Applying ODI Rules

Important Note: I do not own the Outcome-driven Rules. They have been continually in flux over the years and all I know is what I learned at a point in time. These are not guaranteed to be as perfect as the human practitioners of ODI. 😜

## ODI Rules

Now that you have constructed the base statement I’m going to give you a very important further instruction.

There are three (3) formats for the final success statement. These formats are pre-pended to the success statement you generated. These are the three types. Only use the prepends that are inside the quotes. The word “Avoid” should be replaced with a version using the second and third format type. The rest is instructional:

Information: recall that I generate multiple formats of success statements. I lay the same foundation for all of them. This is where I diverge to follow this particular rule set.

1. “Minimize the time it takes to “…(do something) - this should be applied to all statements that are not about avoidance.
2. “Minimize the likelihood that “…(something causes an undesirable result) - this is one of formats used when you are trying to avoid an undesired result.
3. “Minimize the likelihood of “…(something undesirable happening) - this is one of the formats used when your are trying to avoid something undesirable from happening.

Examples: There are only three possible structures, so it was very simple to just show AI what they are, and use a little bit of language to finish them off

The following defines the required structure. Always use this structure.

1. Direction of improvement = Minimize [MANDATORY]
2. Metric = one of the 3 formats above [MANDATORY]
3. Object of control [MANDATORY]
4. Contextual clarifier [OPTIONAL]
5. Example of object of control [MANDATORY] - but never use more than 3 examples

Example structure: Minimize + the time it takes to + verify the accuracy of a desired outcome + with a customer + , e.g., its meaning, its completeness, its exactness, etc.

Instructional Example: I’m teaching AI what the proper structure of an ODI statement is. The only difference is that I’m forcing it to always output 3 examples. I’m a rule breaker! 😈

Finished example: Minimize the time it takes to verify the accuracy of a desired outcome with a customer, e.g., its meaning, its completeness, its exactness, etc.

Output Example: I simply show AI what a real ODI success metric looks like in the wild. I grabbed this from a presentation that uses its own self to demonstrate the form and structure.

Pro Tip: the examples should always explain the object of the action; in this case the accuracy of a desired outcome statement. You’d be surprised, but I’ve seen people describe with a customer. 🤣

Additional examples:

1. Minimize the likelihood of undetected defects or performance issues, e.g., loose components, calibration drift, software bugs, etc.
2. Minimize the likelihood that extra costs are triggered by an unstable connection, e.g., failover to expensive network, manual troubleshooting, etc.

Examples: More examples just to beat the structure and format into the head of AI

Only use a maximum of three (3) examples of the object of control

Instruction: Why am I doing this again? Have you worked with ChatGPT?

Format types 2 and 3 should only account for about 20% of all statements. Also, prioritize “Minimize the time it takes to” statements over the “Minimize the likelihood…” statements.

Instruction: I’m being honest, I didn’t come up with this rule, and it is an extremely subjective subject. This is not how it’s necessarily done in ODI but the topic does come up and I’ve heard various campfire stories. This is my rule. I could add even more to it, but I see no reason to do so

Format types 2 and 3 should not be framed in the negative. In other words do not output a statement like this: “Minimize the likelihood of not reviewing and updating key initiatives as market conditions change” because it is minimizing the likelihood of not doing something. Also, do not include a connective word such as as “and” or “or”. A better format would be “Minimize the likelihood of failing to track key initiatives as market conditions change”

Information: More instruction and examples to attempt to correct problems I was experiencing early on. I’m slowly removing or refactoring these areas in the hopes of optimizing memory more and more.

The next instructions is EXTREMELY IMPORTANT

Instruction: Does it look like I got frustrated with ChatGPT?

Format types 2 and 3 should NEVER use the words or phrases “not”, “does not”, “do not”, “is not”, or “of not”. For example this statement “Minimize the likelihood that the identified data sources do not capture key customer insights” should be stated as “Minimize the likelihood that the identified data sources fail to capture key customer insights”

Instruction: Repetitive, and for good reason. Long prompts are harder for ChatGPT to parse and remember

##END ODI