8 Key AI Learnings for Practical Application

The journey into leveraging AI effectively is iterative and strategic. Here are core principles and actionable tips to get the most out of AI tools while navigating potential pitfalls:

1. Iterate & Refine – Don't Expect Perfection on the First Try:

•  AI as a Starting Point, Not an Endpoint: Treat AI-generated content (text, summaries, ideas) as a strong first draft (aim for 80% done by AI). Your expertise is crucial for the final 20% – refinement, nuance, and strategic alignment.
• Ask "What If?" & "What Else?": Continuously prompt the AI with follow-up questions. "Could you elaborate on this point?" "What are alternative perspectives?" "What's a more creative/conservative approach?"
• Small Steps, Big Impact: Start with simple tasks and gradually increase complexity. Learn how the AI "thinks" with one tool before trying to orchestrate many.

2. Actively Seek Criticism (from AI itself!):

• Challenge the Output: Don't just accept the first answer. Ask the AI to critique its suggestions or find flaws in its reasoning. "What are the potential downsides to this approach?" "Are there any biases in this summary?"
• Request Counterarguments: "Could you argue against this proposal?" This helps uncover blind spots and strengthens your final output.

3. Leverage Multiple Tools Strategically (The "AI Toolbox" Approach):

• No Single "Magic Bullet": Different AI tools excel at different tasks.
  • Example: Use PerplexityAI or a similar research-focused AI for initial information gathering and source identification. Use Claude or ChatGPT for drafting, summarisation, and creative brainstorming. Use Google AI Studio for robust content generation and prompt refinement. Use Canva AI for visual creation.
• Combine Strengths: Take the output from one tool and feed it into another for further refinement or a different type of analysis (e.g., research from Perplexity -> summarisation & structuring in Claude -> critique in Google AI Studio).
• Be Mindful of Data Flow: Understand how information is passed between tools, especially if dealing with sensitive data.

4. Prioritise Data Privacy & Anonymisation:

• Understand Tool Policies: Be aware of how different AI tools (especially free versions) handle your data. Does it become part of their training set?
• Anonymise Sensitive Information: Before inputting client-specific or confidential data into public AI tools, rigorously anonymise it. Remove all company names, and specific project identifiers.
• Use "Placeholders": If you need to work with a structure that involves sensitive data, use generic placeholders (e.g., "[Client Name]," "[Project X Revenue]") when prompting the AI, and then manually insert the real data into the AI's structured output outside of the AI tool.
• Explore Secure/Internal Options: For highly sensitive data, advocate for or explore secure, on-premise, or private cloud AI solutions where data control is paramount.

5. Context is King – Feed the AI Rich Information:

• Don't Assume Knowledge: AI doesn't "know" your specific project, client, or internal context unless you provide it.
• Provide Background: Give the AI relevant background information, your desired tone, the target audience, key objectives, and any constraints. The more context, the better and more relevant the output.
• Use Your Own "Best Practice" Examples: Feed the AI anonymised examples of your own high-quality work (reports, emails, proposals) and ask it to learn your style or use that structure as a template for new content.

6. Fact-Check & Validate – AI Can "Hallucinate":

• Critical Thinking Required: Always critically review AI-generated facts, figures, statistics, and especially references. AI can confidently present incorrect information.
• Verify Sources: If an AI provides sources, double-check their validity and relevance. AI has been known to invent citations.
• Your Expertise has the Final Say: You are responsible for the accuracy and integrity of the final output.

7. Build Your "Prompt Library":

• Save Effective Prompts: When you craft a prompt that yields excellent results for a specific task, save it!
• Categorise & Share (Internally): Create a personal or team library of effective prompts for common project management tasks (e.g., "Prompt for summarising meeting notes," "Prompt for drafting a risk mitigation plan"). This accelerates future work and shares best practices.

8. Start Small, Scale Smart:

• Automate One Annoying Micro-Task First: Identify one small, repetitive, slightly irritating task and see if you can automate part of it with AI (e.g., drafting a standard follow-up email, categorising feedback). Success here builds confidence.
• Resist Over-Complication Initially: Avoid building highly complex, multi-tool AI workflows until you're comfortable with individual components and have a clear value proposition for the added complexity.​

​