Why this matters
AI is no longer just a tooling line item. It is becoming a labor economics question.
Most enterprise AI discussions still live in the language of experimentation: pilots, copilots, productivity uplift, and isolated software budgets. But the more important lens is cost structure. Once AI spend per employee begins to approach salary cost, leadership teams have to rethink where work is created, how teams are staffed, and which workflows still deserve to stay manual.
The chart above compares two curves in the UK tech sector: average annual salary and estimated AI spend per employee. Salary growth has been relatively steady. AI spend growth, by contrast, appears to be compounding much faster.
What the chart says
The crossover is not the end state. It is the signal that the economics of execution are shifting.
Based on the current best-fit assumptions, AI spend per employee converges with average salary cost around late 2026 at roughly GBP 52,000 per employee per year. That does not mean businesses will suddenly replace employees pound for pound. It means the annual budget required to give each employee a powerful AI layer is rapidly becoming comparable to the fully loaded cost of the employee itself.
Salary curve
UK tech salary growth remains broadly linear. Wage inflation is real, but the shape of the increase is familiar and relatively stable.
AI spend curve
AI spend is behaving more like an adoption curve: rapid early acceleration, vendor sprawl, and a growing willingness to budget for model access, orchestration, and workflow tooling.
In other words, the gap is closing not because salaries are collapsing, but because businesses are becoming willing to spend materially more on software that can directly augment or compress labor.
Methodology
A directional estimate, not a claim of false precision.
The salary series uses published digital-sector earnings benchmarks and extends them with a linear line of best fit. The AI series is constructed from aggregate UK AI investment data, company-level spend surveys, and sector workforce assumptions, then modelled with an exponential fit.
- Salary: linear fit chosen to reflect steady historical wage growth.
- AI spend: exponential fit chosen to reflect the compounding shape of enterprise adoption curves.
- Limitation: direct UK tech-sector AI spend per employee is not yet systematically reported by any official body.
The exact crossover quarter should therefore be treated as directional. The more important point is the relationship between the two lines and how quickly the AI curve steepens.
How to interpret it
If this trend holds, the question changes from “Should we buy AI?” to “Which work should still be human-priced?”
For operating leaders, the strategic implication is not just budget allocation. It is workflow design. Once software spend can plausibly rival salary cost, the right comparison is no longer “tool versus tool.” It becomes “software-plus-human system versus human-only system.”
Role design
Teams will increasingly split between work that benefits from judgement and ownership, and work that should be systematized because the software layer is economically justified.
Budgeting
AI budgets should be evaluated against labor substitution, labor augmentation, and margin creation rather than treated as generic SaaS overhead.
Architecture
Point tools will matter less than whether the company has a coherent system that can deploy models, context, and workflow logic where the cost compression is greatest.
Bottom line
The important takeaway is not the exact date. It is the speed at which AI is becoming economically non-trivial on a per-employee basis.
Even if the growth rate moderates, the direction is hard to ignore. AI spend per employee is moving from experimentation-scale budgets toward labor-scale budgets. Leaders who understand that shift early can redesign workflows, teams, and systems before the economics force the issue.
That is why this chart matters. It turns a vague narrative about AI adoption into a sharper operating question: when software starts to cost what people cost, what should your business ask software to do?