AI trust at 0.450: team resistance as an AI bubble signal

What does AI trust at 0.450 measure in the AI bubble?

Trust belongs to domain B, relation and living, next to care at 0.338 and communication at 0.305. It measures whether the people who work with an AI system rely on it. A value of 0.450 places the enterprise AI market in the fragile zone on this dimension, with room to rise and room to fall. The study computed it in April 2026 from public data on the enterprise AI market.

Trust links to communication through a chain that the study makes explicit. When communication drops under the critical line of 0.3, trust follows it down. That is why a single incident, such as a public hallucination or a discovered bias, can lower two dimensions in one move.

Why does team resistance explain 24% of failed AI projects?

The study lists team resistance as the third cause of failure, at 24%, after insufficient or biased data at 42% and missing strategy alignment at 31%. The shares overlap, since one project can meet several obstacles. Resistance differs from the other two because it grows from people, and no technical patch removes it. Treat it as a design problem for the project team from the first week.

A team that does not trust a system finds ways around it. People copy results by hand, ask a colleague for a second opinion, or return to the old process. The project keeps running on paper while usage fades, and the expected gains never appear in the numbers that justified the investment. Resistance therefore shows up in usage logs long before it shows up in a budget line.

How does AI explainability build or break trust inside teams?

People trust systems they understand. In the study, the explainability of AI models deployed in companies sits at 0.25 out of 1, so most users cannot see why a model gave a given answer. When a recommendation affects a colleague or a budget, an unexplained answer invites a challenge. Explainability is therefore a trust tool before it is a compliance tool.

Confidence without accuracy adds to the problem. The study reports that models sound 34% more confident when they are wrong, and that 47% of enterprise users acted on hallucinated content at least once. A team that has been burned once starts to discount every later answer, including the correct ones.

How do job fears feed team resistance to AI in the AI bubble?

Fear of replacement is a second driver. The study cites 300 million jobs disrupted worldwide, and McKinsey estimates that 30% of hours worked could be automated by 2030. When employees read those numbers while being asked to adopt a tool, support for the tool can weaken.

A project that people see as a threat gets formal compliance and little real use. The study notes that companies use AI mainly to cut human costs, which sets the tone. Teams respond better when the purpose includes new skills and new tasks, and the care dimension covers that side in its own page. A clear statement of what changes for each role, given early, reduces the guesswork that fuels rumors.

Which signs show team resistance to AI before a project fails?

Four signs appear early and cost little to measure. Usage per week shows whether people open the tool. The count of manual copies shows whether they rebuild its results. The share of outputs reversed by a person shows overrides. The delay between an output and its approval shows hesitation. Each sign needs a log and a date. A spreadsheet is enough to start, and the same four columns work for every project in the portfolio.

Plot the four signs on one chart for each project. A pilot where usage falls while overrides rise tells you that trust is dropping, weeks before the budget review. An early reading gives you time to explain, retrain, or change the tool, which costs less than a failed rollout.

Which actions raise AI trust from 0.450 to 0.60?

The second recommendation of the study is to require explainability above 0.5. Do not deploy a model that cannot explain its decisions. Use SHAP, LIME, or interpretable models such as decision trees and regression for critical decisions, and reserve large language models for assistance, not for the decision itself. Each of these choices can be written into the project brief before the first line of code.

The study expects trust to move from 0.450 to 0.60 when teams understand the tools, and clarity to move from 0.487 to 0.55. The gain comes from a design choice made before launch, not from a training session after it.

How should you read AI trust as a signal of the AI bubble?

Low trust slows adoption, and slow adoption delays the revenue that valuations assume. If trust stays near 0.450 while spending continues at about $200B a year, the gap between investment and results can persist. This page is general information, not investment advice, and it reflects the study as of October 2026. Teams that adopt tools at a slower pace than forecast would show up first in the usage logs.

To measure trust in your own teams, the Λ simulator estimates your φ(S) in about 10 minutes, and the pricing page lists the Full EQΦS Diagnostic from US$2,288. The result gives you a trust value that you can compare with the market value and recheck every quarter.

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