Equilibriumby Adopt AI
Start a conversation

Equilibrium · the engine behind the AI Adaptation Profile

Team AI readiness measurement you can defend.

Equilibrium is the instrument. It measures how a team adapts to AI at work, and it is deliberately conservative: every score carries its uncertainty, every norm is versioned, and nothing about a group is reported until enough people have answered to make it fair.

Adopt AI Co is a trading name of AJILITY PTY LTD, an Australian company. It builds Equilibrium and publishes the free ApplyAI profile. We are not the Australian Government AI Adopt Program, not an AI Adopt Centre, and not AdOPT AI Pty Ltd.

Built on established psychometric constructs. Reliability and validity evidence is being collected and will be published as it is completed. See our limitations page for what this profile can and cannot support today.

We are talking to a small number of organisations. Start a conversation , or sign in if you already have access.

See a sample reportFor teams

Team Snapshot · sample

n = 7agreement 0.81
Adaptability71likely 59.6–82.4 · 0–100
AI fluency58likely 46.6–69.4 · 0–100
Human-edge skills80likely 68.6–91.4 · 0–100

Illustrative bands. The shaded interval is derived from the standard error of measurement, if two people's bands overlap, they should not be reported as different.

Assessment

~50 min

One resumable assessment, six reporting lenses

Reporting floor

n ≥ 5

Hard floor before any team analytic renders

LLM-produced scores

0

Scores a language model is allowed to produce

How it works

Three things that shape every output

What Equilibrium measures at team level

The same individual profile that produces an AI Adaptation Profile, adaptability, human-edge skills, AI fluency and disposition, is aggregated into team-level views: how much the group agrees with itself, how psychological safety is distributed, and which contribution roles are covered. Pressure-tendency signals now score and report against provisional norms. Drivers and values are collected by the instrument but are not scored or reported yet; that work is in development. Nothing is inferred from job titles or activity data; every index traces back to items a person answered.

Why confidence bands and versioned norms matter

A number without its uncertainty invites comparisons the data cannot support. Every score renders with the interval derived from its standard error of measurement, and where two bands overlap the platform will not describe the people as different. Norm tables and item parameters are stored as versioned data, so any report can be traced back to the exact reference sample and scoring key that produced it.

The n ≥ 5 threshold that protects participants

No team analytic renders until at least five valid responses exist for that group, and managers never see individual item responses. Below the threshold the surface stays suppressed rather than degrading to a smaller, re-identifiable sample. Outputs are advisory: nothing here is designed to be the sole basis of a decision about a person.

Who it is for

Built for the people who have to stand behind the number.

Three colleagues around a table reviewing a printed team report

  • Leadership teams

    A Team Snapshot that says what the group agrees on and where it does not, with the uncertainty drawn.

  • Consultants and advisers

    An instrument you can put in front of a client and defend line by line.

  • Sceptical readers

    The science, the standards and the limitations, published in full.

Where the evidence actually stands

Being plain about this matters more to us than looking finished. Equilibrium is an instrument in active validation. The norm tables currently in the system are provisional and drawn from early samples that are not yet large or representative enough to be described as benchmarks. Reliability and structural evidence is still being collected, and no independent validation study has been published. We do not claim the platform is generally available, benchmarked, or in use at any organisation. If any of that changes, this paragraph changes with it.

Start a conversation

If measuring your team this way is interesting to you, as an early participant, a design partner, or just a sceptical reader with good questions, tell us a little and we will pick it up.

We use these details to answer your enquiry and nothing else. No assessment data is attached to this form.

Adopt AI is operated in Australia by AJILITY PTY LTD, ABN 39 649 782 427, trading as Adopt AI Co. Equilibrium is provided under the registered business name Equilibrium AI, and ApplyAI under Apply AI Co, both trading names of the same company. Adopt AI is not related to the Australian Government AI Adopt Program or to AdOPT AI Pty Ltd. Contact: hello@adoptai.com.au We use these details to answer your enquiry. Privacy policy.

FAQ

Questions worth asking

How a team adapts to AI at work: adaptability, human-edge skills and demonstrated AI fluency, each reported with its confidence band. It does not produce a verdict about any individual.

The free ApplyAI profile is about eleven minutes and 45 fixed questions. The leadership team assessment is about 50 minutes and can be paused and resumed.

The profile and a five-person Team Snapshot are free. The Team Diagnostic is priced by team size from $1,800 ex GST and the platform is sold per seat.

A structured measurement of how a whole team adapts to AI at work, reported as aggregates. Adopt AI's version reports only at five or more people and never shows individual answers to managers.

Each person completes the same profile, and results are aggregated only once five or more people finish: adaptability, human-edge skills and AI fluency, each with a confidence band.

No. Organisations see aggregate patterns only, never an individual's responses or scores, and results cannot be traced back to a person.

No. Zero scores are produced by a language model. Every score comes from items a person answered, scored server-side against versioned norms.

Five. Team results only appear once at least five valid responses exist, and the threshold is enforced on the server.

In a managed PostgreSQL database in Singapore (AWS ap-southeast-1), never used to train AI models. Analytics sit with Google and PostHog in the United States.