A managed services provider is one of the clearest places to watch operating margin track directly to per-person throughput. Service Leadership, Inc., which benchmarks IT solution providers quarterly, put the average MSP's adjusted EBITDA at 11.1 percent at the end of 2024, above the 10.3 percent of 2023, while its 2026 annual report has best-in-class providers at nineteen percent or better for a sixth consecutive year. The firm credits that run to easing wage inflation combined with "a sharper focus on efficiency and AI automation," which is to say the best performers are the ones whose systems carry more of the routine work. An MSP is one instance of the pattern that runs every operating business, where output per person is set by how much of the routine work a system absorbs before it reaches someone qualified to make a judgment. The MSP just makes the relationship unusually legible, because the work is ticketed, the endpoints are counted, and the benchmarks are published.
What the technician's day actually contains
If you watch a service desk for a shift, the technician's time splits into a few repeating activities, and most of them are not the technical fix that the title implies. Tickets arrive and have to be read, classified, and routed to the right queue or person. Monitoring tools fire alerts, the large majority of which are noise or duplicates of an already-open issue. Routine maintenance, the patch windows and agent upgrades, has to be scheduled around client business hours. At the end of the quarter someone has to assemble the quarterly business review, pulling ticket volumes, response times, and asset inventories into a document the client will actually read.
The diagnosis and remediation, the part that needs a person who understands the client's environment, sits inside that surrounding work rather than filling the day. The industry's commonly cited target is roughly 350 fully managed endpoints per technician, and the firms that hit it have taken the classifying, the alert sifting, and the scheduling off the technician's plate.
Classifying and routing the inbound ticket
The first redesign is at intake. A ticket comes in as free text, an email forward, or a portal form, and a person decides what it is and where it goes. An agent reads the same input, assigns category and priority, attaches the client and affected asset from the configuration database, and routes it to the correct queue, leaving the technician to confirm rather than to triage from a blank field. Kaseya's 2025 Global MSP Benchmark survey found about thirty percent of MSPs already using AI to eliminate tedious tasks, and the intake queue is where that tedium concentrates: the manual load is heaviest and the judgment required is lowest. The technician still owns the genuinely ambiguous ticket, the one where the stated symptom and the real fault diverge. That call stays with a person, and the agent's job is to make sure it is the only ticket that needs one.
Reducing alert noise before it reaches a person
Monitoring is where volume overwhelms judgment most plainly. A fleet of agents on client machines generates a continuous stream of events, and a technician asked to read all of them stops registering any of them. The redesign collapses the stream: deduplicate alerts for the same underlying condition, suppress the transient ones that self-resolve, correlate related events into a single incident, and surface only what is both real and actionable, with the supporting telemetry already attached. The work is high-volume and rule-shaped, which makes it the natural first candidate for a system to absorb, and every suppressed duplicate is a context switch the technician never pays for. The escalation, the decision that an alert is the early sign of something worse, is the part a person keeps.
Scheduling maintenance and drafting the review
Two more activities complete the internal workflow. Routine maintenance, patching, agent updates, and scheduled reboots, has to land in windows that do not interrupt the client's working day, and an agent can propose and sequence those windows against each client's calendar far faster than a coordinator working a spreadsheet. The quarterly business review is the other. Assembling ticket counts, mean response and resolution times, endpoint inventories, and renewal flags into a draft is mechanical assembly of data the MSP already holds. The agent produces the draft; the account manager edits the narrative, decides what to flag to the client, and removes what does not belong. This is the same division of labor we describe in a redesigned process inside a staffing firm, where the system carries the assembly and the person carries the judgment and the relationship.
Why the rollup buyers care
The per-technician ratio is not only an operator's concern. Consolidation in this market is concentrated, and the analyst firm Omdia counted 169 managed services acquisitions in 2025, with private equity involved in about sixty-nine percent of them. A buyer rolling several MSPs into a platform is buying, in large part, the spread between the bottom of the industry and the top, and that spread is wide: Service Leadership's benchmarks, presented at ConnectWise's IT Nation conference in late 2025, put the bottom quartile of MSPs at 0.7 percent adjusted EBITDA and the best-in-class average at 23.4 percent. Vertical specialization compounds the opportunity, because a specialized book of business makes the ticket patterns more repetitive and therefore more mechanizable. For an acquirer, lifting an acquired firm's per-person throughput is one of the more direct value-creation moves available after close, which is why the internal workflow, not the sales motion, is where the redesign pays. This is the case we make in why services-as-software firms scale: the economics improve when delivery stops being linear in headcount.
A reasonable counter
A reasonable counter is that MSP margin gains track the broader maturation of the tooling vendors, the RMM and PSA platforms shipping more automation in the box, rather than anything a redesign engagement adds on top. There is real truth in that. Some of the industry's margin improvement is the platforms doing more out of the box, and a firm that turns those features on captures part of the gain. But Kaseya's survey found only about thirty percent of MSPs using AI to eliminate tedious tasks at all, and the firms at the top of the profitability distribution are the ones that pushed past what the platform does by default into the queues, alerts, and reports specific to their book. That last stretch is engineering against a particular firm's actual workflow, not a feature toggle, and it is where the per-technician ratio is won. It is also why the delivery model matters: a deck describing the opportunity does not move the ratio. A running agent inside the service desk does, and what it should look like is set out in what an aligned AI engagement looks like.