Jevons Paradox For MSPs: Why AI Is About to Make IT Services Bigger, Not Smaller

Everyone in MSP-land thinks AI is coming for tech support jobs. The opposite is happening.

• AI commoditises the L1 work (password resets, printer tickets, onboarding). What doesn’t compress is planning the agent’s policy correctly upfront, supervising and verifying what it does, and being the accountable party when something touches security or performance.
• MSPs who build a system to absorb that low-value volume can handle meaningfully more ticket volume on the same headcount, while moving a large share of senior engineering hours into planning, supervision, verification and accountability, most of it recurring rather than one-time project work.
• SuperIT is built for exactly this: it fully automates what it safely can and significantly assists on everything else, with engineers planning, executing and verifying alongside it rather than starting from scratch.
• The strategic question isn’t whether your MSP or IT team survives AI. It’s whether you’re positioned to absorb the demand it creates, or stuck competing on price against the MSPs who got there first.

Our team ran MSP businesses before SuperIT, and the belief we hear most often in this industry, that AI is going to displace tech support jobs, doesn’t match what actually made a service desk good. The thing we remember most was the moments where engineers delivered exceptional service: the best engineers on our desk were superheroes to the businesses they served, the person who knew the customer’s IT environment inside out, the best communicator, the one who unblocked a user banging their head against the desk over something that looked impossible.

AI gives every engineer on the team those same instincts on tap, so that level of service becomes consistent across the desk instead of occasional, and the work expands to match the demand.

Dan Shipper’s essay After Automation (May 2026) crystallised the underlying mechanic. His argument: AI commoditises whatever can be made explicit enough to train on, which collapses the value of default output and creates new demand for what’s different. Automate the predictable middle of the work, and what’s left concentrates at the edges, in the judgment calls only a person can make.

For IT services and MSP businesses, it is the textbook case of what Shipper is describing, and we don’t think most people have seen it clearly yet.

Here is our prediction of what is going to happen over the next three years.

The work that gets cheap

Walk through a service desk on any Tuesday morning. What’s in the queue?

Password resets. Account unlocks. Printer issues. Performance problems. New starter onboardings. Teams permissions. M365 licence assignments. Distribution list edits. Outlook profile rebuilds. The same collection of ticket types, over and over, eating fifteen to thirty minutes a pop, multiplied across every help desk in the country.

This is the “yesterday’s competence” Shipper is talking about. It has been documented in tens of millions of tickets, written up in tens of thousands of KB articles, and demonstrated by hundreds of thousands of engineers over two decades. The pattern is fully explicit. The cost of doing it has been falling for years, and AI finishes the job that scripting and runbooks started.

So the simple and intermediate ticket commoditises. The seat-based MRR margin that depends on humans doing this work compresses. MSPs that try to defend the old model on price will lose, because their competitors who don’t are going to be cheaper per ticket and faster per resolution by orders of magnitude.

The work that explodes

The highest-value work splits into two shapes, and we think a lot of MSPs are about to conflate them.

One shape is project work: a security uplift, a cloud migration, an AI rollout. Real, billable, and valuable. It’s also one-time. You scope it, deliver it, invoice it, and go looking for the next one.

The other shape doesn’t look like a project at all. It’s planning the policy an agent runs under so it behaves like your firm rather than a stranger’s, supervising what it actually does day after day, verifying that a resolution is genuinely correct rather than just closed, and being the party accountable when something touches security or performance, because no client, and no regulator, accepts “the AI did it” as an answer.

That work doesn’t have a start and an end the way a project does. It runs for as long as the agent runs, which is every day the client is your client. It’s currently rare inside most MSPs, not because it isn’t valuable, but because the service desk queue eats every hour that could go toward it.

Two different things are happening to service-desk labour at the same time, and it’s worth being precise about which is which. Execution, the day-to-day work of L1 and L2 support, is compressing. Planning and accountability, the upfront design and the ongoing supervision and verification of what the agent does, is doing the opposite. It’s becoming more valuable, and unlike a project, it renews every month instead of closing out as a line item.

Preparation for the new dynamic

You need to build a system that absorbs the high volume from the bottom of the stack so the senior engineers you already have can do the top of the stack. Free the heroes from the queue and the rest takes care of itself.

This is Shipper’s “human sandwich.” A human frames the work. The system collapses the predictable middle. A human judges, owns, and extends the result. Apply that pattern to a service desk and it looks like this:

• Management and senior engineering frames, sets policy, decides what the system is allowed to do, defines what good resolution looks like for the client.
• An autonomous agent collapses the predictable middle: the initial ticket, the standard service request, the diagnostic-and-fix loop you’ve seen a thousand times.
• A human extends: supervises what the agent actually did, verifies the resolution was genuinely correct rather than just closed, and stays the accountable party if anything touches security or performance. Some of that freed time also goes to projects and the boardroom, but the supervision and verification is the part that recurs every day.

The teams who run this pattern at scale will handle meaningfully more ticket volume with the same headcount, while moving a large share of their senior engineering hours into planning, supervision, verification and accountability, with project work sitting on top as the occasional, one-time piece.

What this means in practice

If you’re running an IT team or MSP, the next three years come down to three questions.

One: do you have a system that can absorb the bottom of the stack?

Not a triage tool, a chat tool, or a summary feature that hands a still-unresolved ticket to a human who then has to read the summary: a system that takes the predictable ticket in production, with safety controls and auditability, and either resolves it end to end or gets it most of the way there before handing it to an engineer with the diagnosis already done.

That’s the system we’re building. It fully automates what it safely can, and for everything else it significantly assists: the plan, execute, verify pattern described above, the human sandwich, is the core of the product, not a fallback. The main driver is drastically increased productivity across the whole queue, not a headline resolution number.

Two: have you built a way to charge for planning, supervision and accountability, not only for projects?

Project work, an AI rollout, a security uplift, a migration, is real and worth pursuing, but it’s one-time: you scope it, deliver it, invoice it, and go looking for the next one. The work that actually compounds is different. It’s being paid, every month, because you’re the party who set the agent’s policy correctly, who supervises and verifies what it does, and who carries the accountability if something touches security or performance. Most MSPs don’t have a name for that today, let alone a price for it. Build the pipeline conversation for both: the one-time projects your account managers should be scoping over the next six months, and the recurring accountability line that should sit on every contract from here on.

Three: are you positioning the business for the work that explodes, or the work that gets cheap?

This is the strategy question. Every IT team we know is positioned for the work that gets cheap, because that’s the work they’ve been selling for fifteen years. The MSPs who win the next decade will be the ones who reposition: same client base, completely different story about what they do for them, completely different margin structure.

You don’t have to commit to the new positioning today. But you have to know it’s coming, and you have to be moving toward it. The MSPs who wake up in 2029 still selling per-seat L1 support against competitors who’ve absorbed the bottom of the stack are going to have a bad year.

The point

Shipper’s piece argues that AI doesn’t replace experts, it creates more situations where expert judgment is needed. The MSP industry is the cleanest case study of that argument anywhere in the economy. The L1 ticket gets cheap. Planning an agent’s policy, supervising what it does, verifying its resolutions, and carrying accountability for its security and performance, all of that explodes, and most of it renews every month instead of closing out as a line item. The one-time work explodes too: the architecture, the compliance programs, the M&A integration, the AI rollout itself. But it’s the recurring accountability that compounds.

The question isn’t whether your MSP or IT team survives AI. The question is whether you’re set up to absorb the demand that AI is about to create, or whether you’re going to compete on price for shrinking margins while the MSPs who got the math right take your best clients.

We built SuperIT to give your IT team superpowers, not by replacing your engineers but by freeing them. Same headcount, more throughput, and the senior hours move from tedious support to planning the agent’s policy, supervising and verifying its work, and carrying accountability, the work that actually grows the client’s business, month after month.

The MSPs who do this well will be doing fundamentally different work for the same clients at materially better margins. The MSPs who don’t will compete on price for shrinking margins.

The paradox is real. IT services is going to explode, and your chance to take advantage of it is now.

Want to see what this looks like against your own ticket queue? Book a demo and we’ll walk through the tickets that matter to you.