The managed services model already went through half an industrial revolution. It just went through the wrong half.
Twenty years ago, most IT support was sold by the hour. Revenue moved with the work, so the provider carried little volume risk. Then the industry productised: per user, per device, per month. A fixed price for a defined service. Clients loved the predictable line item. MSPs loved the recurring revenue.
That was an industrial move. Fixed unit price is how industrial producers sell. An industrial producer can offer it because its production cost per unit is stable. The MSP changed how it sells without changing how it produces. Revenue became fixed. Cost of delivery stayed a function of how many tickets arrive and how many engineer hours they consume.
Fixed price, variable cost. The price per seat is
fixed. The cost per seat is not. The gap between those two is where margin
lives, and margin absorbs every swing in volume.
The pooling assumption
A fixed-price support contract is priced like insurance. Some clients have a quiet month, some have a bad one, and across the book the variance is supposed to even out. That logic holds only while the bad months remain independent of each other.
They are not. And the industry made them less independent on purpose. Standardising on one RMM, one security stack, one productivity suite and one patch cadence is correct practice. It cuts routine ticket volume and makes the estate manageable. It also means clients now share their failure modes. When a widely deployed vendor ships a bad update, when Microsoft 365 has a bad afternoon, or when a phishing campaign hits a sector you serve heavily, the tickets do not arrive from one client. They arrive from all of them, in the same week.
The CrowdStrike update of July 2024 remains the
clearest example most MSPs lived through. The cause was external, the impact
was simultaneous, and the fix was manual work on machine after machine. No
contract was priced for it.
How a month goes negative
Take round numbers. A contract book brings in 100 a month. Delivery costs 70, of which engineer labour is 55. Gross margin is 30.
A correlated incident lands. Ticket volume for the month rises by half. Your engineers are a fixed team, so the extra work is absorbed through overtime, contractors or delay. If the extra tickets cost what normal tickets cost, labour rises by roughly 27 and the margin is close to zero. Overtime and contractor rates are higher, so the real figure is below zero.
That is only the visible cost. Three more follow. SLA breaches trigger credits or damage renewals. Proactive work -the patching, reviews and maintenance that prevent future tickets- gets pushed aside, so next month starts with a higher baseline. And engineers who spend a fortnight in firefighting mode too often are the ones who update their CV.
Revenue in that month did not move. It cannot move,
because the price is fixed. Every cost of the incident lands on the MSP.
Why hiring does not fix it
The usual answer is to carry spare capacity: pay for engineers who are underused in normal months so they are available in bad ones. That lowers the margin every month to protect it in a few, and it assumes you can hire those engineers in the first place. Most MSPs cannot fill the roles they already have open.
The structural problem is that production
capacity -people- is fixed in the short term and expensive to flex, while demand
is variable and increasingly correlated. No staffing model resolves that
mismatch. It can only choose where to absorb it.
Finishing the job
The other half of industrialisation is making production capacity behave like the price does. That requires the routine share of the work to be done by capacity that scales with volume without hiring, overtime, or a contractor rate card.
This is where autonomous agents change the arithmetic.
During a correlated incident, a large share of the surge is routine: the same
diagnosis, the same fix, the same status question from forty users at different
clients. An agent that triages, recognises the pattern, correlates incidents
into problem tickets, applies the known remediation within the limits you set,
and escalates what does not fit handles that surge at close to zero marginal
cost per ticket. Your engineers take the exceptions - the work that actually
needs them. And as agents take on proactive maintenance, that work stops being
the first casualty of a bad month.
The result is a cost of delivery that moves far less with volume. Your fixed price finally sits on a production model that can support it.
That is the promise managed services made twenty years ago when it moved to
fixed fees: predictable IT for the client, and a business the MSP could
engineer. The industry delivered the first half. Completing the
industrialisation delivers the second: the same commercial model you already
sell, on a production model that no longer turns someone else’s outage into
your loss.
This is part of a series of 7 blog posts about the industrialization of Managed Services.
1 - In the 4th and current Industrial Revolution, AI disrupts IT operations fast and MSPs must respond faster
Next in the series: What happens to your L2 engineers when the L1 work is gone.
