Hello. This is Alex Bakker, standing in for Stanton Jones with what’s important in the IT and business services industry this week.
If someone forwarded you this briefing, consider subscribing here.
AI, Managed Services and the Backlog Paradox
How will AI impact the managed services market? The data presents a paradox.
On the one hand, many organizations believe AI will reduce headcount. Executives increasingly expect that AI-assisted development and automation will allow fewer people to complete the same amount of work. At face value, this suggests managed service providers (MSPs) should face structural pressure.
On the other hand, the market is behaving very differently. Deal durations are increasing and total contract values are growing. Enterprises are signing large, long-term transformation agreements rather than focusing on contracts that preserve maximum flexibility.
This tension becomes even more interesting in the context of what is happening in the software market. Over the past week, SaaS valuations have experienced volatility driven by concerns that AI may compress future growth. Investors worry that enterprises will gain pricing leverage over vendors, use AI-assisted development to build more internally, or simply that AI agents will break seat-based licensing models. In other words, AI is challenging the traditional software growth model.
So why are we seeing longer managed services commitments with larger contract values?
Part of the answer lies in enterprise priorities. In our 2026 Budget and Spending Study, technology modernization emerges as a top IT priority alongside productivity improvement and cost savings. Large transformation programs require long-time horizons to generate returns. The structure of current contract activity reflects that economic reality. Big modernization efforts are inherently multi-year, and providers often work with their clients to front-load the savings while spreading out the costs of transformation over more time. In essence, providers offer both transformation services and the financing model to drive those long-term outcomes.
So, while enterprise leaders may expect productivity gains from AI, they are simultaneously committing to long-duration transformation programs with their providers.
Our research also shows that organizations do plan to shift away from traditional staff augmentation and FTE-denominated contracts at the margins. But that shift does not equate to a retreat from managed services. In fact, it often coincides with increased adoption of outcome-oriented, multi-year agreements.
At the same time, when we look at AI use cases, we see that once projects reach production, they tend to deliver improvements on the metrics enterprises choose to measure—whether compliance, efficiency, cost or performance. The pattern is consistent: production AI use cases create value on their intended performance metrics.
That finding is important. It suggests that, while experimentation may be uneven, production AI is not broadly failing to deliver returns. Instead, it is reinforcing confidence in transformation programs.
This situation is best described by the Jevons paradox, which states that technology that improves efficiency of a resource tends to increase demand for that resource.
AI improves developer productivity, which introduces slack in the system. Historically, when project effort estimates were wrong, as they often are, teams had little ability to recover. Deadlines slipped and confidence eroded. Projects were shelved or never funded in the first place because estimation risk was too high. If productivity increases but estimation practices remain unchanged, something subtle happens. Developers gain more capacity to absorb estimation error. In aggregate, more projects that are estimated to take six months actually take six months. Delivery performance improves, not because forecasting becomes perfect, but because AI-driven productivity gains dampen variability in aggregate estimation accuracy.
As delivery reliability improves, leadership confidence increases. Projects that previously seemed too risky to fund become tractable. Efforts that were economically marginal (where ROI was uncertain due to delivery risk) begin to clear hurdle rates.
This is where the backlog becomes central.
Most enterprises are sitting on enormous modernization backlogs: thousands of applications, legacy infrastructure, fragmented data estates, governance gaps, integration complexity and accumulated technical debt. The reason this backlog exists is not a shortage of ideas. It is economic friction. Until now, the economics of “rip and replace” were always unattractive. There were never enough staff, estimates were unreliable, ROI was difficult to model, pricing risk was high, and the outcomes of new software were often materially similar to the old software in all the easily modellable dimensions. All the costs of not doing it were opportunity costs and foregone capability. Over a finite period of time, those opportunity costs are manageable, but in the long run, the economic cost of legacy systems compounds.
In other words, much of the backlog accumulated because the delivery economics did not support addressing it, and the costs incurred were theoretical rather than immediate.
If AI reduces delivery risk – by increasing productivity and improving adherence to timelines – the economic case for tackling legacy complexity changes. Projects that were once too risky or too marginal to justify investment begin to make sense.
This is the Jevons dynamic in IT. Improving the efficiency of software development does not necessarily reduce total demand for development. It increases it. When the effective cost and risk of delivery go down, more projects become economically viable.
And here is the critical distinction between managed services and software.
In software, AI may compress margins or growth by increasing buyer leverage or enabling substitution. Enterprises may negotiate harder or build selectively. In managed services, AI may also introduce price pressure at the unit level. Labor components may decline. Contracts may shift toward outcomes rather than FTEs. Providers are not immune to economic pressure.
But the aggregate demand environment is different.
Current large transformation contracts typically address only subsets of applications or infrastructure—not entire estates. Even with productivity gains, there is no immediate risk of “running out” of work. Instead, AI has the potential to unlock demand that was previously uneconomical to pursue.
Long-duration, large-scale deals reflect the scale of the problem set, not necessarily an absence of pricing pressure. They signal that enterprises are committing to transformation programs that require sustained execution capability. AI may change how that work is delivered – and it may change the economics of labor within those agreements – but it does not eliminate the need for partners capable of orchestrating complex change across legacy environments.
So, do managed service providers win?
They may not be immune to margin compression, but they are operating in a demand environment defined by accumulated complexity, modernization imperatives and newly tractable economics. AI increases productivity. Productivity reduces delivery risk. Reduced risk expands the economically addressable backlog.
Enterprises may need fewer people per unit of work, but they have more economically viable work to do. This backlog may just give the services industry time to transform itself and its delivery with AI.
About the author
