How to move AI plans from paper to practice
On paper, AI strategy looks straightforward: set a vision, define governance, invest in the right foundations, prioritise use cases and launch pilots to unlock value. Alice Keal, senior delivery consultant at Cortex Reply explains why in practice, it rarely unfolds that neatly.
Most organisations are shaping strategy while experimenting at pace. What looks coherent in a document becomes more complex when teams are already building, testing and deploying. Pilots show promise but scaling is where friction appears. So why is AI strategy different?
First, the technology landscape is constantly changing. Model performance improves quickly. Pricing models shift. New tools appear while others fade. Regulation continues to evolve.
Strategies built on fixed assumptions about cost, capability or risk can become outdated within months. This makes long‑term planning harder and can reduce organisational appetite, as decision‑makers hesitate to commit in an environment that feels unsettled.
Second, the work is inherently multi-disciplinary. AI delivery extends well beyond the technology team. It involves data, cybersecurity, legal and compliance, procurement, HR, change management, frontline operations and more. Each function has different priorities and risk thresholds.
If the strategy focuses only on the technical build, it overlooks the coordination required to make AI work in day‑to‑day operations.
Third, capability maturity varies across organisations. Some teams experiment rapidly. Others are still addressing data quality, legacy systems or unclear governance. Confidence in AI therefore differs across the business.
More mature teams push ahead, while others hesitate because the foundations are not yet secure. A single pace rarely works - strategy needs to recognise and respond to these differences.
Where organisations get stuck
Common challenges are rarely technical – they are organisational. A lack of clarity is one of the leading issues for this reason.
AI ambition is often framed broadly: “improve productivity” or “embed AI.” This signals intent, but not direction. Without defined outcomes, measurable targets and clear decision rights, alignment breaks down and success remains ambiguous. This is often driven by limited clarity on where AI creates real advantage, leaving goals abstract and prioritisation weak.
At the same time, misaligned or limited organisational appetite can also hold up firms, because even with defined goals, appetite for AI is rarely consistent. Some leaders favour rapid experimentation; others focus on regulatory, workforce or reputational risk. When this appetite is uneven or unspoken, decisions become inconsistent. Some teams accelerate, others apply brakes - creating friction and mixed signals about how far and how fast the organisation is willing to move.
Meanwhile, teams seek speed and flexibility. Leadership seeks cost control, compliance and predictable risk. Both are reasonable, but they pull in different directions.
When guardrails are unclear, teams either move too fast or wait for approval. Leadership responds with tighter controls, and the organisation swings between open access and restriction. Without a clear operating model, this tension slows innovation and increases unmanaged risk.
AI also evolves faster than annual planning cycles. By the time strategies are refreshed, assumptions on cost, capability and tooling are already outdated. Shorter, more adaptive review loops are required. Finally, capability blind spots are found when strategies often focus on use cases and tools, but underinvest in readiness. Gaps in skills, governance and change management mean strong pilots fail to scale, and AI remains experimental rather than embedded.
What strong AI strategies do differently
Addressing these friction points requires more than refining the strategy document. It requires clear ownership, disciplined prioritisation, defined guardrails and review rhythms that support both innovation and control.
Clarify where AI will create advantage and where it will not. Set priority domains, expected impact and appetite for change so teams align on what matters and how success is measured. One way of doing this is to create a centre of excellence. A multi-function coordination layer that sets guardrails, supports reuse and aligns decision-making. Done well, it reduces duplication, accelerates learning and gives leadership confidence that experimentation is happening within agreed boundaries and ambition.
Actively prioritising a portfolio can also help – balancing quick wins with long-term capability building. This ensures investment is deliberate, trade-offs are explicit and resources are directed toward strategic efforts. Meanwhile, ensuring someone is accountable for tracking model evolution, tooling shifts, regulatory developments and talent market changes means strategies can adapt accordingly.
Feeding into this, firms should adopt faster review cycles, regularly testing assumptions and adjust direction. Regular review cycles provide a structured mechanism to reassess risk appetite and adjust ambition in line with evolving confidence and capability. But above all, the most effective organisations explicitly design for both controlled platforms and flexible experimentation. Secure core environments coexist with sandboxed innovation spaces. This balance allows teams to innovate safely without bypassing governance.
Walking the talk
Cortex Reply works alongside organisations to assess AI maturity and readiness, shape practical and executable strategies, and build the technical and organisational foundations required to scale. Writing the strategy is only the starting point.
The real work lies in building the organisation to deliver it. AI strategy is less about vision statements and more about building adaptive capability. That includes deliberately shaping organisational appetite - creating confidence through guardrails, demonstrating value early and ensuring leaders understand both the risks and the opportunities.
This requires governance that evolves, coordination that scales and leadership that recognises the experimental nature of the field.
