How to scale an MSP without hiring more staff

An MSP can sign more clients on the same service desk, without adding headcount, once the work around each ticket comes off its analysts: reading and sorting new tickets, asking users for the details they left out, answering "where is my ticket", logging time and watching the dates that fall due. That layer, more than the fixing, caps how many endpoints and tickets each analyst can carry while the response times in the agreements are still met. An AI worker can take it on, so each analyst carries more without the SLA slipping. The fixes, the changes in a client's environment and every judgement call stay with your engineers.

When does an MSP need another analyst?

It usually starts when a new client signs. Their endpoints come onto the monitoring, their users start emailing the desk, and within a month the analysts are behind. Vendors publish rules of thumb: Acronis, in January 2025, called 350 fully managed endpoints per technician the gold standard, though it says the number varies. The Service Leadership Index, which ConnectWise's benchmarking arm runs, keeps its figures for subscribers. Your own number is the one that matters. Wherever it sits, the desk copes until it does not, and the usual answer is another analyst, a salary that takes a large share of what the new client pays.

The desk breaks because the work around a ticket grows faster than the work inside it. A client with twice the users sends twice the tickets, and a fair share of them are not faults at all: a question, a request for a status, a monitoring alert that has already cleared. Each one still has to be read, matched to a client and a contract, given a priority and put on a board, by a service desk analyst; the National Careers Service lists recording issues and assigning them among the job's tasks. The sorting is not optional, because the response clock in the PSA starts when the ticket is logged, not when somebody reads it.

What work grows with every new endpoint and user?

Lay out an analyst's day. A ticket comes in with a subject line and no detail. Someone works out what is being asked, checks whether the same fault is already open, sets the category and priority from the matrix, and asks the user which device, since when and what changed. Someone answers the client who emails at three o'clock to ask if anyone has looked at it. At the end of the day, someone makes sure the time went on the ticket, because the PSA will not bill what was not entered and approved. And on a slower clock, someone keeps the list of licence terms, warranties and Cyber Essentials certificates, each good for 12 months.

None of that needs an engineer's skill. All of it is done by engineers, between fixes, and when they are on a call or on leave it waits. The MSPs and IT support page goes through each of those jobs; the question here is how they move the ratio.

What stays with your engineers?

Every fix stays with your engineers, as does every change to a client's environment, every password reset and every access change. So does every phone call, because the roles here work in writing. Every major incident, suspected security incident and complaint goes straight to a named person the moment it is recognised. So do the calls on what counts as billable, what to quote and what to renew. The agent does not touch the RMM, does not log into a client's tenant and does not need to.

What can AI staff take off the service desk?

What follows for each role is the specification agreed with you before anything is built.

For an MSP the first role is helpdesk ticket triage. Its effect on the ratio is that a ticket would reach an analyst already matched to its client, checked for a duplicate, given a priority from your own matrix with the reason written on it, and with the missing details already asked for. The dispatcher hears about a response target at risk before it is missed. Our guide to triaging tickets against the SLA sets out the method.

The second is first-line client support, which takes the status questions. It would answer them from the ticket record, in wording you have approved, and pass anything that is a fault or an access request to the desk with the facts attached. The analyst who was answering those by hand gets back the time the interruptions took.

Each role is filled by an AI worker, with a role specification, KPIs the PSA can count and a named manager on your side.

What should an MSP count before it starts?

Count, for one week, how many tickets arrived with no category, how many needed a question sent back before anyone could begin, and how many messages the desk sent that were only a status. Those three numbers are the capacity to recover. Then write the triage matrix down, because the agent will apply the one you give it and nothing else.

Run triage on new tickets for a month, with the dispatcher checking what it chose and why. Add the status replies once the triage is trusted. Keep renewals and time-entry chasing for later; they are not what breaks the desk. The aim is that the next client's endpoints come onto the monitoring and the desk absorbs them, and the hire you make next is an engineer you wanted rather than an analyst you needed.

How the same logic runs beyond IT support is in how to scale a business without hiring more staff, and what an AI worker is, and is not, is in our explainer on AI staff. Our starting price is on the homepage.

On an MSP's service desk the worker would take the sorting, chasing and answering around tickets, so that your engineers can spend their day on the fixes. Is your service desk the reason you cannot sign the next client? Talk to our agent.