Updated 17th Feb 2026
When Vodafone rolled out Microsoft 365 Copilot to trial users they saw quick wins as people shared what worked in their day to day. "We’ve been really impressed by the ingenuity of our teams using generative AI,” said Scott Petty, Chief Technology Officer. The company has now expanded Copilot to tens of thousands of employees after early productivity gains.
That kind of grassroots momentum is powerful, but it sticks only when a central playbook sets security, compliance and shared tooling. Let local teams surface practical use cases, then use a steering group to standardise, fund and safely scale.
This mix moves organisations from experiments to measurable impact and is needed because most companies still sit in pilots rather than at scale, with only about a third reporting enterprise‑level scaling and 39% reporting EBIT impact according to recent research from McKinsey.
A top-down approach provides strategic alignment, central control and ensures compliance and security. This method benefits from senior management backing, which helps align AI projects with broader organisational goals and ensures the necessary resources are allocated efficiently. However, it can also introduce bureaucracy, slow decision-making and potential resistance from departments that feel dictated to by senior leadership.
Conversely, a bottom-up approach fosters grassroots innovation and local engagement. This method empowers employees at all levels to identify and solve problems using AI, driving higher engagement and adoption rates. However, without central oversight, it can lead to fragmented efforts and inconsistency in AI implementation across the organisation.
By combining both approaches, organisations can leverage the strategic benefits of top-down control while fostering the innovation and engagement driven by bottom-up initiatives. This hybrid approach ensures a cohesive strategy aligned with organisational goals, while also harnessing the collective creativity and insights from employees across all levels.
For example, EY has rolled out Microsoft 365 Copilot at scale and then equipped teams to build their own agents with Copilot Studio. “With Copilot Studio, our people are innovating with agents that help them get more work done, more accurately, and more quickly,” notes Roger Park, Global Business Enablement, AI and Innovation Leader.
Senior management backing: For AI initiatives to succeed, they must have the support and endorsement of senior management. Leadership plays a crucial role in aligning AI projects with the broader strategic goals of the organisation and ensuring the necessary resources are allocated.
Steering group: A dedicated steering group is essential for reviewing AI use cases, maintaining a central use-case register, and setting the strategy for AI tools and products. This group ensures responsible AI adoption, focusing on ethical considerations, regulatory compliance and data security.
Security and compliance: Central control helps ensure that AI initiatives adhere to regulatory requirements and maintain high standards of data security. This is particularly important in industries with stringent compliance needs.
Training and support: Providing comprehensive training and support to employees is vital. Centralised training programmes help upskill the workforce, foster a culture of continuous learning, and enable employees to use AI tools effectively. In 2026, this also means equipping teams to work with AI agents and updated workflows through focused, role-based training, clear guidance on risk tiering and guardrails, and practical exercises for human-in-the-loop supervision.
User groups: Local user groups are key to identifying and implementing AI solutions tailored to specific departmental needs. By empowering employees at all levels to suggest and develop AI use cases, organisations can drive innovation and ensure AI tools address practical, day-to-day challenges.
Training and tools: Equipping local teams with the necessary training and tools to experiment with AI solutions fosters a sense of ownership and encourages widespread adoption. This grassroots approach ensures that AI initiatives are user-driven and relevant to specific business areas.
Feedback loop: Establishing a robust feedback mechanism allows successful use cases and insights from local groups to be shared centrally. This facilitates the refinement of strategies and the replication of successful projects across the organisation. Moreover, once the first few use cases are implemented and knowledge shared, the flywheel effect sees users finding and solving an ever-increasing number of use cases.
The “top-down, bottom-up” approach is particularly effective for leveraging both the big head and the long tail of AI, a key characteristic of successful AI adoption.
Early pilots showed the art of the possible. Moderna’s Dose ID GPT, part of its 2024 OpenAI partnership, helped clinical teams synthesise large datasets to support dose selection, alongside hundreds of internal GPTs. By 2025–2026, big-head projects look more like agentic systems at scale. JPMorgan has begun deploying AI agents to handle multi-step employee tasks as it moves towards a “fully AI-connected enterprise”; a CNBC demo showed an investment banking deck created in about 30 seconds. JPMorgan’s published AI research also points to systematic investment in agents, planning and knowledge systems, and document processing, signalling an enterprise-mature programme rather than isolated pilots.
Incremental improvements driven by many small AI use cases can deliver significant value. Beyond everyday tasks like email triage or expenses, the long tail now includes agent-led micro-workflows such as drafting routine summaries, updating records, tagging documents, preparing packs, and routing approvals. With clear guardrails and basic telemetry (adoption, quality, outcomes), these small wins compound into material efficiency and cost savings. Microsoft’s research found that 70% of Copilot users reported feeling more productive, and the focus now is turning that time saved into measurable results.
A successful AI strategy requires integrating both major projects and incremental improvements. Big-head projects showcase AI’s transformative potential, while long-tail initiatives ensure continuous innovation and adaptability. The top-down, bottom-up approach ensures that all types of use cases are identified and implemented through engagement with the business via user groups, while the steering group ensures that the larger projects receive the right level of funding and support.
Adopting a “top-down, bottom-up” approach helps enterprises leverage the full spectrum of AI benefits. Centralised control ensures strategic alignment, security and compliance, while local engagement fosters innovation and user-driven solutions. This balanced strategy is particularly effective for uncovering and delivering value from long-tail use cases, driving significant cumulative impact across the organisation.
By combining strategic oversight with grassroots innovation, enterprises can navigate the complex landscape of AI adoption and achieve comprehensive business transformation. This approach not only maximises the potential of AI but also positions organisations for sustained success in an increasingly competitive market.
If you want to understand where AI agents can deliver value across your organisation, we can help you identify high-impact workflows and show what those agents look like in Microsoft 365.