What Determines Performance Is Shifting
Enterprise technology capability has become less differentiated. Cloud infrastructure, AI tools, DevOps practices, and observability platforms are broadly accessible. The gap between the technology available to a well-resourced enterprise and the technology available to a similarly resourced competitor has narrowed significantly over the past five years.
The differentiator has moved. Organisations that outperform their peers in technology-enabled business outcomes are doing so not primarily through technology choice but through operating model: the structures, processes, governance, and talent strategies that determine how effectively the available technology is put to work.
This shift is visible in the nature of the transformation work that enterprise technology leaders are prioritising entering 2025. The conversations are less frequently about technology selection and more frequently about capability development, governance design, and operating model change. The technology questions have answers that are widely accessible; the operating model questions require organisation-specific judgment and deliberate design.
The First Trend: AI Integration Moves from Experiment to Operating Model
2024 was the year of AI pilots for most enterprises. The AI coding assistant was deployed to a subset of engineering teams. The AI-powered customer service capability was tested in a limited deployment. The internal knowledge management AI was run as a pilot for a specific business unit. The common characteristic of these deployments was that they were experiments sitting alongside the existing operating model rather than integrated into it.
2025 is the year that the AI integration question becomes an operating model question. Which workflows have AI integrated as a standard component rather than as an optional tool? What does the human-AI collaboration model look like for the knowledge work that constitutes the majority of enterprise white-collar productivity? How are quality standards maintained when AI is generating a significant portion of the output that knowledge workers are accountable for?
The organisations that answer these questions with deliberate operating model design will extract materially more value from their AI investments than the organisations that leave the integration to individual employee choice. The employee who decides independently how to use the AI coding assistant will use it in ways that may or may not align with the team’s quality and collaboration standards. The team whose workflow has been deliberately adapted to integrate the AI assistant produces more consistent outcomes.
The operating model work for AI integration requires the same disciplines as any major process change: workflow analysis, process redesign, role definition, training, and governance. It requires these disciplines applied to AI specifically, which means understanding what AI does reliably and what it does poorly well enough to design workflows that leverage the former and compensate for the latter.
The Second Trend: Platform Engineering Consolidates Around Proven Patterns
The platform engineering movement has generated significant investment and significant experimentation over the past three years. In 2025, the experimentation phase consolidates around the patterns that have demonstrated value and the organisations that have invested without clear patterns face the choice between doubling down on a defined direction or accepting that their current platform engineering investment is not performing.
The patterns that have demonstrated value are identifiable from the leading enterprise platform engineering deployments. The internal developer portal as the central interface through which development teams access platform capabilities, reducing the cognitive overhead of navigating multiple systems and tools. The golden path as the optimised, supported, opinionated path for the most common development use cases, coexisting with the flexibility to deviate for cases that require it. The platform team as product team, with a product management capability that drives the platform roadmap from developer need rather than infrastructure preference.
The organisations whose platform engineering programmes are delivering on these patterns will accelerate in 2025. Those that have invested in platform tooling without the product management discipline that makes the tooling useful to development teams will face a difficult assessment of whether the investment is recoverable or should be restructured.
The Third Trend: Governance for AI Systems Becomes Operationally Required
The EU AI Act’s progressive entry into force, combined with increasing board attention to AI risk following high-profile AI failures and regulatory guidance across financial services and healthcare, makes AI governance an operational requirement for a growing population of enterprises in 2025.
The AI governance that was advisory in 2023 and preparatory in 2024 becomes operational in 2025, meaning it needs to be embedded in the processes through which AI systems are developed, procured, and deployed rather than existing as a separate governance layer applied to already-running systems.
The operating model implication is that the product development lifecycle needs an AI-specific gate for AI-involved products, the procurement process needs AI-specific due diligence for AI vendor relationships, and the risk management process needs AI-specific incident classification and escalation pathways. None of these changes requires building new governance functions from scratch; they require extending existing operating model components with AI-specific design.
The Fourth Trend: Cost Discipline Becomes an Engineering Norm
The FinOps programmes that many enterprises initiated in 2022 and 2023 in response to cloud cost escalation have produced mixed results. The organisations that treated FinOps as a tool purchase or a finance function have not achieved the ongoing cost discipline that the investment promised. The organisations that treated FinOps as a cultural and operating model change have.
In 2025, cost discipline as an engineering norm is increasingly a competitive requirement rather than an optional discipline. Cloud costs that continue to grow faster than the business growth they support are a sustainability risk for technology operating budgets. The engineering culture that considers cost as a normal dimension of design decisions, alongside performance, security, and reliability, produces different cost trajectories than the engineering culture that treats cost as a finance concern.
Building this norm requires changes to how engineering decisions are made, how performance is assessed, and how the platforms that engineers use expose cost information. The operating model design work for engineering cost discipline is practical and achievable; it requires deliberate intent rather than technical breakthrough.
The Operating Model Capability That Underlies All Four
The four trends described above have a common requirement: the organisational capability to change how work is done, rather than just what tools are used to do it. This capability has been described in transformation literature as change leadership, operating model design, or organisational effectiveness, depending on the discipline. The organisations that execute operating model transformation consistently well have it; those that have excellent technology selection discipline and limited operating model execution capability do not.
Building this capability is not a technology investment. It is a leadership investment, a talent investment, and a discipline investment in the systematic management of how the organisation operates alongside the management of what the organisation does.
The organisations that enter 2025 with this capability already developed are better positioned to execute on all four of the trends described here. The ones building it in parallel with executing on the trends face a harder programme design challenge but not an insurmountable one.
The technology is not the difference. The operating model is.