Putting Generative AI Into Daily Practice: Give It 20 Minutes

By Curtis Michelson |  March 19, 2024

Hi! I’m Curtis, and I’ll be your personal GenAI trainer.

With any new technology, there is a common adoption pattern that starts with the early tinkerers, and ends with the laggards admitting they can’t live without it anymore.

Since the launch of ChatGPT in November 2022, generative AI has been flying through that adoption pattern like a meteor— even though some companies still prevent their employees from using web-based generative AI tools, or at least uploading company data.

So how do you get your brain in shape for this new era, and understand the ways it may affect your work, and that of your team? That’s what this article is for.

I want you to find the time for some short experimental “workouts,” perhaps even away from the office. The ask: set aside 20 minutes, and practice some power prompting.

Below are five Gen AI workouts to boost your productivity and understanding. For simplicity, we’ll keep the exercises limited to generic chat boxes; i.e., the standard input boxes you see when using ChatGPT or similar tools like Gemini, Claude, etc. Even if some of those tools support “multi modal” inputs like images, we’ll keep it to text. In fact, I have tested all of these power-ups against multiple vendors to ensure these prompts will work with any of them. In the cases where one tool seems to provide better results than another, I’ll let you know.

And none of the workouts require you to upload anything specific to your job or your company. You may want to use a playful scenario: you’re the chief ice cream designer at Ben & Jerry’s, a Walt Disney Imagineer designing the next great theme park, or a resort marketer trying to lure more people to Fiji.

Okay, ready to begin? Let’s get those fingers loose and relaxed. Deep breaths. Here we go…

Power Up #1: Get Outside The Building

  • Context: Creative work throws us complex problems every day. Sometimes staring at our screen just doesn’t inspire. It often helps to get out of the office.
  • Why It Works: Imagine you had a helpful walking partner who deeply listens, remembers every detail, and can empathize. This partner will take your thoughts, muddled and confused as they might be, and give you a fresh perspective on them.
  • How To Use It: First, install OpenAI’s free ChatGPT mobile app on your phone. Put in your earbuds and start walking. (If walking is not suitable, sit outside and dictate.) For your first prompt input, just say to the app, “For this session, I want you to simply listen to everything I say and reply with ‘Okay.’ Do not offer anything else. I will ask for a summary later.”

    Now, just unload one thought at a time. The mobile app allows you to just speak your inputs and it transcribes and inputs the text. The app has enough memory for an hour walk or longer. Don’t feel capped at 15 minutes.

    When you’ve said your piece, say this final prompt as: “Please replay everything I have said. Summarize my situation into a paragraph, and then follow that summary with a list of my biggest challenges and maybe some helpful solutions.”
  • Key Considerations: Turn phone notifications off for the walking session, otherwise your train of thought will be interrupted by emails, texts, and calls. Speak slowly; don’t rush it. Input one sentence at a time, or a few sentences in a row, but don’t speak for minutes before hitting “submit.” It’s nice to hear your walking partner reply with “Okay.” It really does feel like it’s listening to you, and that will keep you advancing your ideas.
  • Variations:
    • Have it reply with something other than “Okay.” Maybe “Gotcha,” or “Hmm, hmm,” or “I feel ya”?
    • Ask it for a summary in poetic form (Shakespearean sonnets work well)
    • Ask it to change its tone depending on your need (talk to me like a gym coach, talk to me like a therapist, talk to me like a venture capitalist investing in me as a business.)

Power Up #2: Challenge My Thinking Six Ways to Sunday

  • Context: Examining things from multiple perspectives is a hallmark of knowledge work in the modern era. Acknowledging our limited frame of reference and inviting other perspectives can spark solutions to vexing problems. The six thinking hats developed by Dr. Edward de Bono is one such method for doing so.
  • Why It Works: de Bono’s genius was realizing that we needed all six core perspectives (hats) for robust problem solving:
    • White: just facts/info
    • Red: emotions/sentiment
    • Blue: process awareness
    • Green: creative options
    • Black: risks
    • Yellow: opportunities.
  • How: To prime the GenAI, open a new chat thread (called a “context window” in LLM-speak) with “Do you know de Bono’s six thinking hats?” It will of course affirm and validate its understanding with a quick summary of the Six Hats method. Prepare it further by adding this prompt, “I’ll tell you a challenge I’m having, and I want you to then wear each of the hats and come back to me with relevant questions I should be asking from that hat’s perspective about my challenge.” Then input your challenge statement. Make it rich and juicy, including all the characters and constraints involved. The more detail you give, the better its responses will be.
  • Key Considerations: Worth noting I got mixed quality results from four different tools that I tried. All of them were quite good, but the superlative output came from Google Gemini.
  • Variations:
    • Have it reply with the hats in specific orders. I personally like White -> Red -> Blue -> Green -> Black -> Yellow.  Any combination is great.

    • Allow it to go deeper and drill down on one of the hats. For example, Green is about options, so give a follow-up prompt like, “Could you go further with the green hat and give me five more options in a table format?
    • Have it black hat those additional options with, “Repeat that table, but add a Black Hat column to balance it out.”

Power Up #3: Just the Facts Ma’am, Just the Facts

  • Context: The phrase “information overload” has been with us since the 1960s, and after the Internet went mainstream in the 1990s, it only got worse. Now, generative AI tools promise to serve up succinct answers to your questions, but often they detach those answers from their original source. That makes it tough to trust the output.  But what if you could access a tool that would cite quality sources, read them on our behalf, compare and contrast opposing views, and neatly summarize it all for fast sharing with our teammates? That tool is Perplexity.
  • Why It Works: Although all GenAI tools can summarize complex information quickly, Perplexity combines the power of a web search engine with a conversational interface, offering the best of both worlds in one interface. With each response, it suggests “additional questions” to advance your thinking. The web is a loud echo chamber, but unlike Google which prioritizes search results largely on popularity (links to the page), Perplexity’s relevance algorithm indexes on trustworthiness, timeliness, and variety.
  • How to Use: A great place to start is to inquire about a firm — perhaps a competitor, or just a company that you’ve been curious about. You might start your opening prompt with, “Tell me some basics about [company name.]”  You’ll get a well cited summary of 15-20 relevant source docs, plus additional questions you can turn into follow-on prompts; such as, “most popular products,” “business model evolution over time,” “sustainability commitments and actions,” etc.

    Example: I needed some grounding in a commonly-told innovation story about Iridium and how they went bankrupt. I was quickly re-schooled by Perplexity, and learned that the company is healthy and strong today, and actively competes against and partners with Elon Musk’s Starlink and SpaceX respectively. I uploaded iridium’s latest 10K filing, and Perplexity processed and summarized that dense 138-page report, allowing me to query the document for highlights.
  • Key Considerations: Perplexity has a free version, but it has limits that you might hit after a few days of use. It is factually accurate, but in other ways limited. For example, it can analyze pop lyrics but completely misses irony and subtle meanings.
  • Variations:
    • Power feature: limit Perplexity by asking for a specific type of information, like “only academic literature,” or “just YouTube videos”
    • Have it fact-check a document that you upload to it with this prompt: “Please identify any factual errors, and explain the error citing reliable sources.”

Power Up #4: Making Your Presentations Pitch Perfect

  • Context: We humans have only got so much time in the day, and so when a presentation is required, we’re tempted to create something for a broad audience that we can use many times over. But GenAI is already showing its prowess at cranking out several personalized versions of an initial document that you share.
  • Why It Works: The generic form of this prompt is “I’m writing for an audience of [the-role]”, which might be CEOs, doctors, teachers, or kindergartners. Sparing you the technical jargon around table weightings and semantic vectors, this technique works because language transformers are exquisitely sensitive to their inputs. Your input prompts, with even small adjustments, trigger vastly different associations in the machine’s neural net. In summary, small input variations can be harnessed for powerful output variations.
  • How To Use: I like to prime the chat with a context. For example, “I’m writing a pitch for a new internal company process change, can you assist me?”  It will oblige. Then follow on with a prompt like “I need X versions for these X different stakeholders”. Describe each stakeholder group. If they are well known categories or roles (like CEOs, CMOs, CTOs, etc.), their title alone will suffice. Now, paste in your generic pitch, et voila, you get stakeholder-specific results like the ones I generated here in my quick test.
  • Key Considerations: All of the big GenAI platforms do this well, but there are differences between paid and unpaid tiers. When you subscribe to Google’s Gemini or ChatGPT, you can paste in much larger docs, and get faster results. Other commercial tools specifically designed for custom pitching and personalized presentations such as are emerging.
  • Variations:
    • Request output in different languages (Hindi, Spanish, French)
    • Have it embody values, such as, “emphasize corporate ESG terms”
    • Request different lengths, like short, tight summaries for executives, for example.

Power Up #5: Options, I Need Options!

  • Context: Knowledge work and creative teams thrive on their ability to generate innovative options. Prematurely converging on ideas due to cognitive bias or other social pressures works against that goal. How might we create ambiguous spaces for problem-solution concepts to emerge, without prematurely judging or shutting them down? One way is to generate what my friend Keith Hopper at Danger Fort Labs calls “sacrificial concepts”; namely, fast riffs and throw-away ideas that themselves might have a seed within them worth preserving. Those seeds can only be discovered when we have generated a substantial number of cocktail napkin sketches, or sacrificial concepts.
  • Why it works: GenAI is a workhorse, it just doesn’t get tired. Unlike teams of people putting stickies up on a wall and taking a coffee break after 3 hours, LLMs can just keep iterating and iterating and iterating. The volume of ideas can itself overwhelm, but luckily, the genAI can also help you weed though the stack to find common themes and those precious seeds of creative solutions.

    This is an advanced level workout maneuver requiring some special gear. What you’ll need is a list of words or phrases to act as “probability benders”; i.e., concepts that will nudge the transformer to move from an obvious solution to a novel one. This is related to the notion of ‘enabling constraints’ often used in design thinking and it’s where genAI can really shine.
  • How to Use: A simple way to  start might be called a “brand stretch”, where the list you feed it are different company names. Then you will ask the AI to brainstorm solutions ‘as if’ those solutions were an embodiment of that firm’s ethos and style.
    1. Input your first prompt as, I have a challenge I want help with. I will feed you a problem statement and you will ask me for information on who has the challenge. You will act as an innovation facilitator helping fresh new solution ideas forward.  After that, you will ask me for a list of companies and then apply the essence or spirit of that company’s brand and their approach to design to my challenge,. You will generate a novel solution for each company I give you . Ready? 
    2. Input your problem statement when it prompts you.
    3. When it asks, provide a list of companies/people like “Apple, Lego, Taylor Swift, Amazon, and Uber” and see what happens.

      To be sure, these will not be great solutions, but they will spark something fresh. Allow them to be starting points, thought provokers, and in this way you will be truly tapping into the high side of ‘hallucinations’. See my example here.
  • Key Considerations: Be mindful of what you upload. Don’t push company data if you have a corporate policy forbidding it. Experiment with other word/phrase lists, and simply ask the AI to act as a non-literal interpreter.

    For example, I created a custom GPT that uses a list of common words, and then I have ChatGPT riff on the non-literal meaning of each to create some fresh sacrificial concepts. It’s called Binky Innovator. Feel free to play with it, but note that it requires the paid version.

Try each of these exercises yourself — or assign one to different members of your team and then report back on your experiences. Either way, I bet you’ll end this workout series with a much deeper understanding of the possibilities of Gen AI than you had going in.

Featured image by Dall-E. InnoLead members can also download this article in Word or PDF form.