My job at Atolio involves constantly talking with clients, attending conferences, and trying to understand where enterprise AI is headed. After spending time at HumanX and NVIDIA’s GTC, a few themes have emerged clearly:
1. Humans Still Matter
AI isn't ready to run entirely on its own, and it might never fully be. Mars’ CMO, Gülen Bengi, explained how her team manually identified nearly 5,000 obscure British slang terms and names associated with criminals, ensuring none appeared in personalized AI-driven marketing, keeping their brand safe. This isn't just an edge case—it illustrates a universal truth. Good AI systems rely heavily on human judgment. The question every business needs to answer is: How much human oversight do we really need? That balance varies by industry, risk, and specific use cases, but human oversight isn’t optional; it’s fundamental.
2. Experimentation Turns to Consolidation
Enterprise AI adoption initially resembled a gold rush—rapid trials of cloud-based tools with little long-term strategy. Now, the mood has shifted. Companies realize scattered point-solutions are unsustainable. IT teams, formerly gatekeepers, must now become strategic enablers, finding ways to balance rapid adoption with governance and security.
GitHub Copilot introduced many to AI’s power, but tools like ChatGPT quickly demonstrated why enterprises can't ignore governance: CIOs saw the urgency of enterprise licenses after realizing employees would share sensitive data on public models if internal solutions weren’t available.
Matt Foster—who leads AI at Amgen—outlined a practical playbook for scaling AI that makes sense:
- Brainstorm and ruthlessly prioritize use cases. You should be able to come up with hundreds to pick from.
- Pick the right partners aligned with your business.
- Engage end users early—map their daily workflow and show explicitly how AI helps it. Bring them along for the journey.
- Maintain transparency with regular updates—Amgen calls this their "City Plan."
At Atolio, we've observed similar outcomes: clear, focused AI use cases, like shortening ramp-up times for sales teams or reducing friction for subject matter experts, deliver measurable results. This is the way forward.
3. Sovereign AI and Security are Non-negotiable
Enterprise AI deployments now face intense scrutiny on data security. Companies ask pointed questions like, “Where exactly is our data hosted, and who has control over it?”
Rachel Jin, Chief Enterprise Product Officer of Trend Micro, describes this as “Sovereign AI”—a scenario where enterprises retain full ownership and direct control over their data, AI models, and infrastructure. At Atolio, we’ve embraced this fully sovereign approach by enabling businesses to host AI within their own cloud environments and manage their model keys independently. This ensures complete control over data privacy and security, which is increasingly becoming a requirement rather than an option.
Next up, I'll be at Google Cloud Next in Las Vegas. If you’re there, let's chat!
