Is there a real problem here?
Work signal, one bounded example, how often it recurs, and the clearest current cost. A workflow nobody can name an example of, and nobody can price the friction on, is not yet a candidate for anything.
Opportunity Studio · private by default
Understand the work
before selecting a tool.
This guided decision aid helps you identify one responsible next move. Your answers stay in this browser during this prototype and are not transmitted.
Skip the briefing and start question 01What this is
Twenty-three questions, in the order the answers actually matter.
Most readiness quizzes ask whether you are excited about AI. This one asks whether the work is understood, whether the information is permitted, whether anyone can tell a good output from a bad one, and whether a named human has the authority and the time to stop it. The questions run in five groups, and later groups cannot rescue an earlier failure.
23Questions, one per screen
5Answers that override the score
0Answers transmitted anywhere
FreeNo account, no email, no payment
Who and what would this touch?
Whether the process is documented end to end, where human judgment carries the work, whether trust or dignity is at stake, what happens if an output is wrong, and whether an affected person could contest it.
Can the information carry it?
Whether the necessary data is lawful, appropriate and available; how sensitive it is; and whether a person can tell an accurate output from an inaccurate one.
Does anyone have capacity to run it?
Baseline, a measurable definition of better, a named reviewer with protected time, an owner after the experiment ends, absorbable upkeep, and equitable access for the people who do the work.
Could it be stopped safely?
The information boundary no external tool may cross, a working non-AI fallback, a stop condition defined in advance, the smallest reversible test, and one honest statement about present readiness.
How the result is reached
The arithmetic is published, because a hidden score is not a reason.
Every answer carries a weight between −3 and +3. The weights are added. Then five specific answers are checked separately: if any one of them is selected, the sum is disregarded entirely and the result is stop and repair. This is the whole mechanism. You can confirm it by running the Studio and watching a single answer change the finding while the score stays high.
Overrides the score
Five answers that end the assessment on their own.
- A wrong output could affect safety, rights, livelihood, or legal interests.
- The necessary information is not currently lawful, appropriate, and available.
- A person cannot currently verify the output against a trusted source.
- No capable person currently has authority and time to review the output.
- The people doing the work would not have equitable access or usable accommodations.
Repairing one of these should trigger a fresh review. It does not, by itself, make a tool appropriate.
Otherwise, the bands
Five findings, and the score range that produces each.
Stop and repair. At least one non-negotiable condition is missing. Positive answers elsewhere cannot cancel it out.
Any controlling condition selectedTest carefully. Enough clarity, review capacity and reversibility to define a small test — not a recommendation for a particular tool.
Score 21 or abovePrepare first. A plausible opportunity with incomplete operating conditions. Repair those before selecting technology.
Score 7 to 20Repair first. The uncertainty matters more than the tool. Map the work, evidence, ownership and review path first.
Score −5 to 6Human-led. Consequences, relationship value, information boundaries or missing review outweigh the case for a test.
Score below −5
What you leave with
A dated working record you can hand to someone else.
The Private Opportunity Record
A plain-text brief containing the finding, any controlling conditions, and all twenty-three answers with the question each one answered — so a colleague can see why the result came out the way it did rather than being asked to trust it. Copy it, or download it as
aloha-ai-private-opportunity-record.txt.The Monday Plan
Seven fields: one action, an owner, a safe or approved input, a measure of usefulness, a reviewer with authority, a stop condition, and unresolved questions. Anything you leave blank prints as “Not yet defined” in the exported record instead of disappearing.
An optional bounded team handoff
A separate short block containing only the assigned task, owner, reviewer, acceptance measure and stop condition. It exists so you can give a team its duties without circulating every private answer you gave.
One onward route, matched to the finding
A test carefully result routes to the Pilot Design Kit; prepare first to the Decision Record Builder; stop and repair to the learning materials; human-led to the Source Desk. The route follows the finding and never changes it, and nothing on it is for sale.
Boundaries
What this prototype does not do.
It does not send anything.
What holds
Answers and the Monday Plan are written to this browser under the key aloha-ai-opportunity-record-v1 and read back when you return to the page.
What that means for you
Clearing site data, using private browsing, or switching device removes the record. Download the brief before you leave if you need to keep it. Aloha AI cannot recover it for you.
It does not name a tool.
What holds
The finding describes the state of the work: whether it can be tested, prepared, repaired, or should stay human-led.
What it is not
No vendor is scored, ranked, recommended or excluded, and no result should be read as approval to buy anything.
It does not certify anything.
What holds
It organises your own answers into a defensible next step and shows the reasoning that produced it.
What it is not
Not a compliance finding, safety assurance, accessibility conformance statement, legal advice, or a guarantee of savings. A person still decides.
Submission is deliberately switched off.
There is no route on this page for sending your answers to Aloha AI, and that is not an oversight. A submission route would first have to let you preview the exact subset being sent, say what assistance you are asking for, and consent before anything left the browser. Until that exists, nothing is sent. If you want a human to look at a situation, bring it through a nonconfidential conversation instead.
Work signal
01 / 23
Opportunity Studio
What brings you here?
Start with the work—not a software catalog.
Answers stay on this device and are not transmitted.