Most people I train have the same story: they've used ChatGPT a few times, maybe asked it to write an email, maybe tried and failed to get it to do something useful for their job. They know AI is important. They just don't know where to start.
Here's where to start.
These are the tips I give in every training session. They're not theoretical. They're what I've seen work with lawyers, marketers, product managers, photographers, copywriters, compliance directors, and founders. The stuff that actually changes how you work.
Understand What the Machine Is Doing
Before you learn any tips or tricks, understand the mechanism. LLMs are autocomplete machines. They predict the next word based on everything they've seen during training. They don't think. They don't know things. They predict.
This matters because it changes your expectations. You stop asking "why did it get this wrong?" and start asking "what input would produce a better prediction?" That shift — from expecting perfection to engineering better inputs — is the single most important mindset change you can make.
If you want the full explanation, read my What is an LLM? page. But the short version: treat it like a brilliant but overconfident junior colleague who needs clear briefings and careful review.
Start With One Problem, Not One Tool
Don't start by exploring tools. Start by looking at your actual workday and finding the task that's most annoying, most repetitive, or most time-consuming.
Is it writing first drafts of client emails? Summarising meeting notes? Researching competitors? Reformatting data between spreadsheets? Those are your AI opportunities.
Once you have the problem, finding the right tool is easy. Working backwards from tools to problems is how people end up with six subscriptions they never use.
Pick One Tool and Go Deep
You don't need ChatGPT and Claude and Gemini and Perplexity. Not yet.
Pick one. I'd suggest Claude (claude.ai) or ChatGPT (chatgpt.com) — both have generous free tiers and handle most tasks well. (Not sure which? See my Best LLMs in 2026 comparison.) Spend two weeks using it for everything: drafting, research, brainstorming, analysis. Push its limits. Find where it's brilliant and where it falls flat.
You'll learn more from two weeks with one tool than from dabbling with ten.
Write Better Prompts (The Anatomy That Works)
The quality of your output is directly proportional to the quality of your input. Every good prompt has the same anatomy:
Role and context. Tell the AI who it should "be" — a senior marketing strategist, a legal researcher, a financial analyst. But equally important: tell it who you are. "I'm a marketing director at a mid-size B2B SaaS company" or "I'm a solicitor specialising in employment law" gives the model the context it needs to calibrate its language, depth, and assumptions. Without this, you get generic answers. With it, you get answers that actually fit your situation. This isn't roleplay; it's priming the model to draw on the right patterns for the right audience.
Task and constraints. Be specific about what you want: "draft," "compare," "summarise," "debug." Include constraints: tone, length, format, date range.
Input data. Give it the raw material: facts, URLs, text snippets, documents. The more context you provide, the less the model has to guess.
Output format. Say exactly what you want back: "a bullet-point summary," "a table with three columns," "a 200-word email." This eliminates the reformatting loop.
The difference between a vague prompt and a well-structured one is the difference between a mediocre output you throw away and a first draft you can actually use. I've seen this cut iterations by half.
Use Projects to Keep Context
One of the most underused features across all AI tools: Projects (in Claude) or Custom GPTs (in ChatGPT).
Instead of starting every conversation from scratch, create a project for recurring work. Upload your brand guidelines, your tone of voice document, your client brief. Set instructions like "always write in British English" or "our brand never uses exclamation marks." The AI remembers this across every conversation in that project.
This is the difference between an AI that knows nothing about you and an AI that works like a team member who's read the brief. Set this up once and every interaction gets better.
Iterate — Don't One-Shot
Your first prompt will rarely produce the perfect output. That's fine. The model is essentially free to use — treat it like a conversation, not a vending machine.
Ask for a first draft. Read it. Tell the model what you liked and what needs to change: "Good structure, but the tone is too formal. Make it conversational, and add a specific example in paragraph two." Ask it to try a different angle. Ask it to play devil's advocate on its own output.
The people who get the most out of AI are the ones who have three-turn conversations, not one-shot prompts. The cost is zero. The improvement is significant.
Always Verify Important Outputs
LLMs hallucinate. They will state false things with complete confidence. This isn't a bug that's getting fixed — it's a fundamental consequence of how prediction works.
For low-stakes tasks (brainstorming ideas, drafting a first version, reformatting text), this doesn't matter much. For anything that needs to be factually correct — legal research, financial figures, medical information, public-facing content — always verify.
Use research tools like Perplexity that provide source citations. Cross-reference claims with original sources. Never publish or send anything AI-generated without a human review pass. I say this in every session: AI gives you speed. Verification gives you trust. You need both.
Learn to Use Files, Not Just Text
Most people type prompts and read responses. That's level one.
Level two: upload files. Every major AI chatbot now accepts PDFs, spreadsheets, images, and documents. A lawyer can upload a contract and ask "what are the key liabilities for my client in this agreement?" A marketer can upload a competitor's annual report and ask "summarise their strategic priorities for next year." A small business owner can upload last quarter's P&L and ask "what trends should I be concerned about?"
The moment you start giving AI your actual work documents instead of typing descriptions of them, the quality of outputs jumps dramatically.
Build Habits, Not Just Skills
The people who get lasting value from AI aren't the ones who attend a training session and then go back to working the old way. They're the ones who build AI into their daily habits.
Start small. Every morning, use AI to summarise the news relevant to your industry. Before every meeting, use it to prepare talking points from the agenda. After every meeting, paste in your notes and get a clean summary with action items. Before sending any important email, run it through AI for a tone and clarity check.
None of these take more than two minutes. After two weeks, they'll feel automatic. After a month, you won't remember how you worked without them.
Don't Wait Until You "Need" It
The worst time to learn AI is when you desperately need it — a deadline is looming, a competitor has pulled ahead, your new hire is running circles around you with tools you don't understand.
The best time is now, when you have the space to experiment, to fail, to learn at your own pace. AI literacy isn't a one-off skill. It's an ongoing practice, like staying fit. The tools change every few weeks. The people who stay current are the ones who play with new releases, who read what's changing, who keep experimenting.
Organisations should be preparing now for graduates who've used LLMs throughout their entire degree. If your company doesn't have AI tools integrated into workflows, these new hires will be baffled — and inefficient. The skills gap is real. But it's closeable, if you start.
The Fastest Way to Get Started
Reading tips is useful. Having someone show you — on your work, with your tools — is faster.
I run 90-minute 1:1 AI training sessions. In the first 30 minutes, I explain how these tools actually work (the real version, not the marketing version). Then we spend 60 minutes building: your prompts, your workflows, your actual problems.
A recent client was leaving insurance to start a storage business. In one session we used deep research to assess locations and competitors, then built a marketing landing page before the 90 minutes were up. That's not a hypothetical. That's a Tuesday.
You'll leave with a recording, a transcript, working prompts, and a clear understanding of which tools solve your specific problems. Not someone else's problems. Yours.
Book a Session →Written by Riz Pabani, AI Trainer based in London. MIT AI Certified, 20+ years in technology. I help business leaders and professionals cut through the AI hype and focus on what actually works.
Related: What is a Large Language Model? | Best LLMs 2026 | Best AI Tools 2026