Effective Communication with AI
Jump to section
Communication, not prompting
The word 'prompt' is misleading — it suggests there is one magic input that extracts the perfect response. Reality is different. Effective AI work is a conversation: giving input, evaluating the response, refining, iterating. Like working with a very capable colleague who does not know the context of your situation.
Three pillars of effective communication
1. Context: what AI needs to know
AI does not know who you are, what you do, who the work is for, and what you have already tried. The more relevant information you provide, the better the response. Not more words — more RELEVANT information.
Bad:
"Write me an email to a client."
Good:
"I'm a project manager at a web development agency.
The client (a marketing agency, 20 people) asked us
to redesign their e-commerce site. We already have
an approved proposal for 500,000 CZK.
I need an email inviting them to a kickoff meeting next week.
Tone: professional but friendly. Length: 5-8 sentences."2. Framing: HOW you ask
You can ask the same question ten different ways — and get ten different answers. Framing determines the direction of the response. Key techniques: define a role ('as an experienced lawyer...'), specify format ('5 points, each max 2 sentences'), set the level ('explain to an absolute beginner').
3. Iteration: refining the response
The first response will rarely be exactly what you want. And that is fine. The most powerful technique is iteration: 'Good start, but (1) shorten each point to one sentence, (2) add specific numbers, (3) make the tone more formal.' Each iteration refines the result.
When the AI response 'is not good', do not delete everything and start over. Tell AI exactly WHAT is wrong and HOW you want it changed. AI learns from your feedback and the next iteration will be better.
Save your best prompts in a personal library (a simple text file or note). When you find a framing that works well for a specific task, record it as a template. Over time, you build a personal toolkit that saves hours of prompt crafting.
Advanced techniques
Examples (few-shot)
Instead of explaining what you want, show an example. 'Here is an example of how I want the email to look: [example]. Now write a similar email for this situation: [situation].' AI understands format, tone, and structure from examples better than from descriptions.
Role and persona
'You are an experienced financial analyst with 15 years of banking experience. Your client is the CEO of a small tech company...' The role changes AI's perspective — it uses different vocabulary, different level of detail, and different assumptions.
Constraints and limitations
Constraints improve quality. 'Max 200 words.' 'Use only data from this document.' 'No jargon.' 'Answer only yes/no with a one-sentence justification.' The tighter the frame, the more precisely AI works.
Common communication mistakes
- Too vague: 'Help me with a presentation' — about what? for whom? what format?
- Too many instructions at once: give AI one task, not ten
- Ignoring context: AI does not know what you meant — it must see it explicitly
- No examples: show what you want instead of describing it at length
- Giving up after the first answer: iteration is key — the first response is always a draft
Pick a real task from your work (email, report, analysis). Follow these steps: 1. Write the first AI request — no context, just the basic requirement 2. Evaluate the response — what is good, what is not 3. Add context and proper framing. Compare with response 1. 4. Iterate 2-3 times — each iteration refines a different aspect 5. Record how many iterations you need for a satisfactory result Goal: understand that iteration is not failure — it is the process.
Hint
Notice how dramatically quality improves between version 1 (no context) and version 2 (with context). Often the difference is between 'unusable' and 'almost done'.
Choose a single business question (e.g., 'Should we expand into the German market?'). Ask AI the same question 3 times, each time with a different role: 1. 'You are a conservative CFO focused on risk management' 2. 'You are an ambitious head of growth focused on market expansion' 3. 'You are an experienced operations manager focused on execution' For each response: (a) Note the key recommendation (b) Compare the vocabulary and tone (c) Identify which perspectives you would have missed without role switching (d) Write a 3-sentence synthesis combining the best insights from all three
Hint
Role-based prompting is one of the most powerful techniques because it forces AI to adopt different mental models. Use it whenever you need a multi-perspective analysis of a complex decision.
Take a single writing task (e.g., summarize a meeting, draft a proposal outline). Write the same request 4 times, each time with different constraints: 1. No constraints — just the basic request 2. Format constraint: 'Use bullet points, max 5 items, each max 15 words' 3. Audience constraint: 'Write for a non-technical executive who has 2 minutes' 4. All constraints combined plus a style example Compare all 4 outputs: (a) Which is most useful in practice? (b) How did each constraint change the output? (c) Which single constraint had the biggest positive impact?
Hint
Most people find that adding audience context ('who is this for?') produces the single biggest quality improvement. Format constraints are second. Combining both is where the magic happens.
- Effective AI work is a conversation, not a one-shot prompt
- Three pillars: context (what AI needs), framing (how you ask), iteration (refining)
- Examples work better than descriptions — show what you want
- Constraints (max words, format, level) improve output quality
- The first response is always a draft — iteration is the key to good results
In the next lesson, we dive into AI-Augmented Research and Analysis — a technique that gives you a clear edge. Unlock the full course and continue now.
2/7 complete — keep going!