A retailer signed a five-year contract for application maintenance services in 2024. The vendor staffed a team of 40 engineers across different support levels and billed the retailer through a monthly FTE charge. A year later, the vendor introduced AI tools for ticket triage, log analysis, and code fixes. Routine tickets could now be resolved automatically or in minutes. 

The vendor reduced the effective team to 28 people, reassigned the remaining engineers, and continued billing for 40. Service levels remained stable, so the retailer had little reason to question the arrangement. The savings were real, but they were not reaching the customer. 

This is the issue AI is creating for many application maintenance contracts. Tools that classify incidents, suggest fixes, generate code, and answer user questions reduce the amount of human effort required to support applications. Yet contracts based on tickets or headcount do not automatically account for that reduction. 

Unless buyers revisit the commercial model, they may end up paying for work that AI has already made cheaper to deliver. The vendor captures the productivity gain while the buyer continues paying under the old terms. 

How Application Maintenance and Support Services Are Usually Priced 

Most application maintenance and support agreements follow one of a few common pricing models. 

FTE-based pricing. The buyer pays for a fixed number of people at agreed rates. The vendor commits to staffing levels and service levels. This model is common in large, long-term engagements. 

Ticket-based pricing. The buyer pays for each ticket resolved, often at different rates based on priority or complexity. Some contracts also use volume bands with minimum or maximum charges. 

Fixed-price managed service. The buyer pays a fixed monthly fee for a defined scope of services. The agreement may also include service levels and volume limits. 

Time and materials. The buyer pays for the actual hours worked. This model is often used for enhancements and project work that sits alongside ongoing support. 

These models worked well when support effort generally increased with ticket volume. AI is changing that relationship. 

The impact also varies by support level. Level 1 work, including password resets, access requests, and how-to questions, is generally the easiest to automate. Level 2 work, such as known defects and configuration issues, can benefit from AI-generated fixes and faster diagnosis. Level 3 work, including complex code changes and architecture problems, still requires experienced engineers, although AI can speed up investigation and troubleshooting. 

This means a contract that treats every support level in roughly the same way can become increasingly misaligned as automation grows. 

What AI Is Changing in Application Support 

AI affects nearly every layer of application support: 

Each of these reduces the human effort required for a given volume of work, often by a wide margin. Enterprise software is also embedding more of this capability directly. Gartner predicts that 40% of enterprise applications will include task-specific AI agents by the end of 2026, up from less than 5% in 2025. 

Why Current Contracts Let Vendors Keep the Savings 

The problem is built into many existing pricing models. 

With FTE-based pricing, the buyer pays for people rather than outcomes. If AI allows a vendor to deliver the same service with fewer engineers, the vendor can reduce its delivery costs without necessarily reducing the invoice. The buyer may continue to see the same service levels while having limited visibility into how much effort the vendor now needs to provide them. 

Ticket-based pricing creates a different problem. Ticket deflection does reduce the number of tickets and therefore can lower the bill. But the tickets that remain are often priced at rates established when every ticket required significant manual effort. If AI helps resolve those tickets faster, the vendor's cost falls while the rate may remain unchanged. 

With fixed-price contracts, productivity gains generally improve the vendor's margin unless the agreement includes mechanisms for reviewing the price as productivity improves. 

The market is beginning to respond. HFS Research has reported that companies are increasingly renegotiating IT services contracts within 24 months of signing rather than waiting until renewal. Application and infrastructure management agreements are among the areas facing this pressure. Contract discussions are also moving away from FTEs, rate cards, and offshore staffing ratios toward productivity commitments, outcome-based pricing, and gain-sharing. 

Better Commercial Models for Application Maintenance Services 

Several pricing approaches can help buyers share in the productivity gains created by AI. 

Productivity Commitments 

Under this model, the vendor commits to reducing the cost or effort required to deliver the same scope of services each year. For example, a contract could include a 10% to 15% annual price reduction supported by the vendor's AI roadmap.   

The commitment should be written into the contract and reviewed regularly against actual productivity improvements. 

Outcome-Based Pricing 

Here, pricing is linked to measurable service or business outcomes rather than the number of people assigned to the account. 

Possible measures include application availability, mean time to resolve incidents, user satisfaction, and the reduction in recurring defects. 

This changes the vendor's incentive. Instead of earning more by assigning more people, the vendor has a reason to improve the application and deliver better results. 

IDC expects this shift to continue. Its projections indicate that 30% of service provider contracts could be outcome-based by 2029 as agentic AI changes how services are delivered. 

Whatever pricing model a buyer chooses, several clauses can help keep the commercial relationship aligned as AI capabilities evolve. 

Ownership deserves particular attention. 

Knowledge bases and automation scripts created during an engagement often become an important part of the support operation. If those assets remain with the vendor, moving to another provider can become more difficult and expensive. 

What to Expect From an Application Maintenance Company 

When evaluating an application maintenance company, buyers should ask how the provider plans to share the benefits of AI. 

A provider that can answer these questions with specific data and commercial proposals is more likely to be prepared for an AI-driven support model. 

A provider that focuses only on service-level commitments may not be addressing the underlying change in delivery costs. 

The Human Side of AI-Enabled Support 

AI does not eliminate the need for skilled people in application support. 

Complex incidents, business-critical changes, and root cause analysis still require experienced engineers who understand the applications and the business processes behind them. 

What changes is the mix of work. Fewer people may be needed for routine tickets, while more engineering effort can go toward prevention, automation, and continuous improvement. 

Good application support and maintenance services should reflect this change. Providers need engineers who understand the business, manage automation tools, and work with application owners to address recurring problems. 

For buyers, the goal should not simply be to find a provider with the lowest headcount or the cheapest rates. The more important question is whether the provider can use AI to improve the service while sharing the resulting productivity gains. 

Measuring Value After the Change 

After renegotiating the contract, buyers should continue tracking both cost savings and service quality.