📋 What Professional Context Is
Professional context and the set of information that turns a generic AI into one that understands your practice. It includes who you are, what you do, who you work for, how you communicate, and which rules you follow. Without context, the AI responds as if it were talking to anyone on the planet. With context, it responds as if it had worked with you for months and knew every detail of your office, practice, or studio.
Think of context as the "field manual" you give a new employee. Without it, they have to ask everything: "who are our clients?", "how do we reply to emails?", "what's the standard format for an expert report?" With the manual, they can start producing right away. AI works the same way. The difference is that you write this manual once and use it forever.
less rework
set it up, use it regularly
consistent
to put together
🗃️ Your essential information
The professional brief is the document you give your assistant with everything it needs to know about your practice. It doesn’t need to be long. It needs to be specific. Most professionals make the mistake of providing too much information without structure or too little without enough detail. The key is to organize it into 6 blocks clear and objective.
Identity
Your name, education, professional registration (CRM/OAB/CRC/CREA), specialties, and years of experience. Example: "I’m Dr. Maria Santos, a dermatologist, CRM-SP 123456, with 15 years of clinical practice focused on cosmetic and oncologic dermatology."
Specialty
Detailed areas of practice, procedures you perform, methodological approach. E.g.: "I practice family law, focusing on contested divorce, shared custody, and probate. I take a collaborative approach whenever possible."
Target audience
Who your typical clients/patients are, their age range, socioeconomic profile, and common concerns. Example: "My patients are women ages 35-60, upper-middle-class, concerned about aging and skin spots."
Differentiators
What sets you apart from your peers, your work philosophy, awards, and publications. Example: “What sets me apart is my integrative approach: I combine traditional accounting with strategic tax planning for small and midsize businesses.”
Restrictions
What the assistant must never do: diagnose, give definitive legal advice, promise results, cite data without a source. Example: “Never claim that a treatment is guaranteed. Always include ‘schedule an in-person consultation for a complete evaluation.’”
Common cases
The 5–10 most common situations in your week. E.g., "Most frequent requests: reviewing articles of incorporation, starting a MEI business, estate planning, family holding companies, tax consulting."
💡 Practical Tip
Don't try to fill in all 6 blocks at once. Start with the 3 required ones (Identity, Specialty, Audience) and add the others as you notice the assistant needs them. The briefing is a living document that evolves with you.
🎨 Voice and style
O tone of voice is what makes AI’s writing sound like you, not a robot. If you’re a formal attorney, AI needs to write with legal formality. If you’re an approachable nutritionist, it needs to write in a relaxed, educational way. The wrong tone is the #1 reason professionals give up on AI: "it doesn’t sound like me." Getting the tone right saves more time than any other optimization.
✓ Calibrated tone
- ✓"Dr. Silva, as we discussed at yesterday’s appointment, here’s the revised treatment plan."
- ✓"Dear Ms. Costa, please be advised that the appeal deadline ends in 15 business days."
- ✓"Hi, Marina! This week’s workout focuses on core strengthening. Ready?"
✗ Generic AI tone
- ✗"Of course! Here’s a treatment plan for you. I hope it’s helpful!"
- ✗"Absolutely! The appeal deadline is important. Be sure to keep track of deadlines."
- ✗"Here’s a core-focused exercise program. Have a great workout!"
To calibrate the tone, use the technique of "write like this text". Pick 2-3 emails or messages you’ve written that represent your style well. Paste them into the system prompt with the instruction: "Use these texts as a reference for tone and style." The AI will absorb your vocabulary, favorite expressions, and writing rhythm.
📋 Prompt to Calibrate Your Tone of Voice (Copy and Use)
Analyze these 3 texts I wrote and identify my communication style: TEXT 1: [paste an email you sent to a client] TEXT 2: [paste a professional WhatsApp message] TEXT 3: [paste an excerpt from a report/opinion] Based on these texts, identify: 1. Level of formality (1-10) 2. Words and expressions I use frequently 3. Typical sentence length 4. How I start and end messages 5. Punctuation and formatting style Summarize this in one paragraph of instructions that I can paste into my assistant's system prompt.
💡 Tip: the Formal-Informal Spectrum
Choose a number from 1 to 10 for the assistant. 1 = ultra-informal (slang, emoji). 5 = approachable and professional. 8 = formal and technical. 10 = legal/academic language. Add this to the system prompt: "Use a formality level of 7." That alone solves 60% of tone problems.
📚 Knowledge Base
A knowledge base and the collection of documents you provide to the assistant for it to consult when generating content. Instead of making up information, it uses your own templates, standards, and references. This changes everything: the assistant stops being a generic text generator and becomes a writer who knows your office's standards.
📄 Template documents
- Standard contracts and draft agreements
- Opinion or expert report templates
- Business proposal templates
- Reference emails by type
📖 Standards and Regulations
- Professional code of ethics
- Technical standards (ABNT, CFM, OAB)
- The firm's internal regulations
- LGPD applied to your field
📊 Business data
- Price and fee schedule
- FAQ answered for clients
- History of common cases
- Standard customer service scripts
How to do it in practice: in ChatGPT, create a custom GPT and upload the files in the "Knowledge" section. In Claude, use "Projects" and add documents as project context. In Gemini, attach files directly to the conversation. Each platform has its own method, but the principle is the same: give the assistant access to your actual references.
💡 Context limits by platform
ChatGPT-4o (tokens)
~96 thousand words
Claude (tokens)
~150 thousand words
Gemini (tokens)
~750 thousand words
🔧 Fine-tuning
No assistant is perfect on the first try. The fine-tuning is the process of identifying and correcting unwanted behavior. Does the assistant talk too much? Use the wrong jargon? Make up data? Each problem has a specific solution in the system prompt. Knowing how to diagnose issues is the skill that separates those who give up ("AI doesn’t work") from those who optimize until they get real results.
| Problem | Diagnosis | Solution in the prompt |
|---|---|---|
| Response is too long | No length limit | "Responses of no more than 200 words unless I ask for more detail." |
| Wrong tone | No style reference | "Use a formal, technical tone, level 7/10. Avoid slang and exclamation points." |
| Makes up data | No hallucination constraint | "Never cite statistics without a source. If you don’t know, say you don’t know." |
| Inappropriate format | No output format | "For reports, use: Header > Analysis > Conclusion > Recommendation." |
| Ignore ethical restrictions | Missing or weak rule | "NEVER suggest a diagnosis. Always include: 'consult a professional'." |
📋 Diagnostic Prompt (Copy and Use)
I'll send you 3 answers you gave me recently. For each one, evaluate: 1. Is the tone appropriate for my professional profile? (1-10) 2. Does the format follow the standard I defined? 3. Is any information made up or inaccurate? 4. Is the length appropriate? 5. What would you change if you could do it again? Respond in table format. At the end, suggest 3 specific adjustments to my system prompt that would fix the problems you found.
⚠️ Attention
Fine-tune iteratively. Don't try to fix 5 problems at once. Fix one, test it, confirm that it improved, then move on to the next. If you change too many things at once, you won't know what worked and what got worse.
🚀 Perfect context checklist
Use this checklist to audit your professional context. Even if the assistant is already working well, going through each item ensures you haven’t forgotten anything important. Divide items into 3 priority levels: required (it won't work without this), recommended (greatly improves quality) and optional (for advanced refinement).
★ Required (start here)
- ☑Professional identity (name, registration, primary specialty)
- ☑Target audience (who your typical clients/patients are)
- ☑Voice (level of formality, communication style)
★★ Recommended (makes a big difference)
- ☑Ethical and regulatory restrictions of the profession
- ☑Default output format (how you want the responses)
- ☑Template documents (contracts, opinions, reference reports)
- ☑Words and Expressions to Use and Avoid
- ☑Common cases and how to handle each one
★★★ Optional (advanced refinement)
- ☑Competitors and how to stand out from them
- ☑History of decisions (case law, previous cases)
- ☑Internal office approval workflows
- ☑Examples of your writing for detailed style reference
📋 Context Audit Prompt (Copy and Use)
Analise o meu prompt de sistema atual e avalie cada item abaixo de 0 a 10: 1. Identidade profissional esta clara? 2. Publico-alvo esta definido com especificidade? 3. Tom de voz esta calibrado (formal/informal)? 4. Restricoes eticas estao explicitas? 5. Formato de saida esta padronizado? 6. Base de conhecimento esta referenciada? 7. Palavras/expressoes a usar e evitar estao listadas? 8. Casos frequentes estao cobertos? Para cada item abaixo de 7, sugira uma frase especifica que eu possa adicionar ao prompt de sistema para melhorar. Responda em formato de tabela com colunas: Item | Nota | Sugestao.
📅 Monthly review
Set aside 15 minutes each month to review the checklist. Your practice evolves, your clients change, and new regulations emerge. The context that was perfect 3 months ago may be out of date today. Monthly reviews ensure the assistant keeps up with your professional growth.