To work with someone, we have to build a bridge. A bridge that connects us to them. But a bridge between two people is different from a physical bridge. It involves several two-way and internal processes of communication, sharing, learning, and adapting as our understanding of the other person and the task we are doing evolves.
The same applies to working with Generative AI. We have to build a bridge of communication, learning, and so on. And just as we would with a person, we have to learn how Generative AI thinks, learns, works.
There are hundreds of courses now available on prompt engineering. In fact, there are many more being introduced every month. But as true AI experts will tell you, sometimes it’s best not to structure your prompts too much. That’s why I prefer focusing on some basic skills and methods that can help us get the most out of interactions with Generative AI platforms.
Modifying Your Workflow
I have found it useful to follow my regular work process for tasks. For example, my process for research involves several tasks:
- Problem Statement, Research Plan
- Information Gathering
- Reading, Taking Notes
- Fact-checking, Correction, gathering
- Drafting, Analysing
- Editing
When I work with ChatGPT or Claude, I modify my process to include tasks that are done by the Generative AI platform:
| Task | Problem Statement |
| Who | Me |
| Role | Write, Refine |
| Remark |
| Task | Plan |
| Who | Me |
| Role | Draft, Finalise |
| Remark | Sometimes | use ChatGPT/Claude to brainstorm after my draft. |
| Task | Information Gathering |
| Who | Me/Gen AI |
| Role | I design and conduct surveys and interviews to gather first-hand information; I also use Gen AI to gather official, published information from digital sources. |
| Remark | I fact-check every piece of information including what Gen AI generates, rejecting all inaccurate data, prompting till I get accurate info or till it is clear that required info is not available. |
| Task | Reading, Taking Notes |
| Who | Me/Gen AI |
| Role | Gen AI provides drafts of all research it conducts; I extract data and analysis notes. |
| Remark |
| Task | Drafting, Analysis |
| Who | Me |
| Role | Write, rewrite, create charts, graphs. |
| Remark |
| Task | Editing |
| Who | Me |
| Role | Editing and finalising research. |
| Remark |
| Task | Who | Role | Remark |
| Problem Statement | Me | Write, Refine | |
| Plan | Me | Draft, Finalise | Sometimes I use ChatGPT/Claude to brainstorm after my draft. |
| Information Gathering | Me/Gen AI | I design and conduct surveys and interviews to gather first-hand information; I also use Gen AI to gather official, published information from digital sources. | I fact-check every piece of information including what Gen AI generates, rejecting all inaccurate data, prompting till I get accurate info or till it is clear that required info is not available. |
| Reading, Taking Notes | Me/Gen AI | Gen AI provides drafts of all research it conducts; I extract data and analysis notes. | |
| Drafting, Analysis | Me | Write, rewrite, create charts, graphs. | |
| Editing | Me | Editing and finalising research. |
Playbook for Effective Prompting
Promoting is really just making good statements and asking clear questions. But we can break it down to a few clear areas:
- Clarity: be clear what you want, provide details.
- Context: provide background or purpose of the information you are seeking.
- Boundaries: Specify the limits of how much information you want, who it is for.
- Task: Define the tasks specifying what format you want answers in.
- Role: Assign a role for the AI so it knows its perspective.
- Adapt: As the conversation progresses, learn from the AI’ responses and adapt your prompts and queries.
Let’s take a look three different prompts and the responses from ChatGPT.
Prompt 1: Give me the list of best movies.
ChatGPT:
🎬All-Time Classics
- The Godfather (1972) – Crime drama about a powerful Italian-American mafia family.
- Casablanca (1942) – A romantic wartime drama full of iconic lines and emotion.
- Citizen Kane (1941) – Often hailed as the greatest film ever made for its innovation.
🧠Mind-Bending / Thought-Provoking
- Inception (2010) – A dream-within-a-dream heist with stunning visuals.
- The Matrix (1999) – Reality, illusion, and rebellion against AI control.
- Interstellar (2014) – Space travel meets emotional gravity.
💥Action / Adventure
- Mad Max: Fury Road (2015) – Non-stop adrenaline with visionary style.
- Die Hard (1988) – A benchmark for action thrillers.
- Raiders of the Lost Ark (1981) – Indiana Jones’ classic archaeological adventure.
😂Comedy
- Monty Python and the Holy Grail (1975) – Absurdist British humor at its finest.
- Groundhog Day (1993) – A weatherman stuck in a time loop.
- Superbad (2007) – A coming-of-age teen comedy.
😢Drama / Emotional
- Schindler’s List (1993) – A powerful story about the Holocaust.
- The Shawshank Redemption (1994) – A tale of hope and friendship in prison.
- Forrest Gump (1994) – Life, love, and history through the eyes of a simple man.
🎶Musical / Music-Centric
- La La Land (2016) – Jazz, dreams, and love in Los Angeles.
- Amadeus (1984) – The life of Mozart through the envy of Salieri.
- Whiplash (2014) – A young drummer’s intense pursuit of greatness.
👪Family / Animated
- The Lion King (1994) – Disney’s Hamlet for kids, with epic music.
- Spirited Away (2001) – A magical journey by Studio Ghibli.
- Toy Story (1995) – The start of Pixar’s legendary run.
Prompt 2: Which are the Top 5 comedy movies?
ChatGPT:
😂 Top 5 Comedy Movies
- Monty Python and the Holy Grail (1975)
British absurdist humor at its peak. A cult classic that parodies the legend of King Arthur with unforgettable lines and ridiculous scenarios. - Groundhog Day (1993)
Bill Murray relives the same day over and over — darkly funny, existential, and oddly touching. - Superbad (2007)
A hilarious coming-of-age teen comedy with sharp dialogue and genuine heart. Launched Jonah Hill and Michael Cera into stardom. - Bridesmaids (2011)
Female-led comedy that broke barriers — riotously funny and emotionally grounded. Kristen Wiig and Melissa McCarthy shine. - Dr. Strangelove (1964)
A dark political satire by Stanley Kubrick. It turns nuclear war into a chillingly funny farce.
Prompt 3: I am writing an article on popular movies of the year. Please give me a table of the Top 5 movies by box office collections in Asia, Europe and USA.
ChatGPT:
📝 Summary Table
| Region | Asia |
| 1st | Ne Zha 2 (~$2.2 B) |
| 2nd | Detective Chinatown 1900 (~$503 M) |
| 3rd | Minecraft Movie (~$954 M) |
| 4th | Lilo & Stitch (~$915 M) |
| 5th | MI: Final Reckoning (~$547 M) |
| Region | USA |
| 1st | A Minecraft Movie ($423.9 M) |
| 2nd | Lilo &Stitch($391.6 M) |
| 3rd | Sinners ($277.6 M) |
| 4th | Captain America: Brave New World ($200.5 M) |
| 5th | Thunderbolts ($189.6 M) |
| Region | Europe |
| 1st | A Minecraft Movie (€5.68 M wknd) |
| 2nd | Lilo & Stitch (€3.24 M) |
| 3rd | Mufasa: The Lion King (€3.25 M) |
| 4th | MI: Final Reckoning (€2.33 M) |
| 5th | Paddington in Peru (€2.08 M) |
| Region | 1st | 2nd | 3rd | 4th | 5th |
| Asia | Ne Zha 2 (~$2.2 B) | Detective Chinatown 1900 (~$503 M) | Minecraft Movie (~$954 M) | Lilo & Stitch (~$915 M) | MI: Final Reckoning (~$547 M) |
| USA | A Minecraft Movie ($423.9 M) | Lilo &Stitch($391.6 M) | Sinners ($277.6 M) | Captain America: Brave New World ($200.5 M) | Thunderbolts ($189.6 M) |
| Europe | A Minecraft Movie (€5.68 M wknd) | Lilo & Stitch (€3.24 M) | Mufasa: The Lion King (€3.25 M) | MI: Final Reckoning (€2.33 M) | Paddington in Peru (€2.08 M) |
While all three prompts led to information, each response was different and suited for different purposes. The first prompt led to several lists of movies by genre and the results were very subjective. Adding a criteria or more details in expected response could have improved the response.
The second prompt led to a list of 5 movies that one may or may not agree with since there was no qualitative criteria but there was a single list of one genre which is what was expected.
The third prompt extracted a clear and measurable set of information suited well for the task: by region, ranked by revenues.
These are of course examples for Generative AI based on Large Language Models (LLMs), which are reasonably good with conversations and word-based interactions. For LLMs, a good prompt should include:
- Define Task: Start with clear action words: ‘List,’ “Design’, ‘Define’, ‘Explain’
- Details – Include key information, data, or materials to work with
- Output specifications – Define format, length, audience, and style expectations
- Quality criteria – If possible specify what would constitute good or great quality.
Prompting for Generative AI that specialise in generating images is very different. And we will address that another day.