AI strategy
Why your enterprise AI strategy matters more than the AI tools you choose
Most AI projects do not fail on technology. They fail because the organisation bought a tool before it defined the business value it wanted, and never left the experimentation stage.
Start with the business problem, not the tool
The core mistake is treating AI as a technology project rather than a business change. Before any software is evaluated, three questions are worth more than any vendor demo:
- What is actually slowing our growth?
- Where does operational friction cost us the most?
- Which decisions are being made without the information that exists somewhere in the business?
Data is the foundation, and usually the problem
AI output quality is bounded by input data quality. The most common blocker we find is not model capability, it is fragmentation, information spread across CRMs, shared drives and inboxes, with no agreed source of truth.
A workable strategy addresses data quality, governance, management, accessibility, security and sovereignty before it addresses which model to use.
What WhatsApp teaches about competitive advantage
Facebook paid $19 billion for WhatsApp in 2014. It was not buying software, messaging apps were not scarce. It was buying data and engagement signals, and the position those created.
The same logic applies now. Future competitive value sits in the ability to generate, organise and apply intelligence. The winners will not be the companies with better tools; everyone has the same tools. They will be the ones with better data and a clearer idea of what to do with it.
Adoption is not operationalisation
Adoption is testing tools. Operationalisation is embedding them in how work actually gets done, which requires mapping workflows, finding bottlenecks, improving the data underneath, and setting measurable objectives. Most organisations that describe themselves as 'doing AI' are doing the first and calling it the second.
Agents amplify what is already there
This is the part worth internalising before deploying agents: they amplify existing conditions. Good data becomes more valuable. Broken processes become visible, loudly, to more people. Agents need reliable data, defined workflows, secure access and governance controls, and where those are missing, the agent does not paper over the gap, it publicises it.
Sovereignty is becoming a requirement, not a preference
Public platforms offer convenience. They do not necessarily offer control. For proprietary information, customer records and operational data, the question of where processing happens and under whose jurisdiction is moving from an IT concern to a board one.
Four questions before you invest
- Is our data ready?
- Do we have a clear, measurable business objective?
- Are governance controls defined and owned?
- Will this scale beyond the pilot?
If the answer to any of them is no, that is the project, not the tool selection.
Key takeaways
- AI projects stall on strategy, not technology.
- Fragmented data is the most common real blocker.
- Agents amplify existing conditions, good and bad.
- Data readiness, a measurable objective, governance and scale: answer these before buying.