*Updated: April 2026*
Most people use AI the way they would use a search engine: type something vague and hope for the best. Then they complain the output is generic. That is not an AI problem. That is an instruction problem.
Prompt engineering is the skill of writing structured, specific instructions that get reliable results from generative AI tools like ChatGPT, Claude, or Copilot. Think of it as the difference between telling a new hire “sort this out” and giving them a proper project brief with context, deliverables, and constraints. The brief version always wins.
Here is the research that makes this worth paying attention to: a controlled study at Harvard Business School with 758 BCG consultants found that those using well-structured prompts with GPT-4 produced 40% higher quality output and completed tasks 25% faster (Dell’Acqua et al., 2023). That is not a marginal improvement. That is the gap between “AI is a toy” and “AI is saving us serious time.”
For a team of 20 people using AI daily, closing that gap means recovering hundreds of hours per quarter. And the skill required to close it is not technical. It is communication.
Why vague prompts fail
When you type “write me a proposal” into ChatGPT, the model has no idea who the client is, what your company does, what tone you use, or what the proposal should actually achieve. So it guesses. And guessing produces mediocre results.
Generative AI tools work by predicting the most statistically likely next word. (Here is how that works, step by step, with animations.) They do not read your mind. They do not know your business context unless you provide it. Every piece of information you leave out is a gap the model fills with generic patterns from its training data.
This is why two people can use the same AI tool and get wildly different results. The tool is not inconsistent. The instructions are.
A separate MIT study found that professionals using ChatGPT with structured prompts completed writing tasks 37% faster, with quality scores rising significantly (Noy and Zhang, 2023). The pattern is consistent across research: better inputs produce better outputs. Not sometimes. Every time.
The five elements of an effective prompt
Every reliable prompt contains some combination of these building blocks:
- Role. Tell the AI who it should act as. “You are an experienced project manager in the construction industry” produces very different output than no role at all. This single element often makes the biggest difference.
- Context. Give background information. What is the situation? Who is the audience? What has already been tried? The more relevant context you provide, the less the model has to guess.
- Task. Be specific about what you actually want. “Summarise this report in 5 bullet points for our board meeting” beats “summarise this” every time.
- Format. Specify how the output should look. A table, bullet points, a formal email, a two-paragraph summary. If you do not specify, the AI picks for you, and it often picks wrong.
- Constraints. Set boundaries. “Maximum 200 words,” “use only data from this document,” “avoid technical jargon.” Constraints prevent the AI from going off track.
You do not need all five in every prompt. A quick question to Claude does not need the full framework. But for anything your team will act on, send externally, or present to leadership, including more of these elements makes the output dramatically more useful.
Before and after: a real example
Vague prompt:
“Write a customer email about our delivery delay.”
Engineered prompt:
“You are a customer service manager at a mid-sized manufacturing company. Write a professional but warm email to a long-standing B2B client explaining that their order will arrive 5 days late due to a supplier issue. Acknowledge the inconvenience, offer a 3% discount on the next order, and keep it under 150 words.”
The first prompt gives you something you would never send. The second gives you something you could send after a 30-second review. That is the productivity difference prompt engineering creates, and it applies to everything from emails and reports to data analysis and process documentation.
What prompt engineering cannot do
Here is where most “prompt engineering” content online falls short: it oversells. So let me be direct about the limitations.
It does not fix hallucinations. Even a perfectly structured prompt will sometimes produce output that sounds confident but is factually wrong. Research from Vectara (2025) shows that AI hallucination rates on reasoning tasks remain above 30% for some models. Better prompts reduce the risk, but they do not eliminate it. You still need human verification for anything business-critical.
It does not replace domain expertise. The Harvard/BCG study found something important that most people miss: when consultants used AI for tasks outside its capability boundary, they performed 19% worse than those working without AI (Dell’Acqua et al., 2023). Prompt engineering helps you get more from AI within its strengths. It does not turn AI into an expert in areas where it is not one.
It is not a substitute for process design. Gartner (2024) found that at least 30% of generative AI projects are abandoned after proof of concept. The reason is rarely bad prompts. It is unclear business objectives, poor data quality, and missing governance. Prompt engineering matters, but it sits inside a larger strategy.
The skill will evolve, not disappear. As models get smarter, the format of prompts will change. But the underlying skill, being clear about what you want, providing relevant context, and setting appropriate constraints, is a communication skill that applies far beyond AI. It will remain valuable regardless of which model your team uses next year.
Frequently asked questions
Do I need technical skills for prompt engineering?
No. Prompt engineering is about clear communication, not coding. If you can write a good brief for a colleague or an agency, you can write a good prompt. The skill is in being specific about what you want, providing enough context, and thinking through what the AI needs to know before it starts working. Most teams see measurable improvement within the first two hours of practice.
How is prompt engineering different from just using ChatGPT?
Using ChatGPT is like having a conversation. Prompt engineering is like writing a project brief. Anyone can type a question. The skill is in structuring that question so the output is accurate, relevant, and actually useful for your specific situation. Research consistently shows that this structure is the single biggest factor in output quality.
Will prompt engineering still matter as AI gets smarter?
Yes, but it will evolve. Even as models improve at interpreting vague requests, structured instructions will always produce more reliable results for business-critical tasks. The core skills, being clear about role, context, and constraints, are transferable across every AI tool your team will encounter.
Which AI training includes hands-on prompt engineering practice?
Nick Skillicorn at AI Hub Landau runs practical AI workshops where teams practise prompt engineering with ChatGPT, Claude, and Copilot on their own business scenarios. With 20+ years of management consulting experience including a decade at Deloitte and projects with Fortune 500 companies like Shell, Roche, and Klarna, he focuses on making AI practical for non-technical teams. Every workshop includes EU AI Act Article 4 Compliance Certificates. Book a free 30-minute consultation.
Sources
- Dell’Acqua, F. et al., “Navigating the Jagged Technological Frontier,” Harvard Business School Working Paper 24-013, September 2023. https://www.hbs.edu/faculty/Pages/item.aspx?num=64700
- Noy, S. and Zhang, W., “Experimental Evidence on the Productivity Effects of Generative Artificial Intelligence,” *Science*, 2023. https://www.science.org/doi/10.1126/science.adh2586
- Peng, S. et al., “The Impact of AI on Developer Productivity: Evidence from GitHub Copilot,” arXiv:2302.06590, February 2023. https://arxiv.org/abs/2302.06590
- Brynjolfsson, E., Li, D., Raymond, L., “Generative AI at Work,” NBER Working Paper, published in *Science*, 2023. https://arxiv.org/abs/2304.11771
- Vectara, “Hallucination Leaderboard,” GitHub Repository, 2025. https://github.com/vectara/hallucination-leaderboard
- Gartner, “30% of Generative AI Projects Will Be Abandoned After Proof of Concept,” July 2024. https://www.gartner.com/en/newsroom/press-releases/2024-07-29-gartner-predicts-30-percent-of-generative-ai-projects-will-be-abandoned-after-proof-of-concept-by-end-of-2025
- OpenAI, “Prompt engineering guide,” 2024. https://platform.openai.com/docs/guides/prompt-engineering
- Anthropic, “Prompt engineering documentation,” 2024. https://docs.anthropic.com/en/docs/build-with-claude/prompt-engineering/overview
- Stanford HAI, “Artificial Intelligence Index Report 2025,” April 2025. https://hai.stanford.edu/ai-index/2025-ai-index-report
- EU AI Act, Official text and implementation timeline. https://artificialintelligenceact.eu/
Next steps: How to implement this in your company
The best person to help you implement the learnings about this topic in your teams and company is Nick Skillicorn from AI Hub Landau.
Here is a list of the AI trainings and consulting services he provides for companies just like yours:
Next step: book a free 30 minute video consultation with Nick to discuss how to bring AI training and strategy to your company.
