AI clarity at 0.487: why return on investment stays unproven in the AI bubble

What does AI clarity at 0.487 measure in the AI bubble?

Clarity belongs to domain C, action and matter, with stability at 0.412 and energy at 0.588. It measures how clearly an organization states what an AI project must achieve and how it will know. The value of 0.487 is the highest of the five dimensions that map to failure causes, and it still sits 0.013 below the 0.5 viability threshold. The study computed it in April 2026 from public data.

That small distance explains the tone of the study. Clarity is the dimension closest to a passing mark, so it responds fast to simple work. A team that writes down its target and its measure moves the value more quickly than a team that buys another tool.

Why does AI return on investment stay unproven in 22% of failed projects?

The study lists return on investment not demonstrated as the fourth cause of failure, at 22%. It follows insufficient data at 42%, missing strategy fit at 31%, and team resistance at 24%. Shares overlap, so a project can fail on several counts, and an unproven return often travels with a weak strategy link.

Return stays unproven when nobody wrote the target. A project launches with a vague promise such as better productivity or smarter decisions. Months later no one can say what changed, by how much, or against which baseline, and the sponsor has nothing to defend at the budget review. The project ends for lack of evidence, even when some value existed.

How can you define AI return on investment before a project starts?

Four items make a return measurable. A baseline records the situation before the project. A metric names the number that should change, such as hours per case or errors per hundred. An owner answers for the number. A review date sets the moment when the team compares the result with the baseline and decides to continue, adjust, or stop. Keep each item to one sentence, so the paragraph stays readable by a non-specialist.

Write the four items in one paragraph and approve them before the first purchase. The paragraph becomes the test for every later discussion. When a pilot ends, the team reads the metric, compares it with the baseline, and records the verdict, which turns an opinion into evidence that a board can audit.

What does $200B a year of AI spending mean for proving a return?

The study states that the AI industry spends $200 billion per year on research and deployment, and that MIT found 95% of generative AI projects failing in production in 2025. At that scale, a missing return is a market-wide issue and not a local one. Each unproven project adds to the gap between spending and results.

The study also cites Bain & Company: failed digital transformation projects cost $2.3 trillion per year worldwide, with generative AI now a main component. The amounts are large because many organizations buy tools before they define what the tools must achieve. Clarity at 0.487 is the structural measure of that habit.

How does AI clarity connect with explainability and trust?

A return can only be proven when the result can be explained. If a model recommends an action and no one can say why, the team cannot separate the model’s effect from luck. The second recommendation of the study, requiring explainability above 0.5, is expected to lift clarity from 0.487 to 0.55 and trust from 0.450 to 0.60. The same reasoning applies to a vendor demo, where a polished result needs a baseline before it counts as proof.

This link gives the work an order. Explain the model first, measure the effect second, and report the return third. Skipping the first step makes the next two fragile, since a measured effect that nobody can explain is hard to defend in front of a finance team or an investment committee.

How much value could clearer AI return on investment recover?

The study sums the 4 recommended corrections. Together they lift the market score from 0.337 to 0.401 and lower the predicted failure rate from 94% to 82%. The study turns this into money: recovering 40% of the $200 billion invested each year means $80 billion of recoverable value per year for the world market.

The figure is an estimate built on the study model, and it shows the order of magnitude of what clearer targets could change. A company can apply the same logic at its own scale. Take its AI budget, apply the same failure rates, and compare the loss at 94% with the loss at 82%. The exercise takes minutes and gives your finance team a first order of magnitude.

How should you read unproven AI return as an AI bubble signal?

A bubble forms when prices assume returns that evidence has not yet shown. Unproven return on investment is the plainest version of that gap. If clarity stays near 0.487 while spending continues, the distance between price and proof can persist. This page is general information, not investment advice, and it reflects the study as of October 2026.

To prove a return in your own organization, the Λ simulator estimates your φ(S) in about 10 minutes, and a 30-minute call with the EQOS Asia team can review your AI projects against the four items of a measurable return. You leave with a clarity value and a list of projects without a baseline.

Frequently asked questions

About the author

Stéphane Jambu

President, EQOS Forecast Asia

Stéphane Jambu leads EQOS Forecast Asia and works as a consultant in e-commerce, search visibility, and AI. LinkedIn profile