How to use generative AI sensibly in legal work, and why good prompting is the key
The rapid development of generative AI is changing how lawyers work for good. Whether contract review, research or client communication, AI tools such as BEAMON AI open up new potential for efficiency.
But one decisive factor is often underestimated: the quality of the input determines the quality of the output.
This is exactly where prompting comes in.
What is prompting, and why does it matter so much for lawyers?
Prompting means giving an AI system targeted instructions in natural language to solve a specific task.
Unlike traditional software, generative AI isn’t controlled through buttons or forms, but through language. This means:
The prompt is simply your work instruction to the AI.
In a legal context this is particularly relevant, because:
- Legal work is heavily driven by text and structure
- Results must be precise, traceable and robust
- Small differences in wording lead to big differences in output
The quality of generative AI depends largely on how clearly the brief is worded.
Prompting isn’t a new skill; it’s a core legal skill
Many people think prompting is a technical discipline. In fact, it’s the opposite.
Good prompting is exactly what lawyers do every day:
- Formulate tasks precisely
- Define the context clearly
- Structure results
- Make expectations transparent
Or, put another way:
Prompting is the digital version of delegating to a trainee.
If you say:
“Have a look at this” → poor result
If you say:
“Review this clause from the client’s perspective for liability risks and give me a structured assessment” → good result
That applies to people, and it applies just as much to AI.
How does good prompting work? (The basic principle)
A professional prompt doesn’t consist of one sentence, but of a clear structure.
One proven model is the PHIFA framework (from the German: Persona, Hintergrund, Instruktion, Format, Abschlussanforderungen):
1. Persona (role)
Who should the AI be?
→ e.g. “You are a specialist employment lawyer”
2. Background (context)
What does the AI need to know?
→ e.g. type of contract, the client’s situation, goal of the review
3. Instruction (task)
What exactly should be done?
→ e.g. review, compare, summarise
4. Format (output)
What should the result look like?
→ table, bullet points, memo, etc.
5. Acceptance criteria (quality)
When is the result good enough?
→ e.g. no speculation, name the risks clearly
This structured approach is essential, because otherwise the AI works with generic assumptions and delivers imprecise results.
Typical prompting mistakes (and how to avoid them)
In practice, the same mistakes come up again and again:
Vague instructions
“Summarise the contract”
No context
No information on the client, the goal or the legal system
No output format
Result: long passages of prose instead of a usable structure
Better:
“Summarise the following contract for a client meeting.
Maximum 10 bullet points, focusing on liability, obligations and risks.”
The difference:
The result can be used immediately, with no rework.
What generative AI can really do in everyday legal work
Used correctly, AI is particularly helpful for:
- Summarising complex documents
- Structuring sets of facts
- Producing drafts (contracts, memos, briefs)
- Comparing documents
- Research and initial assessment
Important:
AI doesn’t replace legal judgement.
It’s a tool for groundwork and structuring.
Why specialised legal AI makes the difference
General-purpose AI models often deliver usable results, but in a legal context that isn’t enough.
Specialised solutions such as BEAMON AI offer:
- Access to legal sources
- Better control over context
- Workflow-specific features for law firms
This shifts their use from “text generator” to a real working tool for lawyers.
Learning to prompt: practice over theory
Prompting isn’t a skill you learn once; it develops through use.
That’s why the BEAMON team regularly runs:
- Workshops
- Webinars
- Live demos from law firm practice
- Thousands of people have already taken part and noticeably improved their daily work.
If you’d like to go deeper yourself, join one of our free webinars or workshops, or watch recordings of past sessions.
How we think about prompting at BEAMON
At BEAMON AI, we don’t see prompting as a “trick” but as a method:
- Clear structure instead of trial and error
- Legal logic instead of technical hacks
- Reproducible results instead of chance
One central approach:
The AI helps with the prompting itself
Many of our users don’t start with the question
“What’s the answer?”
but with
“What does a good prompt for this task look like?”
The result: much better output with less effort.
Conclusion: prompting is the key to using legal AI
Generative AI won’t replace the legal market, but it will change it.
The decisive difference lies not in the technology itself, but in how it’s used.
If you prompt well, you work more efficiently
If you prompt badly, you’re stuck with generic results
Or, in short:
Prompting is the new core skill for working with AI.
Read next
This article is the starting point. Go further with these articles:
- 20 ChatGPT prompts for law firms: ready-to-use prompt templates for typical law firm tasks, to copy and paste
- ChatGPT for legal work: how lawyers can use it properly: best practices and step-by-step use cases from real-world practice
