Scaling AI For Enterprise: Inflection AI’s Roadmap to Human-Centered Solutions with CEO Sean White
Sean White is the CEO of Inflection AI, a pioneer in human-centered artificial intelligence. With a career spanning decades in tech innovation, Sean has been at the forefront of computer vision and AI technology. His experience includes key roles at Mozilla as Chief R&D Officer, work on augmented reality at the Smithsonian and Columbia University, and contributions to early web-based email systems. Sean’s passion for human-computer interaction and collaborative intelligence drives Inflection AI’s mission to create AI systems that enhance human capabilities and improve organizations.
OUTLINE:
0:00 – Introduction and Sean’s background
4:09 – Inflection AI’s position in the AI landscape
8:43 – Balancing consumer and enterprise AI products
10:14 – Inflection AI Studio approach
12:59 – Emotional intelligence in AI development
15:12 – Open source philosophy in AI
18:09 – Ideal use cases for Inflection AI in enterprises
21:00 – Rollout strategy for enterprise AI solutions
23:16 – Closing thoughts and call to action
Episode Links:
Inflection AI: https://inflection.ai/
Request Inflection AI API Access: https://docs.google.com/forms/d/e/1FAIpQLScM9Iz1KzaRlfgDrYrldoPDnXbhO5LW3-hqmQCd56YpheEN7g/viewform
Sean White’s LinkedIn: https://www.linkedin.com/in/seanwhite/
Sean White’s Twitter: https://twitter.com/seanwhite
PODCAST INFO:
Podcast website: https://www.humainpodcast.com
Apple Podcasts: https://apple.co/4cCF6PZ
Spotify: https://spoti.fi/2SsKHzg
RSS: https://feeds.redcircle.com/99113f24-2bd1-4332-8cd0-32e0556c8bc9
Full episodes playlist: https://www.humainpodcast.com/episodes/
SOCIAL:
– Twitter: https://x.com/dyakobovitch
– LinkedIn: https://www.linkedin.com/in/davidyakobovitch/
– Events: https://lu.ma/tpn
– Newsletter: https://bit.ly/3XbGZyy
Transcript:
David: Welcome back to the Humain podcast, your tech insider podcast on the data economy. We live in a data-first world, from smartphone chips to GPT models. Humane interviews the founders, investors, executives, and tech leaders creating the world we live in.
Today we bring you Sean White, CEO of Inflection AI. Sean, thanks for joining us.
Sean: Thanks for having me, David.
David: I’m excited to have you here because you’ve been at the forefront of computer vision and AI technology throughout your career. Could you share your background and what sparked your fascination with human-computer interaction?
Sean: It’s interesting you framed it that way. One of my first introductions to computers beyond experimenting as a kid was at university. I started working with Professor Terry Winograd, one of the fathers of AI in the ’70s and ’80s. They even named a test after him, the Winograd test.
Terry observed it was just as important to include the human in how we think about artificial intelligence as it was to consider the computational aspects. Very early on, I was influenced by his writings, classes, and our research together, even before the web with early internet systems like Gopher.
That thread stayed with me for decades in the projects I cared about, whether helping people communicate with early web-based email systems, augmented reality work at the Smithsonian and Columbia University, portable devices at Nokia, or at Mozilla where we worked with millions of users on internet interactions.
All of those experiences focused on how technology can make our lives better and access the world around us in different ways. When the opportunity to work with Inflection AI came up, it was a straightforward consideration. I’d already been spending time with AI and neuroscience, co-founding a group looking at how we use neurotech to help people.
Inflection stood out among the small set of companies doing large-scale models. They weren’t only trying to create the smartest model, but also taking into account what it would mean for it to be human-centered in how the AI was created.
David: It’s incredible to hear about the focus on human-centered models. Taking a step back, could you give a broad overview of where Inflection AI fits in the current AI landscape?
Sean: The AI landscape is broad. Even within AI, most of what we talk about now tends to be around large language models, but that’s just one type. We’ll focus there because that’s the wave we’re all experiencing right now. Arguably, this wave is changing society, organizations, and enterprises as strongly as the internet wave in the ’90s.
At Inflection, we’ve been building not just one of the largest language models, but an entire platform, framework, and system to reach directly to end users and enterprises. Some companies focus on making a large language model or fine-tuning. We’ve been building the full stack, orchestrating everything from user interaction to different models, fine-tuning for different contexts, and inference for end users.
We initially have a product called Pi. If you haven’t used Pi, you should try it. It’s amazing, largely because it’s not just giving book reports, but using collaborative intelligence to provide the kind of dialogue you’d want. We’re now bringing that to our new focus on enterprise solutions.
Where some view these as purely computational systems, we see it as both the user interface and computational system. We believe many historical user interfaces take time away from cognitive tasks we care about. These can be brought together under this new interface while still having computational benefits.
In terms of scale, we work with models over 350 billion parameters in size. That scale takes expertise and almost a kind of alchemy that only a few companies in the world have. It makes a difference in the experience when using systems from OpenAI, Anthropic, Microsoft, or Google. These are a different scale from smaller models.
Scaling also comes from inference. Starting with Pi gave us fantastic experience doing inference with millions of users. Handling that scale of inference is enterprise-scale, not just a laptop experiment.
David: How do you plan to balance maintaining Pi as a consumer product while focusing on enterprise offerings?
Sean: My mental model is that we can do both, similar to an early startup I did called Who Where. We had a consumer product called Mail City and white-labeled to other companies. This let us evolve the core technology, get direct user feedback, and learn from paying customers.
We’re still learning a lot from what we release in the consumer product, so we don’t plan for that to go away. The focus is certainly on enterprise, but Pi continues to provide valuable insights.
David: Can you tell us about Inflection AI’s studio business as you transition to an AI studio for enterprises? How do you approach crafting, testing, and fine-tuning custom generative AI models for commercial customers?
Sean: There are a couple ways to think about it. My partner Ian McCarthy often talks about how our product is a dialogue with users. In the consumer space, that means putting something out there, learning quickly, iterating, and growing.
For enterprise and the studio, we’re listening to both end users and enterprises. We have Voice of the Customer events where we bring in large companies from different industries to discuss what they need from AI. This lets us craft the fine-tuning for each organization.
Pi has a strong collaborative sense and is great at dialogue. A Berkeley study found it was the top model for EQ. That was all done through fine-tuning – taking particular cultures, data, and beliefs to create the response style and behaviors.
For enterprises, we fine-tune to their culture and unique datasets. This is part of the studio activity – tailoring the AI to each company’s individual needs and data.
Inflection AI has emphasized emotional intelligence alongside cognitive abilities. How has this dual focus influenced your approach to enterprise AI solutions?
Part of this stems from being a public benefit corporation. We have that double bottom line because AI is at an inflection point where it has the potential to change not just small workflows, but entire organizations in unpredictable ways.
If we’re creating tools with that kind of impact, we want to ensure they’re crafted to make us all better. It’s not just about productivity, but feeling good about what you’re doing. The emotional intelligence or collaborative intelligence is about co-evolving with something that helps us and organizations be better.
If we make brutalist tools, we end up with brutalist organizations. If we make collaborative, supportive tools that help people think at a higher cognitive level, those are the kinds of organizations we create.
David: How does your philosophy around open source influence Inflection AI’s approach to AI development and deployment?
Sean: At Mozilla, I learned you don’t want to open source by default, but by design. Think about what you want to be open and why. If you care about transparency or security, that’s one form. If you want to bring stakeholders together, that’s another form with different governance.
For us, it’s a differentiator in how we share source, whether through licensing, sharing portions, or data ownership for users. We recently worked with DTI on data transfer initiatives so users could own and export their data.
We apply this spirit to enterprises – they should own their data and intelligence. These influences continue as we set up partnerships with Inflection AI.
David: For enterprises evaluating LLMs, what use cases do you think are most ideal for Inflection AI’s offerings and why?
Sean: One simple but meaningful use case is for organizations with vast amounts of interrelated data who want a better internal interface. While some use a RAG approach with vector databases, fine-tuning is needed when data pieces need to work together. We excel at this with our platform and framework.
We also have a fantastic user interface layer. For learning, dialogue and collaboration are more effective than just asking questions and getting book reports. Some use cases we’re working on involve interfaces to multiple heterogeneous data collections that need to be interrelated.
Another interesting area is where you care about aggregating agentic aspects across an enterprise. It’s not just interfacing with one activity, but a collection across heterogeneous providers. You need something to aggregate that so you can think at a higher level and have a dialogue that goes beyond small tasks to the cognitive flow of getting things done.
David: How do you envision the rollout of AI to enterprise users, both initially and in the near to midterm?
Sean: We’re already doing this through two approaches. First, we created an API with over 10,000 organizations signed up. We’re gradually letting them test and build with our system. This helps us learn about the diverse use cases people want, from talking to spreadsheets to changing review cycles to creating wellbeing chatbots.
On the other end, we have Voice of the Customer meetings with large Fortune 500 enterprises. We spend more time on overall solutions and fine-tuning aspects. Some partners have regulatory constraints, so we can provide on-premises instances and hardware.
This gives us exposure to many new experiments and enterprises, as well as deeper integrations for companies wanting to own their AI, intelligence, and data. We’re working through each of these, which is why I’m hiring rapidly to grow our capacity to work with different partners.
David: What’s the most important takeaway for listeners, and what’s one actionable step they can take to engage further with Inflection AI’s work?
Sean: I ask everyone to think beyond chatbots and pure computational systems. Be creative in considering how you could aggregate activities in your enterprise so people can work at a higher cognitive level and accelerate what they’re doing. As we work with partners, this creativity is bringing out exciting new use cases and applications.
Once you’ve thought about that, reach out to us. Go to inflection.ai, sign up to be an API user, or contact our team. We’re excited to work with you to build the next generation of AI in a way that makes our commercial enterprises and society a better place.
David: Thank you for joining us on Humain, Sean.
Sean: Thanks for having me.