sides to really invest in this.
What are the common pitfalls in corporate-startup partnerships?
Having grown up in Uber’s startup days, I get that startup optimism — the “of course we can do this, we can launch by tomorrow” mindset. I was there myself 12 years ago. And sometimes you really can move that fast.
But where things often break down is when startups over-promise and under-deliver because they don’t truly understand the complexity of working with a large enterprise. When you’re dealing with tech platforms, it’s not just about the product. It’s about servers, load testing, and operations. These are the technical details that can make or break a partnership.
That’s why I always approach these relationships as true partnerships. I tell any startup or smaller vendor that I work with: “Help me see the blind spots. Let’s build a solid operational plan for how you’ll actually scale with us.” Being optimistic is great, but when you’re a bigger company, the stakes are much higher.
I always approach [startup] relationships as true partnerships. I tell startups: ‘Help me see the blind spots. Let’s build a solid operational plan for how you’ll actually scale with us.’
What other emerging technology are you excited about?
We’re currently spending a lot of time looking at “modern knowledge” technology that enables higher accuracy and quality content through governance and knowledge health diagnostic tools. Simply put, it’s tools that can help us govern and manage our content ecosystem more effectively. Right now, we’re super focused on diagnostic tools that can spot issues or inaccuracies in our Support content – with a goal of ensuring highest accuracy and quality.
Here’s a good example of what we’re dealing with: We have very specific legally approved language around earners that can differ across states and countries due to different rules and legislation. Right now, we’re manually ensuring everything stays in sync and accurate. With these new diagnostic tools, a lot of them GenAI-based, we could actually detect inconsistencies, gaps, or outdated information automatically. It’s like creating our own funnel to proactively pre-identify issues.
We’re currently assessing what capabilities we will buy versus build for this; and it fits nicely with a much larger area of work we are doing currently to define what our tech stack needs to look like between now and 2027 to ensure we can support the needs of both humans and AI.
Paulina Karpis leads Early-Stage Platform for B Capital, a global multistage venture firm investing in B2B startups.
Photo by Viktor Avdeev on Unsplash
