Practical AI for Everyone

From years of running AI projects, I’ve seen where AI is useful – and where it isn’t. AI is brilliant in some areas and quirky in others. It can draft beautiful poems or compress tons of data, yet it can trip over simple “trick questions” or nuances…

Knowing both sides isn’t always intuitive. But it’s critical to apply AI where it adds the most value and keep humans in the loop where we matter most. So, I wrote this “guide” to help “manage some expectations” for both professionals and everyday users.

You can use these sections like a checklist to spot where you can (or should) lean on AI in your work, studies, hobbies etc. – and when not. Done right, AI makes you more efficient and creative. But misused, it wastes time, nerves and resources…

Even though discussions tilt toward “in vogue” tools like ChatGPT & Co., the article still applies broadly to all “flavors” of AI: traditional, generative, agentic etc. Ready?

Table of Contents

What AI Does Well

AI “feels home” in structured, well‑defined environments. Think coding, math, diagnosis, protein synthesis or even chess etc. Let’s check out how state-of-the-art AI can make your life easier – beyond running ~24/7 without coffee:

Sift noise from signal effortlessly

Long reports or big data piles overwhelm humans easily. AI compresses them into digestible insights – from predictive signals (e.g. quantitative forecasting) to language models summarizing the key points of long reads.

I use this “AI superpower” daily. For example, after workshops to condense messy whiteboard notes into crisp one-pagers with a simple ChatGPT prompt like: “Summarize [the whiteboard in the photo] in bullets and highlight potential follow‑ups.” This can also help with research papers, minutes/transcripts, brain dumps etc. Details can get “lost in compression”, so double-check important outputs.

Expand ideation efficiently

Creativity benefits from diverse angles. GenAI accelerates ideation cheaply and can even surface outside‑the‑box suggestions you may not see. This lets you explore and narrow down the “solution space” rapidly – without getting stuck at “blank page syndrome”.

For instance, I used AI to come up with new formats for communication campaigns, blog article topics, project risks etc. I once asked ChatGPT for 20 “creative” slogan ideas for a new product and refined one into the final version: “Ideate 20 options for [goal] with one-liner reasons for each.” Most ideas are weak but the value lies in instantaneous volume. The last call is always yours.

“Transform” content for different occasions

One message rarely works for all audiences. AI helps adjust tone, style, length etc. on demand while keeping the core intact. This lets you cheaply “repurpose” any materials for different situations or stakeholders without starting from scratch.

For example, I used GPT to turn a dense strategy paper into a briefing for a senior leader: “Rewrite [this doc] for a busy executive in <200 words; bullets; short active sentences”. That also works for multilingual content, reusing blog articles as social media posts (😉) etc. Cultural nuance, sensitive topics etc. of course need review.

Automate repetitive workflows tirelessly

Routine tasks often drain time and aren’t always enjoyable… AI – even simple, accessible tools like ChatGPT – help automate many of these processes. AI agents (like ChatGPT’s Agent Mode) autonomously do deep research on any topic (maybe AI trends?) and send you the summary like a newsletter. To me, handing such tasks off frees up scarce time for more interesting stuff.

As another example, I use ChatGPT to write all “alt texts” for images I upload to my blog: “Create an alt text description reflecting this [image’s] content and purpose in <120 chars.” But automation can also replicate errors (on steroids), so don’t blindly throw it at bad processes… Better re-think your (now AI-powered!) workflows – and stay in the loop for critical outputs.

Challenge biases with different perspectives

Plans can miss fatal blind spots if they are only seen from one lens. AI can broaden our horizon by “role‑playing” stakeholders and exposing what may remain hidden. That’s useful for interview prep, debate practice, market research (customer personas), overcoming our own biases etc.

Using GPT as a sparring partner to prepare project pitches is one of my favorite use cases. Practicing questions from CFOs, IT Security or other stakeholders, I can find (and close) any “holes”. Try this prompt: “Play devil’s advocate from the perspective/s [CFO, IT etc.] For each, highlight potential flaws in and suggest fixes for my pitch: […]” AI simulations, even if they feel realistic, only reflect training data – not lived experience – and can miss real human’s motives and expertise.

Lower barriers for designing and creating

Creative and technical tasks often need specialist skills. AI lowers the threshold for design, coding and even product creation widely. Simple consumer tools like Lovable and Claude Artefacts let everyone take an idea and shape it into something tangible now. Websites, apps, presentations, visuals, product demos etc. Endless possibilities… (Although I think AI outputs don’t have the heart to be true art. Unless humans build on and add meaning to it. Debate me.)

I experienced this power when I used Google’s AI Studio to create a fully working app – in one afternoon – that turns my LinkedIn posts into visual carousels. With limited TypeScript experience, I couldn’t believe my eyes. Prompts like “Help me create [a simple prototype] for [goal]. Provide code plus short, simple explanations.” get the ball rolling. While “vibe coding” promises speed, AI outputs are rarely perfect. AI-made software can contain vulnerabilities. Review properly before “shipping”!

What AI (Still) Struggles With

AI stumbles when problem spaces are messy, goals vague or boundary conditions change. (Yep, that reflects a big portion of “real life”…) Let’s check out some limitations and how to navigate them.

Don’t outsource critical thinking or learning

Real understanding forms when you wrestle with ideas yourself. AI can assist your learning by generating additional explanations, quizzes etc. But it won’t do the hard thinking for you. If you skip the struggle, the learning won’t stick. Research shows: “Use it or lose it” also counts for your mental muscles!

Letting AI answer your homework questions, for example, may save some time in the short term, but prevents skill development. Better: Let AI tools like e.g. NotebookLM turn course materials into flash cards. Then use these yourself to practice with study techniques like active recall. As Steve Jobs may beautifully put it: Use AI as a bicycle for your mind – not to substitute your brainpower.

Fact-check – AI outputs probabilities, not guarantees

Language models generate what’s likely, not what’s “true”. Hallucinations – aka “AI confidently outputting falsehoods” – (and knowledge cut‑offs) are common pitfalls. Mind that what may look like human “reasoning” is just high‑powered pattern matching under AI’s hood.

I learned this the hard way when I asked ChatGPT for a travel itinerary in Dublin during a business trip… At first glance, everything looked plausible – until I stood in front of a supposed historical cathedral… that simply doesn’t exist. While technical approaches like “RAG” and “chain-of-thought” can improve AI accuracy, there is no perfect solution yet to fully prevent hallucinations. To derisk, consider working with acceptance rules like “no claims without sources”.

Don’t hand over high‑stakes or precision calls

Some decisions need more than speed – they require accountability and accuracy. Areas like health, law, finance or hiring fall in this category. AI can’t take responsibility; it doesn’t wrestle with “what should I do” in grey zones. People do. Generative AI also tends not to be very precise given its probabilistic nature… While AI can surely draft or suggest, the final call stays with humans when the cost of a wrong answer is too high.

For instance, an innovative HR tool screening candidates may seem efficient, but left unchecked it will just amplify bias beyond good and evil. Injustice. Massive lawsuits. These socioeconomic risks are solid reasons why e.g. the EU AI Act severely regulates such use cases. Use AI only in a “supporting role” for sensitive topics. Don’t let the sweet comfort of automation lead you down an ethical slippery slope…

What’s Next – Changing Frontiers

Today’s possibilities aren’t carved in stone. As you likely noticed, AI is evolving erratically – unlike any other technology really. Tasks impossible half a year ago may already work today etc.

While I don’t have a magic 8-ball here, I foresee major developments along these 3 lines/buckets…

1. Model improvements: Capability advances of AI models themselves. Maths and logic are becoming more reliable. Context windows grow “from pages to bookshelves”, while reducing the “needle in a haystack” problem. “Omnimodality” as text, image, audio, video, “world simulations” etc. improve and blend more smoothly.

2. Systems integration: AI is no longer about a single model. “Compound systems” already connect models with databases, APIs and web access. Agentic AI orchestrates increasingly sophisticated software workflows autonomously. Robotics link AI models (“brains”) with physical machines (which could soon have their own “ChatGPT-moment”). Altogether, AI gets woven deeper and deeper into our everyday lives.

3. Emergent behavior & human learning: With continued experimentation, we discover a) new capabilities of AI systems and b) how to apply AI better. Transfer of use cases across domains (e.g. from customer service to HR) becomes more common. We’re gradually learning how to adopt AI meaningfully in our hobbies, work, education etc. (That’s basically my full-time job…)

The distant horizon is “AGI” (aka human-level artificial intelligence) – still aspirational but a key driver of frontier research. (The upcoming advances in robotics could be a key enabler for this – e.g. due to “richer” feedback loops and learning from sensors and physical actions etc.)

Conclusion & Tips for Smarter Human-AI Collabo

In a nutshell, AI is strong at the “what” – predicting, generating, transforming “things” etc. But humans own the “so what” – with context, intuition, empathy etc. This contrast is an opportunity: where AI falters, our brains “coincidentally” tend to excel (and often vice versa!) I think of it that way: The value of my work isn’t shrinking but shifting. (Check out my primer on Human-AI Collaboration and how to adapt to these changes.)

Let’s finish with 3 practical takeaways for you:

1. Complement each other – know AI’s strengths as well as your own; use it to extend, not replace. My “rule-of-thumb workflow”: my idea → AI contributes → my final touch. You own the results and your mental muscles stay fit.
2. Stay in the loop – AI’s potentials are a “moving target”, so try keeping up with the latest trends. I’ve shared tips on filtering noise from signal efficiently in another piece.
3. Experiment individually – AI isn’t one‑size‑fits‑all; it depends on your personal context, goals, skills, values etc. So, keep testing tools for your own tasks and interests. For inspo, check out my (free) library of ChatGPT use cases/prompts.

If this “the art of the possible” was helpful for you, please share it with someone who’s figuring out how to make the best use of this tech. What are your thoughts on those (or other) “wins” and “sins” of using AI? Drop a comment below or get in touch.

Cheers,
John