Shopping cart

Your cart is empty

Total:
£0.00
continue booking

News

M365 Masters Conference

The Microsoft roadmap for 2026
17-21 August-Online

read more
Become a Microsoft TRAILBLAZER!

Learn jaw-dropping Microsoft tips!

read more
Training Options

Choose the best training solution for you!

read more
20-01-2026

Five Custom Instructions That Transform AI for Busy EAs

The Custom Instructions Every EA Should Use To Make AI Work Like an Assistant

The Custom Instructions Every EA Should Use To Make AI Work Like an AssistantExecutive Assistants do not have the luxury of wasting time. When you support a senior leader who receives more than one hundred emails a day, attends back‑to‑back meetings and expects information to be prepared quickly and accurately, every minute counts. AI can help, but only if it understands how you and your executive work. Without the right setup, AI produces vague summaries, generic drafts and inconsistent structure that can cost more time to fix than it saves.

This is where custom instructions become indispensable. When configured correctly, custom instructions teach an AI system to behave far more like a junior assistant who knows your preferences, recognises your executive’s style and understands your workflow. Once set, these instructions improve every email draft, every summary, every set of minutes and every briefing note you generate.

For EAs, custom instructions are not optional. They are the foundation that makes AI reliable.

Why AI output often disappoints EAs

AI systems are powerful, but they can only work with the information they are given. Most assistants try AI once, give it a short prompt like “Summarise this” or “Write an email about this topic” and receive something bland, incomplete or totally mismatched to the executive’s expectations.

The problem is not the EA.

The problem is that the AI has no sense of tone, role, structure or context by default.

For example:

  • It does not know that your executive hates waffle.
  • It does not know that you prefer bullet points for internal briefs.
  • It does not know that your leadership team expects clear next steps.
  • It does not know which messages should sound firm and which should sound diplomatic.

Custom instructions solve this instantly by giving AI a permanent profile of how you work and what you need.

The four essential areas your custom instructions must cover

There are hundreds of ways to write custom instructions, but almost all of them fall into four essential categories. When these four are covered, AI becomes dramatically more consistent and accurate.

  1. Define your role and responsibilities clearly

AI must understand you are an Executive Assistant, not a general user. When it understands your role, it automatically produces output that supports EA‑specific tasks: email drafting, meeting preparation, action tracking and briefing summaries.

Example instruction:
“I work as an Executive Assistant supporting senior leaders. I manage complex inboxes, prepare meeting briefs, write communications, summarise information, track actions and organise documents. Tailor your output to these tasks.”

Why it matters:
This stops the AI from giving generic advice and ensures every output aligns with your actual responsibilities.

  1. Explain your executive’s preferences and tone

Executives often have very specific expectations. Some like direct communication, others prefer warmth. Some want bullet points, others want prose. Some prefer very brief summaries, others expect full context.

Instruction examples:

  •  “Use a professional, concise tone when drafting emails for my executive.”
  • “Avoid unnecessary adjectives or overly casual language.”
  • “Always include next steps or actions when summarising meetings or emails.”

Why it matters:
This reduces rewrites and makes drafts feel like they were written by someone who knows the leader well.

  1. Tell the AI how you want information structured

Structure is one of the easiest wins when using custom instructions. Once the AI knows your preferred format, everything becomes faster.

Examples:

  • “Use clear headings and bullet points when presenting information.
  • “When summarising email threads, include three sections: context, decisions, actions.”
  • “For briefing notes, follow this structure: overview, key points, risks, recommended actions.”

Why it matters:
Structure determines clarity. When AI formats information correctly, half your job is already done.

  1. Set boundaries and quality expectations

AI should know what not to do. It should also understand that accuracy and professionalism matter.

Examples:

  • “Do not invent details or assume information.”
  • “Highlight uncertainty instead of guessing.”
  • “Ask for clarification if essential information is missing.”
  • “Keep summaries under 150 words unless otherwise stated.”

Why this matters:
Boundaries prevent AI errors and protect your professional reputation.

A real EA scenario: with vs without custom instructions

Imagine you support a finance director who is preparing for a stakeholder meeting. They forward a long email thread with the instruction: “I need a summary before my 4pm call.”

Without custom instructions

You prompt AI: “Summarise this thread.”

The output:

  • Too long
  • Missing key decisions
  • Doesn’t match your executive’s tone
  • Includes irrelevant details
  • Does not clearly identify risks or next steps

You now rewrite the summary manually.

With strong custom instructions

You prompt AI with the same instruction: “Summarise this thread.”

The output:

  • A crisp overview
  • Key decisions extracted
  • Risks clearly highlighted
  • Action list separated and prioritised
  • Tone aligned to the finance director’s expectations

At most, you tweak a sentence or two.

Time saved: 20 minutes.

Stress saved: immeasurable.

This is what custom instructions unlock: reliability.

How custom instructions transform EA productivity

Once your AI system understands your preferences, every task becomes faster:

  • Email responses begin as strong first drafts.
  • Meeting notes come formatted your way every time.
  • Summaries are concise without further editing.
  • Briefing packs take minutes instead of hours.
  • Reports appear in your executive’s preferred tone.
  • Information retrieval becomes accurate and instant.

Most importantly, you stop repeating yourself.

AI remembers what you expect.

Practical steps to create your own EA custom instruction set

Here are five steps you can apply immediately:

1.Write a paragraph describing your role as an EA

The clearer you describe your responsibilities, the better the AI supports them.

2. Describe your executive’s communication style

Include tone, formality level, preferred formats, and typical audiences.

3. Define your structural rules

Headings, bullet points, short paragraphs, action lists, etc.

4. Specify what you do not want

Guessing, unnecessary details, informal language, invented facts.

5. Test and refine weekly

Custom instructions get better with small adjustments over time.

These small steps can transform how you use AI every day.

Conclusion

AI becomes a valuable professional partner only when it is given the clarity and expectations an experienced EA naturally provides. Custom instructions act as a permanent foundation that helps AI work the way you work. For Executive Assistants who want faster clarity, fewer rewrites and stronger support for their leaders, mastering custom instructions is one of the highest value skills available today.

If you want guided, practical training on using AI, prompting and Microsoft 365 tools at a professional EA standard, you can learn these skills in depth through:

Today’s PA Academy
and
Microsoft Copilot Masterclass

Latest News

Trust and risk: safe Copilot use with sensitive information, compliance, and auditability
15-01-26

Trust and risk: safe Copilot use with sensitive information, compliance, and auditability

Copilot changes how work gets done in Microsoft 365. It makes searching, summarising, drafting, and reusing information dramatically faster. That is the opportunity. The risk is just as clear: if your organisation handles sensitive information, Copilot will accelerate both good work and bad habits. Not because Copilot bypasses security, but...
read more
Copilot readiness checklist: data, permissions, governance, and what to sort before rollout
10-01-26

Copilot readiness checklist: data, permissions, governance, and what to sort before rollout

Most Copilot projects do not fail because the AI is weak. They fail because the organisation is not ready for what Copilot reveals. Copilot does not create new access to information, but it makes existing access dramatically easier to use. That is the shift leaders often underestimate. If people already...
read more
Why Copilot outputs go wrong: the 7 most common failure patterns and how to fix them fast
05-01-26

Why Copilot outputs go wrong: the 7 most common failure patterns and how to fix them fast

Copilot is capable. But it is not a mind-reader. When people say Copilot is not that good, what they often mean is this: the output did not match the real need. Usually, that is not a model problem. It is a workflow problem. Copilot outputs go wrong in predictable ways....
read more
How to enable JavaScript in your browser