The End of the Proof of Concept?

When an initiative begins with a solution already chosen, a team is asked to prove that it can work. Success then creates momentum towards further investment.

That sequence assumes technical feasibility is the first uncertainty worth resolving.

Sometimes it is. But management may still not know whether the problem matters, whether users will adopt the solution or whether the organisation can operate it economically. A technical demonstration cannot answer questions it was not designed to test.

The decision is not whether to abandon the proof of concept. It is whether a technical PoC should be the automatic first gateway.

What has changed?

Signal 1 — Technical experimentation is more accessible

Paid cloud adoption among EU enterprises rose from 17.8% in 2014 to 52.7% in 2025. Cloud platforms and reusable services have expanded access to development infrastructure.

This does not directly measure the cost of a PoC. Generative AI provides another emerging but mixed signal: some controlled studies report faster completion of bounded coding tasks, while research involving experienced developers in mature repositories found the opposite. Technical demonstrations can be produced more readily in some settings. That makes their evidential purpose more important.

Signal 2 — The implementation gap is more visible

The distinction between feasibility and investment readiness is not new. Research into 77 highly innovative products identified five dimensions in go/no-go decisions: strategic fit, technical feasibility, customer acceptance, market opportunity and financial performance.

Recent AI evidence exposes the continuing gap. RAND practitioners identified wrong problem selection, weak business alignment, inadequate data and missing infrastructure among recurring causes of failure. The OECD found that 58% of nearly 1,500 EU public-sector AI cases remained planned, in development or in pilot. This does not explain each case, but it shows how much activity remains short of implementation.

Technical success is meaningful. It is not investment readiness.

Signal 3 — Evidence for problem-first methods is developing

The UK Government Service Manual has long required teams to investigate users, needs, constraints, value and alternative interventions before committing to build. Discovery may legitimately conclude that no service should be developed.

Recent research gives qualified support to Design Thinking. A 2025 study of 246 projects associated early experimentation with greater innovation effectiveness and efficiency. A 2024 systematic review found predominantly positive effects on teamwork but limited empirical support connecting those effects to performance.

Design Thinking is useful. The evidence does not establish it as a universal replacement for a technical PoC.

The decision that needs rethinking

The decision principle should move from PoC-First to Uncertainty-First.

If problem importance or user desirability is uncertain, begin with structured discovery. If scientific or technical feasibility is uncertain, begin with a technical PoC. If organisational execution is uncertain, run a realistic pilot.

These approaches may proceed sequentially or in parallel. Their order should follow the material risk, not a standard innovation ritual.

Resolving the first uncertainty does not complete the investment case. Material scaling still requires proportionate evidence of value, operational readiness and accountable ownership.

What hasn't changed

Technical PoCs remain essential when feasibility is genuinely uncertain. In breakthrough research, users and applications may not yet be identifiable. Proving that something can work may be the first useful evidence available.

Design Thinking has boundaries too. It does not prove production economics, security, integration, resilience or maintainability. Its value depends on competent execution and supportive organisational conditions.

The objective is not to replace one automatic gateway with another. It is to commission the experiment most capable of changing the investment decision.

Executive Questions

• What is the most consequential uncertainty this initiative must resolve next?
• Are we testing whether the problem matters, the technology works or the organisation can operate the solution?
• What must be proven about value, operational readiness and ownership before material scaling?

Three Signals to Monitor

Bottom Line

The proof of concept is not ending.

Its automatic position at the front of the investment process should be questioned.

Where feasibility is established, Design Thinking may precede a technical PoC, run alongside it or make it unnecessary. Where feasibility remains the decisive unknown, the technical PoC still belongs first.

The better discipline is to identify the uncertainty that matters most, then demand the proof designed to resolve it.

The first proof should address the greatest uncertainty—not automatically the technology.

References

1. Eurostat — Cloud computing use by EU enterprises, 2025. View source
2. Peng et al. — GitHub Copilot controlled experiment. View source
3. Paradis et al. — Google developer-productivity trial. View source
4. METR — AI and experienced open-source developer productivity. View source
5. Carbonell-Foulquié et al. — Go/no-go criteria for innovative products. View source
6. GOV.UK — How the discovery phase works. View source
7. RAND — Root causes of AI project failure. View source
8. OECD — Implementation challenges for government AI. View source
9. Magistretti et al. — Experimentation in Design Thinking projects. View source
10. Robbins and Fu — Performance impact of Design Thinking. View source
11. Schlott — Systematic review of Design Thinking and teamwork. View source
12. Magistretti et al. — Design Thinking in early technological R&D. View source