To share an AI method, keep the instructions, sources and criteria behind a good result. Separate the method from case-specific data, then ask a colleague to try it. Shared contexts and skills bring these elements together in Witkeep.
Context explains the work
For an account plan, context might include the sales strategy, available offers and how the team tracks opportunities. It places the request in a business setting. Customer data then adds case-specific details.
- A prompt
- An instruction given to AI for a task or a stage of work.
- A skill
- A reusable method combining the instructions needed for a particular task.
- A business context
- The information, rules and resources that place AI within the team’s activities.
- An agent
- A system that uses AI and tools to perform tasks within a defined scope.
The method must work on a new case
Remove information specific to the first customer and state what the user needs to provide. Define the expected result, then ask a colleague to try the method. If they must guess which source to use or how to verify the answer, the method still needs more context.
Sharing includes responsibility for maintenance
An offer changes, a procedure evolves, a source becomes outdated. Assign an owner to the method and retain its versions. Use team feedback to improve it, rather than creating multiple diverging copies.