Spotting maintenance risks in SaaS platforms early
A software company with 2,400 employees runs more than 190 SaaS platforms; 260 maintenance tasks arise every day and risks remain hidden.
- Technology
- IT-Service-Management
Key metric
Time per maintenance case
2011min
Average assessment effort per maintenance case
At a glance
Starting point
The company develops and operates its own SaaS products as well as customer-specific platforms. In total, more than 190 platforms are running with different technology stacks, tenants and service levels. In IT operations, 28 employees are responsible for maintenance and stability. On average, 260 maintenance tasks arise every day, spread across the ticketing system, monitoring alerts and change requests. For each maintenance case, the teams spend an average of 20 minutes clarifying dependencies, risks and time windows. That adds up to around 85 hours a day, mostly tying up senior specialists. Information on tenants, contractual obligations and technical dependencies is scattered across several systems and documents.
- 190
- SaaS platforms
- 260
- maintenance cases per day
The challenge
Maintenance tasks range from security updates and configuration changes to infrastructure migrations. Each platform has its own tenants, SLAs and integrations with third-party systems. Whether maintenance can take place during the day, which tenants are affected and whether regulatory obligations are involved currently has to be checked manually. The IT teams jump between the ticketing system, monitoring, the contract repository and configuration documents. As a result, decisions are delayed, critical risks are recognized too late and maintenance is moved to night windows as a precaution. This increases costs and puts a strain on on-call duty. At the same time, the number of platforms keeps growing without additional senior capacity becoming available.
Typical symptoms in the organization
Why this remained unsolved
The problem had long been known but was accepted as an unavoidable complexity of running platforms. A classic rule set would have had to cover thousands of combinations of tenants, SLAs and integrations and would quickly have become outdated. Monitoring and the ticketing system had already been expanded but remained separate silos. A central tool for automatically assessing maintenance risks seemed too big an intervention in ongoing operations and was therefore postponed again and again, even though the effort for manual checks kept rising.
Our solution
An agentic workflow was introduced that bundles maintenance tasks from the ticketing system and monitoring, enriches them with platform data and contract information, and creates a structured risk assessment for each task. The workflow identifies affected tenants, relevant SLAs, technical dependencies and possible conflicts with other changes. The IT teams receive prioritized maintenance cases with suggestions for time windows, communication needs and escalation paths. Generative components condense scattered free text from tickets, runbooks and contract documents into clear assessments that can be reused in the change process. This turns many individual signals into a consistent decision framework for maintenance.
Where generative AI comes in
Generative AI analyzes free text in tickets, runbooks and contracts, recognizes patterns relating to risks, tenants and SLAs, and writes structured assessments for each maintenance task. It suggests time windows, communication templates and escalation paths and makes complex context usable faster, which reduces queries in the change process.
- Integration of the ticketing system, monitoring and platform inventory with consistent identification of the affected SaaS platforms and tenants.
- Automatic enrichment of maintenance cases with SLAs, contract information and known technical dependencies from configuration documents.
- Generative analysis of free-text descriptions in tickets and runbooks to produce structured risk assessments and recommendations.
- Prioritization and bundling of maintenance tasks by urgency, affected tenants and possible conflicts with other changes.
- Assessments delivered directly in the ticketing system, including suggested communication templates for customers and internal approvals.
How the maintenance workflow runs today
Collect signals
Tickets, monitoring alerts and planned changes are collected continuously, and platforms and tenants are assigned automatically.
Enrich context
The workflow adds SLAs, contract data, known dependencies and past incidents from the connected systems to each maintenance case.
Assess risks
Generative analysis summarizes free text, assesses risks and suggests suitable time windows and communication needs.
Prioritize cases
Maintenance tasks are prioritized and bundled by urgency, affected tenants and conflicts with other changes.
Document the decision
The chosen approach is documented traceably in the ticketing system, including the rationale and escalation path.
Measured: after six months in operation
Period: after six months in operation.
| Metric | Value | Basis |
|---|---|---|
| Time per maintenance caseKey metric | from 20 to 11 minutes per case | Time tracking before/aftermeasured |
| Senior time in the maintenance process | Reduced by around 70 hours per week | Time tracking before/afterestimated |
| Corrections needed to assessmentsQuality metric | Manual corrections needed in around 15 percent of cases | Sampleestimated |
Cost-effectiveness
The solution assesses risks and suggests time windows, but it does not replace technical approval by the people responsible. For new platform types, the workflow has to learn first, and the quality of results depends on well-maintained platform inventories and SLAs.
Integration and operation
Transferability
The pattern suits any situation where many maintenance or change tasks have to be coordinated across different tenants, SLAs and technical dependencies. Today, relevant information is often scattered across several systems and documents, and experienced specialists spend a lot of time on assessment. An agentic workflow bundles these signals, analyzes them generatively and delivers structured recommendations without replacing existing approval processes.
A similar case in your organization? Let's talk.
Briefly describe your process. We'll get back to you with an initial assessment of whether the pattern can be transferred and what it would realistically take.