Listen to the article
AI tools could help organisations handle sprawling application estates by sorting tickets, spotting risks and supporting incident response. Their value depends on sound data, careful trials and people retaining control of consequential decisions.
As organisations add cloud platforms, integrations and newer technologies to already sprawling systems, keeping business applications reliable has become a more demanding operational task. Traditional application management has benefited from monitoring, robotic process automation and offshore delivery, but these measures may not be enough when systems are numerous, interconnected and constantly changing. Boston Consulting Group’s research found that organisations managed more than 1,000 applications on average in 2024, with only a minority connected for data exchange, highlighting the integration and governance burden behind large portfolios.
AI-driven Application Management Services are presented as a way to extend, rather than discard, existing operating models. Software agents can sift through logs, rank incoming tickets and flag unusual behaviour, while predictive analytics can help identify patterns that may precede a disruption. PagerDuty’s account of AIOps describes similar techniques for correlating events across monitoring systems, reducing alert noise and supporting earlier intervention. Their effectiveness, however, depends on usable data, integration with existing tools and careful operational planning.
The model also relies on people. Human specialists remain responsible for checking machine-generated recommendations and making decisions where the consequences are significant. A review of AIOps research published on arXiv describes a field spanning detection, prediction, root-cause analysis and remediation, while noting that approaches and evaluation methods are not yet standardised. That makes oversight and testing important elements of deployment, rather than optional safeguards.
Moving from individual automated tasks to coordinated systems is a further step. Multiple agents may be assigned different responsibilities across a service tower, while knowledge-retrieval tools help staff apply internal expertise during incident resolution. The broader AIOps principles outlined by PagerDuty offer context for this approach, although they do not establish that a particular application-management deployment will deliver a defined result.
A staged introduction is advised in place of an abrupt overhaul. Organisations can first identify repetitive work, then trial agents in tightly defined cases such as ticket sorting or retrieving relevant guidance. If pilots perform reliably, AI can be extended across a service area under human supervision and later incorporated into routine delivery. BCG’s analysis of enterprise application strategy reinforces the need to make such decisions in light of business requirements, integration demands, security and long-term costs, rather than assuming one arrangement will suit every organisation.
Technical controls need to develop alongside the operating model. The supplied article identifies unreliable AI outputs and the protection of sensitive information as risks, suggesting that organisations consider where different data and models are handled. The arXiv review’s account of unresolved challenges in AIOps also underlines the need to evaluate systems carefully. Meanwhile, the accessible abstract for Gartner’s 2026 enterprise applications study confirms that application portfolios are being benchmarked, but does not disclose figures that would substantiate any specific estate size.
The commercial gains cited by CGI should be treated as reported experience, not a universal forecast. The company says advanced implementations it has supported have achieved productivity improvements of 25 to 40 per cent, reductions in mean time to repair of up to 35 per cent, and operating-cost savings of 15 to 25 per cent at service-tower level. Those claims are consistent with the potential benefits associated with less manual work and faster incident handling, but the AIOps material from PagerDuty and the academic review do not independently verify those figures. Results will depend on the systems, data and processes involved.
The change is therefore organisational as well as technical. Staff need to adapt to working alongside automated agents, with responsibilities and escalation routes clearly defined. F5’s historical discussion of enterprise application estates illustrates how diverse hosting arrangements and ongoing integration, security and maintenance work can add to operational demands, though its 2016 figures are not a current benchmark. As application environments continue to evolve, AI may help organisations manage that workload more proactively, provided it is introduced incrementally and remains subject to meaningful human control.
Source Reference Map
Inspired by headline at: [1]
Sources by paragraph:
Source: Fuse Wire Services


