Which Half of the MSP Industry Dies First
I have spent a significant chunk of my career inside the managed services and MSSP model. I built the MSSP platform at Kudelski Security; I ran product and technology for a PE-backed MSP/MSSP group as CTO. I know the commercial structure, the delivery economics, and the renewal mechanics from the inside, which is why I am not writing this as an outsider predicting doom. The MSP sector is not dying; half of it is, and the other half is about to get stronger, and the question that matters for operators and their sponsors is which half they are actually in.
The economic model rests on three revenue lines. Managed services itself is two sub-lines operators talk about as one: a remote layer (L1/L2 ticket desk, SOC monitoring, patch cycles, backup verification) and a physical layer (racking the switch, running the cable, swapping the failed drive, keeping the on-premise environment alive at the branch office); the two have completely different economics and the AI shift lands on them very differently. Technology resale is the second line: twelve to eighteen points on Microsoft, SentinelOne, Sophos, Mimecast, Okta, and the rest. The third, and in most mid-market MSSPs the largest, is professional services: DLP rollouts, IAM deployments, SIEM onboarding, Microsoft 365 hardening, zero trust architecture. Each line is under a different kind of AI pressure, and they break on different timelines.
The remote layer goes first and hardest. Password resets, patch verification, alert triage, log lookups, firewall rule reviews are the workflows Copilot, ChatGPT Enterprise, and vertical AI agents now handle in seconds; the MSP charging €150 an hour for an L2 engineer to close a ticket in twenty minutes is competing against a client-side agent closing it in forty seconds at near-zero cost. Technology resale erodes alongside it, quietly, as AI buying agents strip out the procurement friction those twelve to eighteen points were a tax on. Professional services does not collapse; it polarises, with the middle band of standard implementations contracting sharply while complex regulated work gets more valuable. I argued in Death of the MSP Dashboard that the portal is no longer the proof of value; what comes next is harder still — when the dashboard, the ticket volume, the resale margin, and the middle band of professional services are all eroding at once, the client asks what they are actually paying for, and most providers do not have a defensible answer.
Four categories come through in better shape. First, and most overlooked, the MSP weighted toward the physical layer of managed services — racking switches, running cable, swapping drives, keeping on-premise environments alive — because the agent cannot drive to the site. Second, the vertical specialist in legal, healthcare, defence, or financial services, where regulatory depth and domain fluency outweigh generic automation. Third, the platform operator who stops selling hours and starts selling outcomes against a fixed data-and-agent layer. Fourth, the advisory-led boutique that sells judgement, delivers automation through partners, and restructures the delivery layer underneath to match. The category that does not come through is the generic mid-market provider whose P&L depends on remote ticket labour, licence resale margin, and a wide bench of mid-level consultants.
The pushback comes from the relationship-led MSP arguing their vCIO and vCISO roles are stitched into the client’s leadership cadence and they are selling judgement, not tickets. That is defensible at the top; a trusted advisor in leadership meetings is not replaced by an agent. But the advisory layer is not what pays the bills; the retainer covers a sliver of the MRR, and the rest is the same remote managed services labour, licence resale, and platform management the AI shift is dismantling. The vCIO survives; the vCIO selling an unchanged stack underneath does not.
The useful question for an MSP or MSSP CEO is not whether the model dies; it is which of the revenue lines in your own P&L is under what kind of pressure, and on what timeline. These are not technology problems; they are operating model problems, and they require CTO-level authority and a sponsor willing to underwrite an eighteen-month transition. The providers still in business at the end of this decade are the ones who accept that the model they have been selling is the model that is dying; the ones who defend the existing structure are financing their own decline.