AI could make the traditional MSP model obsolete without replacing the MSP

The monthly contract has an advantage

Calling every traditional MSP a break-fix business misses an important commercial distinction.

Under an hourly contract, reducing the time needed to complete a task can reduce the supplier’s revenue. Under a fixed monthly agreement, the same reduction can improve the supplier’s margin. An MSP charging per user already has a reason to prevent faults and automate repetitive work, as long as service quality holds.

AI therefore threatens different contracts in different ways. It can undermine the sale of technician hours while strengthening the economics of a managed subscription.

ConnectWise makes that distinction visible in its modelling of AI and automation in the MSP industry. Its scenarios explore lower labour costs alongside higher tooling costs. They also show how aggressive price cuts can erode the financial benefit. These are modelled scenarios, not measured returns that another MSP can assume it will achieve.

The immediate argument between supplier and customer will concern who receives the savings. An MSP can reasonably retain a share to fund better coverage, training and recovery capacity. A customer can reasonably ask why a service costs the same if the supplier needs less labour to deliver it.

Neither position requires abandoning per-user pricing. A predictable monthly bill remains useful to a small business. What becomes harder to defend is a fee whose only explanation is the amount of work the provider used to perform.

Compare the Cloud has previously argued for selling recurring managed outcomes rather than relying on hardware margin. AI adds a further test. Once delivery changes, the provider has to show whether the customer receives faster recovery, fewer interruptions or broader support.

Changing the invoice label to “outcomes” proves none of those things.

Automation is entering the work itself

The strongest evidence for disruption sits in specific workflows.

Rewst describes how Air IT built a custom workflow to classify incoming requests, including work associated with its leaver process. Its account explains how ticket information goes to OpenAI for categorisation and then updates the service-management system. That is a concrete connection between AI and the administrative work surrounding support. It is also a supplier’s account, rather than an independent evaluation of Air IT’s entire service operation. Rewst’s explanation of automated ticketing supports the narrower claim.

Classification matters because it can start the next action. Once a system identifies a request correctly, a defined workflow can route it, collect missing information or prepare an approved change. The commercial opportunity comes from joining those steps together without requiring a technician to move information between systems.

But identifying a leaver request and safely completing an employee’s departure are different achievements. A buyer needs to know which accounts were disabled, which records were retained and who authorised the action. Faster classification alone does not answer those questions.

Atera is making a more ambitious commercial promise. On 5 May 2026, it announced that Robin would independently resolve 50% of an enterprise’s Tier 1 and complex Tier 2 technical tickets within 90 days of onboarding, or it would waive all fees. Atera’s announcement is evidence of that offer, not independent proof that customers consistently achieve it.

Even as a promise, it changes the buying conversation. A software supplier is inviting comparison with completed technical work. For an MSP, the relevant questions become how the offer defines resolution, which tickets qualify and what happens when an apparent fix fails.

Britain also has a useful example of a provider changing its own economics. In November 2025, Infinity Group chief executive Rob Young told IT Europa that the company had grown 35% over 12 months with the same number of people. The report describes changes across sales and project delivery, alongside a narrower Microsoft focus and repeatable services. Infinity’s account of growth through automation illustrates how a provider says it has expanded capacity.

It does not isolate AI’s contribution from those other changes. Nor does it show an MSP being replaced. It shows an incumbent using software and a more standardised business to serve additional demand without proportional hiring.

That distinction could define the next stage of competition. A provider does not need to automate every support task to put pressure on a rival whose costs rise with each new customer.

Abroad, the contest is over who owns delivery

International developments suggest two competing futures.

In one, customers buy a larger share of technical work directly from software platforms. In the other, MSPs use those platforms to widen the services they sell. Both can happen at once, even within the same customer account.

Bessemer Venture Partners’ October 2025 analysis of Indian IT services argues that AI puts pressure on businesses tied to billable hours. It also expects continuing demand for services around legacy systems and customer-specific work. This is an investor’s thesis, with a commercial interest in new entrants, rather than a demonstrated forecast for British MSPs.

Its useful insight is the incentive problem. A supplier paid for time spent has to work through the consequences of selling something that reduces that time. A new entrant can design its price and staffing around automation from the outset.

Pax8 is pursuing a channel-based response. At its June 2026 event in Salt Lake City, it announced a Managed Intelligence Provider programme, an Agent Store and services that partners could deliver with Pax8 experts or under their own branding. Pax8’s launch announcement describes a model in which the partner retains the customer relationship while buying technical delivery behind it.

That route could help a small British provider sell work it cannot staff alone. It also raises a question about bargaining power. If the platform supplies the automation and specialist delivery, how much control does the local provider retain over its own costs?

An MSP can remain the name on the invoice while becoming more dependent on another company’s pricing and product decisions. Consolidation need not arrive through an acquisition. It can arrive through a shared platform that becomes difficult to replace.

For customers, the practical consequence is that the support contract may conceal a longer chain of suppliers than it once did. Knowing who operates the automation becomes part of knowing what you have bought.

The strongest counterargument deserves to win part of the debate

A capable MSP could emerge from this transition with better margins, happier customers and more work.

It already has the customer relationship. It knows which applications matter, where documentation is incomplete and which changes require caution. Under a fixed fee, it benefits when routine work takes less time. Customers may prefer to let that existing provider adopt AI rather than take on another supplier themselves.

There is no sound basis for declaring that model obsolete.

Nor does physical support disappear when a helpdesk becomes automated. Replacing a failed device still requires a physical service. A customer can buy that separately, but somebody must arrange it and decide whether it will solve the problem.

The more difficult counterargument concerns responsibility during an ambiguous failure. Consider a hypothetical British distributor whose warehouse cannot release orders. Its connectivity dashboard looks normal. The business application supplier reports no incident. An automated assistant suggests restarting equipment, but the warehouse manager cannot tell whether doing so will interrupt the remaining working terminals.

That business needs someone authorised to investigate across suppliers and agree a course of action. Its immediate concern is who will stay with the incident until orders move again.

The NCSC’s guidance on choosing an MSP puts practical weight on clear responsibilities, response arrangements and liability. It also asks buyers to examine access controls, logging and what happens if the MSP itself suffers an incident. Those questions remain relevant when software carries out more of the work.

Yet they do not guarantee the survival of today’s provider. Another company can accept the same responsibilities while using fewer technicians. Customers can also divide the work between a platform, an internal manager and a specialist recovery service.

Human involvement protects the need for accountable service. It does not protect a particular supplier’s prices or staffing structure.

British buyers could get better support and a worse contract

A cheaper ticket is valuable only if the underlying problem is resolved.

Imagine a hypothetical 50-person company reviewing an AI-enabled support offer. Its management team sees a lower monthly price and a promise of immediate responses. The useful comparison includes the work that remains outside the automated service. Who handles a failed laptop? Who restores an accidentally deleted record? Who challenges an application supplier when each dashboard says the other system is at fault?

If the office manager inherits those jobs, the company has transferred work internally. The saving on the supplier’s invoice needs to be weighed against that person’s time and the business’s ability to recover.

Automated closure rates need similar scrutiny. A closed ticket can represent a successful fix, an abandoned conversation or a problem that returns the following morning. Buyers need evidence that distinguishes those outcomes.

I would start a renewal discussion with a sample of real incidents. Ask the provider to show which steps were automated, where a person intervened and whether the user needed to ask again. Include an awkward case that crossed supplier boundaries. A demonstration built around those records tells a buyer more than a broad claim about AI adoption.

Access deserves equal attention. An assistant drafting a reply has a different role from an agent permitted to alter accounts or devices. The contract and service design need to make that distinction visible. Customers need an agreed route to stop automated actions and reach a person with authority to take over.

None of this means demanding that the provider pass through every labour saving. Retaining enough staff to handle a difficult incident costs money even during a quiet week. The buyer’s objective is to understand what capacity remains available and whether the fee supports it.

Small providers are not automatically the losers

Size alone is a poor predictor of which MSPs will adapt.

The government study classified 65% of the 10,843 providers for which it could allocate a firm size as micro businesses. That is a large part of the supplier population, but it says nothing by itself about technical ability or customer trust. The report’s size analysis cannot support a conclusion that small providers lack specialist knowledge.

A small firm with well-documented customers and a narrow service offering may find it easier to automate than a larger business carrying incompatible systems from successive acquisitions. Conversely, a provider dependent on one technician’s memory has work to do before delegating changes to software.

Access to shared automation could reduce the advantage of employing a large support team. It could also increase dependence on the platform supplier. Both effects deserve attention before forecasting that only large MSPs will survive.

Britain’s wider challenge is the training route into technical work. If providers need fewer people to handle straightforward tickets, they must decide how new entrants acquire the experience needed for difficult incidents. Reading an AI-generated explanation is not the same as diagnosing why an apparently reasonable fix failed.

The government’s January 2026 assessment of AI and the UK labour market distinguishes success on precisely specified digital tasks from the ambiguity of real work. It also records uncertainty about how task-level gains translate into wider economic effects.

For MSP owners, a practical response is to redesign junior work around supervised investigations and recovery exercises. Let automation remove repetitive administration, while preserving opportunities to inspect failures and explain decisions. Otherwise, a business can improve this year’s delivery costs while weakening the team that will handle its future exceptions.

That is a more immediate skills question than predicting a national total of jobs lost.

Watch the renewal, the exception and the exit

The next convincing evidence will come from ordinary service records and contract renewals.

Watch whether reported automation rates survive a consistent definition of resolution. Include reopened incidents and the time spent repairing an automated mistake. A system that handles a narrow class of requests reliably may be more valuable than one claiming wider coverage with poorly recorded exceptions.

Then watch what customers receive at renewal. Lower prices would be one sign that competition is passing savings through. Broader coverage or demonstrably better recovery could be another. Growing margins alone would tell us that providers had captured the benefit, without settling whether their customers were better served.

Exit is equally revealing. A customer needs to know whether another provider can use its documentation and continue the automated workflows. If leaving means rebuilding the service from scratch, the new model may exchange dependence on a technician for dependence on a platform.

My expectation is a managed-services market in which routine work consumes less human time and providers compete harder over the remaining responsibility. That can support stronger businesses and better services. It can also leave a provider with an expensive sales operation wrapped around work another company increasingly controls.

AI could make the labour-heavy MSP model obsolete. A successful successor will have to prove its value when automation reaches its limit. The customer will still need a number to call, and the person answering will need more authority than the ability to open another ticket.