This is not a warning. It is an invoice.
And your clients are already deciding who pays it.
Challenger reported 97,006 U.S. job cuts in May – the highest May since 2020 – with AI cited in 40% of them.
But the bigger story is not just layoffs. It is a business model reset.
We see this up close in IoTSecOps.
For years, the play was to manage the edge at scale:
Fleet onboarding, policy enforcement, automated patching, alert triage, device health, compliance checks.
But autonomous ai agents change this.
They do not just automate the workflow. They collapse the headcount the workflow needed!
That reshapes the economics for large enterprises – and for the many labor-intensive security integrators and service providers whose models are built around deployed engineers, monitored seats, managed tickets, and support hours.
The next wave will not arrive as a layoff announcement.
It will arrive as a client question:
→ Can this be automated?
→ Why are we paying for headcount instead of outcomes?
→ Where is the ROI?
This applies directly to security integration, managed detection, SOC services, OT deployments, field operations, and every labor-heavy service model.
But here is the conflicting trend nobody is pairing with it:
The same enterprises cutting labor are also getting burned by their own AI spend.
There is a name for it now: token maxing.
· AI sandboxes with no cost discipline.
· Long context windows.
· Expensive models for simple tasks.
· Ungoverned agents.
In SecOps, the risk is sharper: when an ai agent loops across telemetry from a million edge devices, it does not just run up a bill, it can drown the very signal it was meant to surface.
So, no – the answer is not: “Replace the SOC with AI agents.”
And it is definitely not: “Give every remaining employee an AI sandbox.”
That is how thin teams create more cost, more noise, more risk, and a shadow AI bill nobody can explain.
This is the opening for service providers: Your clients are squeezed from both sides:
Labor is too costly to scale. DIY agentic AI is too costly to leave unmanaged.
“We use AI” is no longer enough.
The winning message is:
· We know where agents help.
· We know where they do not.
· We cut detection and response time.
· We reduce cost.
· We improve accuracy.
· We govern consumption.
· We prove the number.
The old model was labor arbitrage. The new model is governed value engineering.
Especially in security (as an example): Stop selling capacity. Stop selling activity. Focus on governed, measurable business outcomes.
Run one AI spend-and-efficacy audit (for a client) this quarter. Find the wasted tokens. Find the fragile workflows. Find the automation blind spots. Find the outcomes worth pricing around.
Interesting times indeed…