The AI Question That Silences Every Leadership Meeting
One question will tell you whether your AI programme will deliver returns this year. It has nothing to do with which model you're running.
Every AI strategy meeting follows the same script. Energy is high. Decks are polished. Everyone agrees AI is transformative.
Then someone asks:
"What percentage of our AI investment is going into people versus platforms?"
The room goes quiet.
The Ratio worth examining
When less than 70% goes toward people — training, workflow redesign, change management, giving teams space to work differently — it raises a question worth sitting with: are we solving the right problem?
Most organisations I've seen succeed get this right. The rest get it backwards. They spend 80% on platforms, models, and infrastructure. Then they wonder why adoption is low, value is invisible, and the board keeps asking "where's the ROI?"
The technology was never the bottleneck. Readiness was.
You're spreading thin and calling it strategy
Here's another pattern worth naming: AI deployed thinly across every function. A chatbot here. A summarisation tool there. A coding assistant in engineering.
Each produces a modest gain. None produces transformation.
The freed-up capacity goes unmonetised because nobody designed what happens next with the time saved.
The teams I've seen ship actual AI value all did the same counterintuitive thing: they picked three domains or fewer, went deep, and refused to expand until the economics were proven. Everyone else is running twenty pilots and measuring nothing.
What does this look like in practice?
- A financial services firm concentrating AI on credit decisioning, fraud detection, and client onboarding — three domains, measurable results in six months.
- A manufacturer applying AI to demand forecasting, quality inspection, and supply chain optimisation — repeatable capability before expansion.
The principle: depth before breadth. Build the muscle in one area, prove it pays, then replicate.
If you had to pick only three areas where AI must deliver measurable revenue or cost reduction this year — which three would survive the cut?
Where the returns actually are
The productivity gains are real. But they follow a specific pattern.
For developers: Claude Code, Kiro, OpenAI Codex — these aren't incrementally faster. They're structurally different. What once required an eight-person squad now collapses to a product owner and a full-stack engineer working alongside AI. I've seen teams report 3-5x throughput on well-scoped tasks. Not on everything. On well-scoped tasks.
For decision-makers: Strategy teams synthesise market signals, model scenarios, and pressure-test assumptions in hours rather than weeks. The value comes not from replacing judgement but from compressing the time to exercise it.
For operations: Document processing, compliance checking, scheduling, reporting — redesigned end-to-end rather than patched with point solutions. Admin teams that spent 60% of their time on manual processes now spend that time on exception handling and relationship management.
The common thread? None of these gains came from choosing a better model. They came from redesigning how people work.
The Ceiling Isn't Technical
Put simply: the ceiling on your AI impact isn't the technology. It's your organisation's readiness to absorb and scale it.
The organisations capturing value aren't those with the fanciest tools. They're the ones with an execution engine — the combination of talent, operating model, and cultural readiness that allows them to repeatedly turn AI capability into outcomes that show up in the numbers.
Boards are no longer asking "Are you doing AI?" They're asking "Where is it appearing in the financials?"
If your answer is still a slide deck — the window is narrowing.
The question
Next time you're debating which AI platform to adopt or which model to fine-tune, try this instead:
"What percentage of our AI investment is going into people versus platforms?"
If it's below 70% on people — you've found your problem.
What's your ratio? I'd genuinely like to know.
#ai #discuss #career #programming
Top comments (0)