I sat through another vendor demo last month. Different logo, same pitch: "just add our AI agent for X". I've had a version of this conversation a dozen times in the last six months, and every time, I think about all the HR leaders I've worked with over the past twenty years who bought a "must have" module one year and couldn't tell me what it actually did the next.

Agent sprawl isn't a new problem. It's tool sprawl wearing a nicer name.

‍We Already Ran This Experiment

Before anyone said the word "agent," HR had already built itself a sprawl problem. The average organization now runs 26 HR technology modules, more than double the 10 it ran five years ago, according to Sapient Insights Group's HR Systems Survey. 76% of those organizations report data silos between systems that were each bought to solve one specific problem in isolation.

That's the punchline. We didn't need AI to build a fragmented stack. We were already good at it.

‍The Interface Changed. The Instinct Didn't

Here's what "agent sprawl" actually looks like in practice: a recruiting agent here, an onboarding assistant there, a benefits chatbot, AI generated performance summaries. Each one gets added because it solves a real, narrow problem. None of them go through a central filter. Josh Bersin has been tracking this shift toward an agentic HR architecture, and the pattern he's flagging is the one I'm seeing on the ground: six months in, you've got five or six autonomous agents running across disconnected systems, each with only a partial view of the employee.

That's not a technology failure. It's the same "buy first, ask questions later" instinct that built tool sprawl, just running faster and with less human oversight watching it happen.

AI Doesn't Fix a Bad Stack. It Amplifies It.

This is the part that should worry HR leaders more than which vendor to pick next. A Gartner survey of 114 HR leaders found that 88% say their organizations haven't realized significant business value from the AI tools they've already bought.

I don't read that as "AI doesn't work." I read it as proof that AI only works when it sits on top of a stack built for a reason. A well-integrated stack gets faster and smarter with AI layered in. A fragmented one gets a fragmented AI problem, with less oversight sitting on top of the mess.

Integration Just Became Non-Negotiable

Forty six percent of HR professionals now name lack of integration as their single biggest HR technology challenge, according to HR.com's State of Today's HR Technology and Integrations 2025 research. That number was already high before agentic AI entered the picture. It's about to get a lot more expensive to ignore.

The Old Questions Still Come First

None of this means don't use AI. It means the same three questions that should have come before any tech purchase in 2020 still come before any AI purchase in 2026:

• Is the business growing, contracting, or changing direction?
• What are the real business challenges, named precisely, not vaguely?
• What does HR actually need to start, stop, or change because of them?

Answer those first. Then, and only then, ask where AI belongs.

A Final Thought

Nobody's going to look back on 2026 and say the HR leaders who won were the ones running the most AI agents. I think they'll say something different: the ones who won did the unglamorous work first, and only then asked where AI actually fit.

The tech stack was never the hard part. The thinking that comes before it is.

I go deeper on this, including the full five-step audit I now walk clients through before they touch a single AI vendor, in my new whitepaper: "AI Isn't Your HR Tech Stack Problem. Your Strategy Is." Download it by clicking on the graphic below.

‍