IT leaders embrace hybrid AI strategies amid rising challenges

IT leaders are increasing their use of artificial intelligence (AI) for business-critical operations, with a majority deploying AI models and turning to hybrid application strategies, according to findings from the F5 2025 State of Application Strategy Report.

The report, based on responses from global IT decision makers, reveals that 96 per cent of organisations have now deployed AI models, a substantial increase from a quarter in 2023. This points to a notable shift in approach, as leaders trust AI to perform functions ranging from traffic management to cost optimisation.

Nearly three-quarters of respondents (72 per cent) expressed intentions to use AI for optimising application performance. Additionally, 59 per cent indicated support for leveraging AI to assist with cost optimisation and to integrate security rules, enabling automatic mitigation of zero-day vulnerabilities.

The adoption of AI gateways—tools that connect applications to AI services—has also risen. Half of organisations presently use AI gateways, and another 40 per cent anticipate adopting them within the next year. The principal application of these gateways includes protecting and managing AI models (62 per cent), serving as central points of control (55 per cent), and preventing sensitive data leaks (55 per cent).

Lori MacVittie, Distinguished Engineer at F5, commented on the findings: "This year's SOAS Report shows that IT decision makers are becoming confident about embedding AI into ops. We are fast moving to a point where AI will be trusted to operate autonomously at the heart of an organisation, generating and deploying code that helps to cut costs, boost efficiency, and mitigate security problems. That is what we mean when we talk about AIOps, and it is now becoming a reality."

Despite heightened enthusiasm, the report highlights ongoing operational barriers. Security of AI models remains the top concern for organisations currently deploying such models. Operational readiness, in particular, is a challenge, with 60 per cent citing manual workflows as a hindrance and 54 per cent reporting skills shortages that complicate AI development efforts.

Budgetary constraints also persist. Forty-eight per cent identified the costs associated with building and operating AI workloads as problematic, up from 42 per cent last year.

Data practices continue evolving, with a higher proportion of organisations indicating that their data handling is not yet scalable (39 per cent, compared to 33 per cent in 2024). Trust in AI outputs, due to potential bias or erroneous results, is another issue, with 34 per cent expressing a lack of trust, compared to 27 per cent previously. However, there has been an improvement in perceived data quality, as 48 per cent reported concerns this year, down from 56 per cent the year before.

The increased integration of APIs also brings its own difficulties. Some 58 per cent of respondents noted APIs as a pain point, with certain organisations dedicating as much as half of their time to managing complex API configurations and coding languages. The most time-consuming tasks involve vendor APIs (31 per cent), custom scripting (29 per cent), and integrating with ticketing or management systems (23 per cent).

MacVittie observed, "Organisations need to focus on the simplification and standardisation of operations, including streamlining APIs, technologies, and tasks. They should also recognise that AI systems are themselves well-suited to handle complexity autonomously by generating and deploying policies or solving workflow issues. Operational simplicity is not just something on which AI is going to rely, but which it will itself help to deliver."

The report identifies a shift towards hybrid cloud architectures, with 94 per cent of organisations running applications across multiple environments, including public and private clouds, on-premises data centres, edge, and colocation facilities. This approach seeks to balance scalability, cost, and compliance needs.

Adaptability was cited as a major advantage of multi-cloud deployments, with 91 per cent of decision makers noting the ability to respond to changing business requirements, followed by improved application resiliency (68 per cent) and cost savings (59 per cent).

Most organisations now use a hybrid deployment approach for AI workloads as well, with 51 per cent maintaining models across both cloud and on-premises environments. An increased number of organisations have also repatriated one or more applications from public cloud to on-premises or colocation for reasons relating to cost, security, and predictability—79 per cent reported having done so, up significantly from 13 per cent four years prior.

The complexity of managing hybrid environments is not without its challenges. Inconsistent delivery policies were reported by 53 per cent, while fragmented security strategies were noted by 47 per cent of respondents.

Cindy Borovick, Director of Market and Competitive Intelligence at F5, said, "While spreading applications across different environments and cloud providers can bring challenges, the benefits of being cloud-agnostic are too great to ignore. It has never been clearer that the hybrid approach to app deployment is here to stay."

Data from Asia Pacific, China, and Japan (APCJ) reflects these global trends. Almost half (49 per cent) of APCJ organisations already employ AI gateways, with a further 46 per cent set to follow within the coming year. Their main objectives are protecting AI models (66 per cent), preventing sensitive data leaks (61 per cent), and monitoring AI application demand (61 per cent). Over half (53 per cent) struggle with data maturity, and 45 per cent are concerned about the cost of AI deployments. The hybrid model introduces additional complexity, with 79 per cent reporting inconsistent security policies, 59 per cent noting inconsistent delivery, and 16 per cent citing operational difficulties.

The report suggests a way forward through the creation of programmable IT environments that standardise and automate application delivery and security. By 2026, AI is anticipated to move beyond isolated tasks to managing entire IT processes. Platforms with natural language interfaces and programmable features are expected to streamline workflows, reducing the need for conventional management consoles.

Borovick added, "Flexibility and automation are no longer optional—they are critical for navigating complexity and driving transformation at scale. Organisations that establish programmable foundations will not only enhance AI's potential but create IT strategies capable of scaling, adapting, and delivering exceptional customer experiences in the modern age."