Jun 20, 2025
Is too much time being spent on re-prompting and editing AI results, even for the same tasks?
Many wonder why their AI tools are not improving despite doing the same thing repeatedly. This is a common challenge we often encounter at Crux Comms.
We have developed a systematic approach to training AI models for continuous improvement over time.
Here are three proven methods for training your AI models that deliver real improvements, not just faster results:
1️⃣ Feed in better examples Stop using generic templates. Provide your AI with your best past work, such as successful media pitches, the perfect press release, updated strategy documents, and spokesperson materials. Better input leads to better output.
2️⃣ Improve your instructions over time After each AI session, note what worked and what didn’t. Update your prompts based on these lessons. Use simple frameworks like PFSET (Persona, Format, Style, Example, Task) to structure your requests more clearly.
3️⃣ Use the memory function to create feedback loops If you're using a tool like ChatGPT with the memory function enabled, actively ask the model to remember what worked. Once you have edited or refined a result, summarize your final version and explain to the model why it was successful. It creates a developing reference point that sharpens your results over time.
The goal is to elevate your AI implementation, not just automate existing processes.