Skip to main content
Business AI
What Is an AI Strategy?
Most companies have AI aspirations, not AI strategies. Here's the difference, the three most common failures, and a five-step framework for building one that delivers real results.
The reality check
72%
Of organisations report having an AI strategy — but only 25% report measurable business impact from AI [McKinsey]
3.5x
Higher ROI for companies with a formal AI strategy vs those with ad-hoc AI adoption [MIT Sloan]
85%
Of AI projects fail to deliver their intended business value — primarily due to poor strategy and change management, not technical failure [Gartner]
There's a gap between "having an AI strategy document" and "having a strategy that creates competitive advantage." Most companies have the document. The gap is in specificity, resourcing, and execution discipline — the kind of ground-level work covered in how to use AI at work. A slide deck that says "we will leverage AI across our operations to improve efficiency" is not a strategy — it's a platitude.
A real AI strategy answers: which specific use cases, with which tools, owned by which teams, measured by which KPIs, completed on which timeline. Without those specifics — and without the evidence from real productivity data to back them — you have a vision, not a strategy.
The three failure modes
Technology-first thinking
Starting with "which AI tools should we adopt?" instead of "which business problems do we need to solve?" Technology choices should follow problem definition, not precede it. Teams that buy AI platforms first and then search for use cases have a poor track record.
Boiling the ocean
Trying to transform every department at once. Companies that prioritise 2-3 high-impact use cases outperform those attempting broad simultaneous deployment. Concentrated effort produces faster results and builds the internal capability needed for the next wave.
No data foundation
AI strategy without data strategy fails. AI outputs are only as good as the data fed into them. Companies with fragmented, inconsistent, or inaccessible data get far less value from AI investment than those who've established clean data pipelines first.
The 5-step framework
1
Audit your highest-value problems
Interview department heads to identify their 3 biggest time or money drains. Look for patterns: repeated manual processes, information bottlenecks, customer friction points, decisions made on incomplete data. AI solves specific problems, not general inefficiency.
Output: A prioritised list of 10-15 candidate use cases
2
Score use cases by impact and feasibility
Rate each candidate on: potential business impact (revenue gain or cost saving), data availability (do you have the data AI needs?), technical feasibility (is the problem well-defined?), and organisational readiness (will the team actually use it?). Prioritise the top 3 quadrant items.
Output: A 2x2 impact/feasibility matrix with 3 priority use cases
3
Pilot fast, measure rigorously
Run 90-day pilots for each priority use case. Set specific baseline metrics before starting (current cost, time, error rate). Use off-the-shelf tools where possible — custom AI development is expensive and slow. Measure against baselines weekly. Kill pilots that don't show signal at 45 days.
Output: Go/no-go decisions at 90 days with hard data
4
Build capability, not just tools
The bottleneck in AI strategy is usually human capability, not tool availability. Invest in AI literacy training across the organisation. Identify internal AI champions who become super-users and advocates. A team that can prompt effectively and evaluate AI output is more valuable than any individual tool.
Output: AI training curriculum, champion network, internal playbooks
5
Govern and scale
Establish governance: who approves new AI tools, what data can be sent to third-party AI systems, how AI outputs are reviewed before customer-facing use. Then scale proven pilots systematically — document what worked, replicate across teams, connect to budget cycles. Build a portfolio of AI initiatives rather than individual experiments.
Output: AI governance policy, scaling playbook, annual AI budget
AI maturity stages
| Stage | Description | Typical characteristics |
|---|---|---|
| 1. Experimenting | Individual employees using AI tools informally | No policy, ad-hoc usage, bottom-up adoption |
| 2. Piloting | Structured experiments in specific functions | Defined use cases, measurement, limited governance |
| 3. Scaling | Proven use cases rolled out across organisation | Formal policy, training, AI budget line |
| 4. Integrating | AI embedded in core business processes | Custom models, proprietary data assets, AI in product |
| 5. Transforming | AI central to competitive strategy | AI-native workflows, new business models enabled by AI |
The honest answer
Most organisations in 2026 are at Stage 1-2. The companies at Stage 4-5 started in 2022-2023 with deliberate, consistent investment — not splashy announcements. For smaller companies, the same logic scales down in AI for small business. You can't skip stages. Focus on progressing one level at a time with rigour rather than leaping to stage 5 ambitions on stage 1 infrastructure.
FAQ
How much should a company budget for AI?
Benchmarks vary widely. For initial strategy and tooling: $50K-$200K for a mid-market company covers a strategy consultant, tool subscriptions, and basic training. For ongoing investment: 2-5% of IT budget directed toward AI tooling is increasingly common. Enterprise organisations pursuing AI transformation allocate 10-15% of IT budgets to AI initiatives.
Who owns AI strategy — IT, operations, or C-suite?
The most successful AI strategies have C-suite sponsorship (typically the CEO or COO, not just CTO), operational ownership for execution, and IT as an enabler. When AI is "owned" solely by IT, use cases tend to be technical rather than business-value driven. Cross-functional governance with executive sponsorship is the pattern that works.
How do I convince leadership to invest in AI?
Show, don't tell. Run a small, self-funded pilot on a problem leadership cares about, measure the result, and present the ROI case. One $5K pilot showing 40% time reduction in a specific workflow is more persuasive than any slide deck about AI potential. The path from scepticism to commitment goes through demonstrated results, not theoretical benefits.
Sources
[McKinsey] McKinsey — "The state of AI in 2025" (annual survey)
[MIT Sloan] MIT Sloan Management Review — "AI Strategy and Business Value" (2024)
[Gartner] Gartner — "Why AI Projects Fail" research note (2024)
04 — Don't watch from the outside
Veltrix Collective
Stay ahead of
the curve
the curve
Weekly briefings on AI tools, adoption trends, and what actually matters for practitioners. No hype. Just signal. Join readers navigating the shift.
Weekly, every Tuesday · No spam · Privacy policy · Unsubscribe anytime
Written by Luke Madden, founder of Veltrix Collective. Data synthesis and analysis by Vel.