Agentic AI Is Everything — But Should You Let Someone Else Own It?

At The Six Five Summit: AI Unleashed 2025, Aneel Bhusri — Executive Chairman and Co-founder of Workday — made a bold claim:

“Generative AI was a head fake for the enterprise.”

His reasoning? While GenAI dazzled with summaries and chat interfaces, it didn’t transform how work actually gets done in the enterprise. But agentic AI — autonomous, decision-making software agents that operate across HR, finance, procurement, and beyond — will.

And he’s right. But he’s also triggering a conversation we need to have: who controls the platform these agents run on — you, or your vendor?

From Copilots to Digital Workers

This next chapter of enterprise AI isn’t about assistants. It’s about autonomous agents that complete business processes end-to-end:

• Reviewing and submitting expenses

• Conducting financial audits

• Optimizing procurement flows

• Managing employee onboarding or offboarding

Workday’s vision is to treat these agents like “digital employees.” That means provisioning, performance management, and org chart integration — all within their HR and finance systems.

But here’s the question:

If agents become your new digital workforce, should their OS be owned by a software vendor?

Vendor Approaches to Agentic AI: A Snapshot

To provide some context, here’s how major players are positioning themselves in this space:

The takeaway: most vendors want to own the agentic runtime. Few offer enterprises true sovereignty.

The Strategic Tradeoff: Speed vs. Sovereignty

There’s no denying vendor platforms deliver value:

• Canonical data models for HR/finance

• Fast time-to-value

• Security, compliance, and governance baked in

But the more you rely on vendor agents, the more control you surrender over:

• Business logic

• Workflow orchestration

• Security posture

• Custom integrations

• Cost and exit strategy

And the cost? Consider this:

A 2024 study by Gartner noted that:

“Enterprises with high SaaS lock-in report 20–40% higher long-term costs when modifying workflows or exiting platforms.”

Agentic AI amplifies that risk. Every agent becomes another thread in your operational fabric — harder to untangle over time.

So What Does “Owning Your Agent Layer” Look Like?

It doesn’t mean rejecting vendors. It means building a hybrid model that protects your flexibility:

• Orchestrate agents via open-source tools (e.g., LangGraph, Kamiwaza, Open Agent Protocol)

• Use identity and access controls you manage

• Wrap agent logic in APIs under your governance

• Decouple data models from vendor schemas

This approach gives you optionality — to build in-house when it matters and leverage vendor speed when it doesn’t.

The Talent Gap Is Real

Let’s be honest: building and managing this kind of infrastructure takes skill.

You’ll need:

• Machine learning engineers or prompt architects

• Security engineers for AI-specific threats

• Platform teams to build observability and audit layers

• Product leaders who understand agent workflows

For many organizations — especially mid-market — this is a steep hill. That’s why vendor platforms will win early. But over time, more orgs will need to build agentic capability into their core engineering strategy.

Questions Every CTO Should Be Asking

1. Who defines the logic and lifecycle of our agents — us or our vendor?

2. What’s our exit strategy if the agentic platform becomes a bottleneck?

3. Can we track, audit, and debug agent decisions across systems?

4. Are we investing in internal capability — or buying temporary convenience?

5. How will we adapt if agent workflows become our IP?

Ethical & Accountability Dimensions: The Next Frontier

We’re not just talking about architecture — we’re talking about responsibility.

As agents move from suggestions to autonomous action, we must ask:

• Who’s liable when an agent causes harm or financial loss?

• How do we ensure agents don’t reinforce bias from historical enterprise data?

• What governance exists for agent-to-agent interactions?

• Can we trace and explain agent behavior when decisions are challenged?

These aren’t hypotheticals. In HR, Finance, and Legal workflows, these questions are inevitable — and the answers need to be baked into the platform you choose.

In vendor-owned ecosystems, your visibility may be limited. In sovereign platforms, you build the rails yourself.

Either way, accountability can’t be an afterthought.

Final Thought: Agentic AI Is Inevitable — But Ownership Is a Choice

Aneel Bhusri is right: Agentic AI is the next operating layer of the enterprise.
But just like with cloud and DevOps before it, we must ask:

Who owns the platform? Who owns the logic? Who owns the risk?

Speed matters — but so does sovereignty.
Convenience is great — until your competitive advantage depends on someone else’s roadmap.

Let’s move fast. But let’s not lose control.

You need help with your AI journey and want to engage the CTO Advisor beyond blog posts and video content? Subscript to Keith On Call via Gumroad or email me directly [email protected]