Markets with AI · Materials prepared with support from AI · Updated 3 October 2026 · Method and limits.
The paper should retain the previous plan
If you write the hypothesis after knowing the result, you can reinterpret why you acted. Save the original condition, date and information available before decision. Then add execution and result to separate fields.
A response from AI can sum up the paper and find incomplete fields. He cannot prove his prior intention from a posteriori written account. Nor should I diagnose an emotional cause just because an operation resulted in loss.
Fields that allow an operation to be audited
| Before | Later |
|---|---|
| Instrument, product, currency and date | Inputs and departures with their dates. |
| Status and ground for invalidation | Data to check compliance. |
| Quantity and assumed cost | The quantity and cost actually registered. |
| Projected retirement and expiry | Registered ground for closure and deviation. |
| Absence and assumption | Result, limits of explanation and checks. |
To compare markets, it keeps the specifications for each product. Two equal monetary performance can come from different exposures. Do not add up different percentages or asset units without a consistent definition of the metric.
Example of an anti-implementation plan
Invented case of a linear product: Input plan 100,20 €, adverse output 99,70 €, 40 units and cost 5 €. Retail registration: Input 100,25 €, output 99,60 €, 40 units and cost 6 €.
The intended loss was 25 €and that recorded (100,25 − 99,60) × 40 + 6 = 32 €. The difference is 7 €. You can decompose it at 2 €for the worse entrance, 4 €for the worse output and 1 €additional cost. That decomposition describes quantity and does not prove why they happened.
Promises to review the paper
That's what I'm talking about. That's what I mean. That page doesn't check a model automatically.
What can you ask the AI to check several rows
It starts with detecting different units, empty costs, duplicated rows and out of order dates. He then calculates metrics with an explicit definition. If you ask for a success rate, clear what you do with zero results. If you demand an expectation, check that gain and loss includes costs in the same way.
To evaluate compliance, it provides the rule and information about each case. The AI cannot conclude "the plan was fulfilled" just because the result was positive. Keep a compliance column and a cash result column and they can disagree.
Exercise: An operation earned money but there's no record of the previous rule. Can the AI certify that it was a proper execution?
No. He's got to admit that his previous plan lacks evidence. The benefit describes the result and doesn't itself prove compliance.
Labour privacy and maintenance
Use an anonymous copy with the fields that exercise needs. Eliminates account identifiers, credentials and personal data. Keep the original paper separate and don't let a summary replace the data.
After receiving the review, check two rows on your own and ask to correct concrete errors. Write down what's changed and why. You can start with the Fictitious case of free intradia course before applying the method to other asset logs.
Practice the complete process
Solves tasks with data, prompts and independent controls at free course.
Go to related courseContinue with another task
Sources to check
The examples are computed with fictitious data. The sources do not support a sign or a performance result of these exercises.