Yvonne Gando spent a decade designing interfaces for people to interact with digital assistants, augmented and virtual reality, and advertisements. Then, something shifted after she moved to Salesforce in 2023. The interfaces were still there, but what she was actually designing had become harder to name. “We’re moving from designing interfaces to designing behavior,” she said. Today she leads conversation design as Senior Director of User Interface, User Experience. She and her team shape how agents reason, collaborate, recover, and earn trust. What readied her for this new kind of role, one that enables AI to act less like a piece of technology and more like a trusted collaborator?
“Poetry, teaching, and growing up speaking multiple languages taught me to pay attention to meaning, ambiguity, and what gets lost in translation,” said Gando. “Because a critical part of designing effective agents is preserving the human bits that often get lost as systems interpret intent, context, and meaning.”
She’s not alone. Many of the roles emerging inside companies like Salesforce are evolving as the company builds the infrastructure to help humans and AI work well together. The scale of that shift is significant. Chief Human Resources Officers project a 327% growth in agent adoption by 2027 and plan to redeploy nearly a quarter of their workforces into new roles as digital labor becomes standard, according to Salesforce research. The World Economic Forum projects that the skills needed for work will change by 70% by 2030, accelerated by AI. Salesforce has been living inside that transformation as its own Customer Zero by deploying the same agentic tools it sells to enterprise clients across its own workforce. It is also using a four-part framework to manage what that means for people by redesigning how work gets done, reskilling people, redeploying talent, and rebalancing work between humans and agents. The evolving roles are, in many ways, more human than ever.
Part of that transformation requires figuring out where human judgment belongs, said Austin Jackson, Director of Workforce Innovation at Salesforce. “There are times when I should have been thinking: AI is going to try to answer the question even if it doesn’t quite have the right information,” he said, recounting a research project where he had to scrap an AI-generated first pass and start over manually. “Knowing when to deploy it, knowing the extent of its capabilities, and pulling it in for the right use cases, that’s underrated.”
In other words, humans are the ones who need to decide what an agent should and shouldn’t do and how. Humans are the ones who have to review customer support conversations to identify where the human touch was lacking. Someone has to know when not to use AI; humans need to ask whether the thing being built at speed is actually the right thing at all.
No single background is sufficient for this work; you need people who come at the problem from different angles.
Philipp Hänggi, Senior Agentic Experience Specialist
That can require a certain comfort with nuance. As Gando noted, “I increasingly think what distinguishes people building human-centered AI is intellectual humility: the willingness to let evidence, users, and even failure prove you wrong so you can get closer to what’s right.”
Here is a look at a few of these evolving roles: what they’re called, what the work entails, and what skills it took to get there.
Yvonne Gando, Senior Director, User Interface, User Experience
What She Does: I increasingly describe the work as designing agent behavior. As AI moves from generating responses to taking action, the role has expanded from simply scripting interactions to shaping how agents reason, collaborate, recover, and earn trust across chat, voice, Slack, and emerging headless experiences. At its core, my role is about representing the human consequences of system design and helping make AI more useful, trustworthy, and human-centered.
Previous Role: I’ve spent my career in UX leadership, including a decade working on ads and across socially networked products, digital assistants, hardware, ecommerce, and AR/VR. Now at Salesforce I’m focused on agentic AI.
Key Skills: My UX background gave me research, systems thinking, and interaction design. Working on ads taught me decision-making in complex systems. My work on ecommerce taught me how trust shapes action at moments of commitment. And my work with AR/VR and digital assistants pushed me to think beyond screens. That mix of systems thinking, language sensitivity, and translating technical possibility into human value has been foundational. Part of designing good agents is deciding what should never be lost in translation.
Philipp Hänggi, Senior Agentic Experience Specialist
What He Does: I’m on the Forward Deployed Engineer team at Salesforce, which is a roundabout way of saying I make sure that when our customers build AI agents on Agentforce, those agents are something people actually want to use, not just things that technically function. AI projects don’t fail because of the tech. Rather, the agent works, but it feels mechanical, off-brand, or doesn’t quite fit how people really do their jobs, so it just sits there.
Our team’s role is to close that gap. We can do that in a few ways: running workshops with customers to pressure-test whether they’re solving the right problem in the first place; mapping out the end-to-end journey; designing the conversations and interaction logic so they feel natural; running user tests; and prototyping the interface so the whole thing plays to each modality’s strengths. The main thing is making sure the human experience is thought through holistically, not bolted on at the end.
Previous Role: Linguistics is actually where I started. I did a Ph.D. in conversation analysis, which is the empirical study of talk and social interaction. Think about all the subtle things we do when we talk and engage in conversation — how we take turns, repair misunderstandings, or adjust our language based on context. I then decided the theoretical armchair wasn’t for me, and I wanted to apply some of those principles to actual products. I moved into industry first as a Language Engineer at Apple, and most recently I was at UBS, leading conversational UX for the bank’s multilingual GenAI assistant. The work was essentially helping them migrate from legacy, deterministic chatbot and automated phone systems to LLM-powered conversational interfaces. Anyone working in this space knows that migration is nontrivial, and the lessons from that project shape how I think about agent design now.
Key Skills: My particular slice sits at the intersection of conversational AI, UX, and linguistics. What that really comes down to is treating conversation itself as a design material. The linguistics side brings that mindset. The UX side trains you to empathize with and research users rather than design for assumptions. The engineering side keeps you honest about what’s technically feasible right now versus what isn’t yet. The other thing I’d add is comfort with ambiguity. The space is incredibly volatile, no one has all the answers, and being willing to say let’s run an experiment and find out is, I think, a real skill in itself. No single background is sufficient for this work; you need people who come at the problem from different angles, and our team reflects that.
Sarah Khalid, Forward Deployed Engineer, Director
What She Does: Implementation teams build solutions. The FDEs make sure those solutions drive value. We go deep into a customer’s org where we figure out adoption gaps, configuration issues, and basically have technical outcomes be aligned with executive alignment. It’s something that was created by the AI era and requires deep technical expertise but even deeper human judgment.
Previous Role: I was a Success Architect at Salesforce for three years and before that I was a Technical Architect and Lead Developer.
Key Skills: Coding definitely is table stakes, but judgment is the most important part. When we are assigned to a customer, we need to go in and make a lot of judgment calls, figure out the issue the customer is running into. Pattern recognition is so important — when we see a problem with one customer, it’s important to understand if it’s a common problem that would be affecting other customers so we can share those insights internally before it becomes a problem. And then there’s executive alignment. You need to really be able to speak hard truths to executives while still maintaining confidence.
Being a success architect really teaches you perspective and how to look at things at a high level. When you’re deployed as an FDE with a customer, the first thing you need to do is take a step back and look at things from the big picture before you go in deeper. The developer side taught me depth, and the architect side taught me perspective. The FDE role combines both of these together to drive business outcomes in the Agentic Enterprise. I didn’t plan on becoming an FDE. It’s something that I grew into.
Kristina Huffman, Agentic Experience Specialist
What She Does: I’m on the FDE team, designing AI solutions for our customers by creating a vision for them of how AI can support their business or consumers and deliver the conversation, persona, and interface design for their AI agent chat interactions. To support me with prototype creation I use a combination of tools like FigJam, Figma, Cursor, and Claude Code so my customers can interact with and align on the vision.
Previous Role: I used to be in Professional Services at Salesforce where I designed experiences for customers across all of our products and industries. I did co-creation and vision workshops, researched and championed user needs, created UI mockups, and supported delivery of the solutions.
Key Skills: New AI products are not only changing how I deliver (in using new tools) but changing what I deliver: AI-driven solutions. That takes curiosity, community-building, and courage. Curiosity because tackling the learning curve of this takes a growth mindset and curiosity. Then, having a small team of experts who all bring different skills to the table, encouraging each other to succeed is the only way we are able to manage the ambiguity and pace of change. And courage because in the agentic world, while there are some emerging best practices and examples, the customer requirements and direction aren’t as established.
Jason Du, AI Delivery Lead
What He Does: My goal is to help my organization get value out of AI, which goes beyond just adopting it but using it well. In practice that means three things: figuring out how our product and engineering teams need to work differently as AI becomes part of how we build; defining what success looks like as the tools keep changing; and helping people across the org build the skills and confidence to keep up.
Previous Role: My role previously at Salesforce was Agile Coach, enabling the teams I worked with to make decisions faster. More broadly, half of my career was spent in the agile space and the other half as a Technical Program Manager in software and hardware organizations.
Key Skills: Four things matter most in this role. Ownership because you have to be willing to drive something all the way through a large organization and actually make the change stick. Then problem-solving to see the issue, bring in the right people, and push it to resolution. You need to be able to think critically, to look at problems from every angle, think about what AI can do now and what it might do next, and not to get too attached to how things have always worked. And finally, this space moves faster than anything I’ve seen before. Staying current and making decisions quickly isn’t just helpful; it’s the job.