Week 3: Advanced tactics for AI chatbots
Tuesday, October 28 at 9am PT | 12pm ET | 4pm UK
Overview
In this session, we took a deeper dive into AI chatbots, including advanced prompting tactics and the latest advanced features. And Fiona briefed all on the Capstone Project.
Quick Links
📺 Session recording
🌟 Key points
💻 Session slides
💭 Workout
📚 Further resources
👩💻 Capstone Project brief
Key points
Advanced prompting techniques:
Prompt chaining: Start with a base prompt, refine through iterative follow-ups (e.g. “make it shorter”, “rewrite less formal”). Also used to repurpose content (e.g. turn a speech into a follow-up email).
Quick probing replies: Use fast follow-ups like “argue the other side” or “what risks exist?” to deepen reasoning and expand creativity—your “cognitive trampoline.”
Tree of Thoughts (ToT) prompting: Complex decision-making method involving generating multiple options, scoring them, and iterating. Best for big personal or strategic decisions.
Megaprompting + meta-prompting: Mega = highly detailed, interactive prompts. Meta = ask AI to help write your mega prompt collaboratively. Time-consuming upfront, but saves time for large tasks (e.g. survey analysis, strategy planning).
Advanced AI chatbot features
Top 3 to use right now:
Advanced Voice: use it to have a human-like, real time conversation with your AI chatbot. Try using it to talk through challenges, draft on-the-go, or rehearse difficult conversations.
Notebooks (Google and MS Copilot only): a digital “binder” with many docs or sources that you can then chat with, learn from, analyze, synthesize. Great for learning or for querying a large number of docs.
Copilot Pages or ChatGPT Canvas: a dedicated window on the right side of your screen with a substantial piece of content you’re creating. Allows you to more easily edit in line while also iterating with AI. Good for complex docs.
Other features to make use of:
Advanced Reasoning Models: Use for complex tasks like onboarding plans or career development—more nuanced outputs.
Advanced Research Models: Ideal for sourcing vendors, writing briefings, or digesting technical content.
Connected apps: Access your desktop and drive docs from within your AI chatbot. Comes as standard with Enterprise Copilot and Gemini accounts.
Memory: Use custom instructions to preserve tone, voice, and preferences.
Capstone Project Brief — you have three options, debrief due November 21; share progress in WhatsApp.
Beginner: Prompting playbook
Intermediate: Build your own AI assistant
Advanced: Automate your workflow (e.g. via Zapier or Power Automate)
Top ways for EAs to use advanced models:
Advanced Reasoning models
Decision support: Compare options or scenarios using structured frameworks (e.g. SWOT, ToT)
Calendar optimization: Restructure your exec’s week to reduce context switching and protect deep work time.
Meeting summaries & next steps: Turn raw notes into clear, prioritized follow-ups with owner/action/deadline formatting.
Message crafting & tone shaping: Tailor comms (internal or external) with the right voice, structure, and framing for impact.
Critical thinking partner: Pressure-test your assumptions, explore alternative solutions, or work through complex challenges step-by-step.
Advanced Research models
Industry & trend intelligence: Summarize market shifts, AI use cases, or regulatory updates tailored to your exec’s domain.
Meeting & stakeholder briefings: Build insight-rich profiles on new investors, speakers, board members, or partners.
Tool & vendor evaluation: Compare software or service providers side-by-side, based on your team’s needs and constraints.
Event & venue scouting: Research high-end, well-matched locations for offsites, strategy days, or executive dinners.
Synthesize across sources: Combine learnings from multiple reports, articles, or decks into a single digestible briefing with strategic insight
Session slides
Workout
AKA your post-session actions to take
1️⃣ Create a writing style guide to use in future prompts, using the following prompt:
Act as an experienced copywriting coach. Take the below samples of my [emails] and analyse them. Create an extremely comprehensive tone of voice and style guide that describes my particular style of [email] writing. Include specifics like commonly used phrases and emojis, structure of [emails], and any other details to describe the [email] writing style with as much specificity as possible.
2️⃣ Try one advanced feature in your AI chatbot, share reactions in Whatsapp
Further resources
Bonus demos!
Copilot: create an email style guide (part 1 of your 2-part Workout)
Grok 3’s Think (reasoning model): help me vet an event my exec is considering speaking at
ChatGPT Canvas: help me draft and polish a health & safety policy
Carve COIF Prompt Writer Custom GPT demo (link to access the GPT is below)
Perplexity Comet browser: summarize web content, create your own content based off of it
Two more prompting resources: