Emails, Proposals, Minutes: AI for Everyday Office Writing in SMBs

Tuesday, 2:30 p.m. The project manager at a small IT services company has just finished a client call. On her notepad: five action items that need to go out as meeting minutes. In the inbox: a complaint that needs a diplomatic reply. On the to-do list: the status update for the client and the scope description for a new proposal. All due by tomorrow.

It is not the big projects that eat your day. It is the many small writing tasks in between. Every email needs to be composed, every set of meeting notes turned into readable text, every proposal properly drafted. Each one takes 20 to 40 minutes. Spread across the day, that adds up to hours stolen from the actual work.

AI Can Handle Everyday Writing. But Only with the Right Briefing.

AI language models have become remarkably good at exactly this type of text: emails, summaries, reports, proposals. The tools are available, the costs manageable, the access straightforward.

Yet I often see disappointment when people try them. The reason is almost always the same: the briefing is missing. If you just tell a language model “Write me an email to the client,” you get generic output that is barely usable as a rough template. The result sounds like AI, not like your company.

The difference between useful and useless output comes down to the context you provide: Who is the recipient? What is the situation? What tone should the text strike? What information must be included? The more precise the briefing, the better the result.

Working with AI Is Not Copy-Paste

One misconception persists: that AI-assisted writing means typing a prompt and using the output as-is. In practice, it works differently, and actually better.

The workflow looks like this: you provide bullet points, context, and the desired tone. The AI delivers a draft. You review, adjust, and approve. This three-step process does not just save time; it often improves quality as well. The AI does not forget any points, keeps the structure consistent, and writes complete sentences where you might cut corners under time pressure.

Tip: speak your notes instead of typing them. Most smartphones and laptops have a built-in dictation function. Just speak your notes into the microphone instead of typing them out. If the speech recognition gets a word wrong here and there, that is not a problem: the AI understands the context and still produces clean output. This way, a quick monologue after a meeting or phone call becomes the basis for a polished text in seconds.

Below, I walk through four everyday situations from SMB office life with concrete before-and-after examples. The prompts are deliberately written so you can adapt and try them right away.

One important point upfront: pseudonymization. Before passing text containing client names, company names, or other personal data to an AI tool, replace them with placeholders. “Customer Müller” becomes “Mr. [CLIENT],” “Müller GmbH” becomes “[COMPANY].” The AI still produces good results, and sensitive data never leaves your organization. In the final text, you simply swap the real names back in. It takes seconds and quickly becomes second nature. I demonstrate this step consistently throughout the examples below.

Example 1: Customer Email for a Complaint

A customer complains that a delivery was incomplete. You know what happened and how you will resolve it. Now you need to turn that into an email that sounds professional, shows understanding, and offers a clear solution.

Before: Your Bullet Points

Customer Müller (fictitious name) only received 3 of 5 packages. Remaining 2 got stuck with the shipping provider. Will be delivered tomorrow. Tracking number available. We cover shipping costs for the redelivery.

What You Send to the AI (Pseudonymized)

Replace the client name with a placeholder and provide the AI with role, context, and tone:

You are a customer service representative at a mid-sized trading company. Write a reply email to Mr. [CLIENT], who complained about an incomplete delivery. Tone: friendly, solution-oriented, not overly apologetic. Facts: [CLIENT] only received 3 of 5 packages. Remaining 2 got stuck with the shipping provider. Will be delivered tomorrow. Tracking number available. We cover shipping costs for the redelivery. Length: short and to the point.

After: AI-Generated Email Draft

Dear Mr. [CLIENT],

Thank you for reaching out. I completely understand that an incomplete delivery is frustrating, and I want to resolve this for you as quickly as possible.

We have looked into the matter: two of your five packages were delayed by the shipping provider. The cause was a sorting error at the logistics center, and the goods themselves are not affected. The redelivery has already been initiated and will arrive tomorrow. Your tracking number for the reshipment is: [TRACKING-NO].

We will of course cover the shipping costs for the redelivery. Should anything still be missing after tomorrow’s delivery, please contact me directly so I can take care of it personally.

Thank you again for your patience. We are working to ensure this does not happen again in the future.

Best regards

Final step: Replace [CLIENT] with “Müller” and [TRACKING-NO] with the real number. Then check: are the facts correct? Does the tone match your company? The AI delivers the draft in seconds. With pseudonymization and fine-tuning, the whole process takes two minutes instead of twenty.

Example 2: Scope Description in a Proposal

You need to write a proposal and know exactly what needs to be done. But turning internal notes into text that the client understands while still looking professional takes time.

Before: Your Internal Notes

Website relaunch for Meier & Sons Ltd. (fictitious name). Migrate from WordPress to Next.js. Responsive design, SEO optimization, contact form with database integration. Content migration from existing blog (approx. 50 articles). Hosting on managed server. Timeline approx. 8 weeks. Post-launch support 3 months.

Prompt Approach (Pseudonymized)

Define the audience, technical depth, and scope. The company name is replaced with [COMPANY]: “Write a scope of services for a client proposal to [COMPANY]. Audience: managing director of a mid-sized company with no technical background. Use technical terms only when necessary and explain them briefly. Length: approx. 150 words. Structure: short introduction, then list of services, then note on support. Facts: [paste your pseudonymized notes here].”

After: Scope Description

As part of the website relaunch, we will migrate your existing web presence to a modern, high-performance platform. The new site will display optimally on all devices, is optimized for search engines, and offers your visitors a contemporary user experience.

Our services at a glance:

• Technical migration from WordPress to Next.js (a modern web framework for fast, secure websites)
• Responsive web design for desktop, tablet, and smartphone, ensuring your website works flawlessly on every device
• Search engine optimization (SEO) for the entire website, including site structure, loading times, and meta information
• Contact form with database integration for direct customer inquiries
• Migration of all existing blog articles (approx. 50 posts), including adaptation to the new design
• Hosting on a managed server with regular security updates and daily backups

Implementation will take approximately eight weeks. During the project, we will coordinate with you through weekly brief updates so you always know the current status and can provide feedback. After go-live, we will support you for three months with questions, adjustments, and onboarding to the new platform.

The difference: Bullet points become text that the client immediately understands. The AI handles the writing, you retain content control. In my article “How a 10-Person Company Saves 15 Hours per Week with AI”, I describe the full proposal workflow with an internal knowledge base in more detail.

A Quick Note on Data Privacy

Proposals often contain sensitive business data: pricing, terms, client names. The pseudonymization we use throughout these examples is the simplest protection: replace personal data with placeholders before it goes to the AI. For particularly sensitive content (e.g., pricing structures or cost calculations), it is also worth looking into local models running on your own infrastructure. I cover this topic in detail in the article on AI time savings, including practical tips on data processing agreements and self-hosting.

Example 3: From Meeting to Finished Minutes

Meetings are part of daily life, but writing up the minutes afterwards often feels like a chore. Sorting bullet points, writing them out properly, assigning action items. The good news: there are now several ways to significantly reduce the effort.

Transcription as a Starting Point

Many video conferencing tools like Teams, Zoom, or Google Meet now offer built-in transcription. In addition, specialized meeting assistants (e.g., Otter, Fireflies, tl;dv) can automatically join meetings and create summaries. The trend is moving toward meeting minutes being generated almost as a byproduct.

Data Privacy for Meeting Recordings

For SMBs in particular, one point is non-negotiable: data privacy. Anyone who wants to record or transcribe meetings needs to follow some rules:

• Consent from all participants: Recording and transcription only with explicit consent. Most tools signal the recording transparently (recording badge, chat notification).
• External participants: Inform clients or partners in advance, ideally in writing. Not everyone is comfortable with AI transcription.
• GDPR-compliant providers: Check the data processing agreement, consider the server location. Prefer European providers or those with EU data processing.
• Alternative without audio: If you cannot or do not want to use transcription tools, your own handwritten notes plus AI post-processing will already get you very far. No recording needed.

Before: Your Notes from the Meeting

Participants: Anna (PM), Ben (Development), Carla Meier (Client, fictitious name). Topic: Migration project status. Database migration 80% complete, rest by Friday. Frontend adjustments need another week. Carla wants weekly status update via email. Next meeting: Wednesday 10 a.m. Ben to clarify open API question by tomorrow.

Prompt Approach (Pseudonymized)

Specify the desired structure explicitly. Before sending, replace external names: Carla Meier becomes [CLIENT_REP]. Internal team members (Anna, Ben) can remain if your company policy allows it. “Create structured meeting minutes from the following notes. Structure: 1. Participants and date, 2. Topics discussed with summary, 3. Decisions, 4. Action items with responsible person and deadline, 5. Next meeting. Notes: [paste your pseudonymized bullet points here].”

After: Structured Meeting Minutes

Minutes: Migration Project Status

Participants: Anna (Project Management), Ben (Development), [CLIENT_REP]

Topics Discussed:

• Database migration: 80% complete. The remaining tables concern inventory data and will be migrated by Friday. A full test run is scheduled for Friday afternoon.
• Frontend adjustments: Expected to require one more week. Main effort lies in adapting the responsive views for tablet and smartphone. Design sign-off by [CLIENT_REP] is still pending.
• Client communication: [CLIENT_REP] requests a weekly status update via email, every Monday. Content: progress from the past week, planned tasks, open items.

Decisions:

• Weekly status update to [CLIENT_REP] will be implemented (starting next Monday).
• Design sign-off for responsive views will be scheduled as a meeting next week.

Action Items:

• Ben: Clarify open API question with the interface provider (deadline: tomorrow)
• Ben: Complete database migration and document test run (deadline: Friday)
• Anna: Schedule design sign-off meeting with [CLIENT_REP] (deadline: end of week)
• Anna: Send first weekly status update to [CLIENT_REP] (deadline: Monday)

Open Questions:

• Clarification of the API interface may impact the timeline for the frontend adjustments. Ben will inform the team by tomorrow about the outcome.

Next Meeting: Wednesday, 10:00 a.m.

Final step: Replace [CLIENT_REP] with “Carla Meier” before distributing the minutes. Even rough bullet points are enough for the AI to work with. If you jot down quick notes during the meeting and run them through AI afterwards, you have your minutes in five minutes instead of thirty.

Example 4: Project Status Update for the Client

Clients want to know how their project is progressing. But writing a good status update that shows progress without drowning in technical jargon, and that is honest about delays without causing panic, requires finesse.

Before: Your Internal Knowledge

Project: Website relaunch Müller GmbH (fictitious name). Sprint 3 of 4 completed. Design approved, all pages implemented. Performance optimization in progress. Blocker: Client has not delivered image assets (2 weeks overdue). Without images we cannot finalize the homepage and About Us page. Go-live will likely shift by one week.

Prompt Approach (Pseudonymized)

Focus on client-friendly language. Company and contact names are replaced with placeholders again: “Write a project status update as an email to the contact person [CONTACT] at [COMPANY]. No technical background. Tone: professional, positive, but honest. Avoid jargon. Structure: What was achieved, where we stand, what happens next, what we need from [COMPANY]. Facts: [paste your pseudonymized internal knowledge here].”

After: Status Update

Dear Ms. [CONTACT],

Here is a quick update on your website relaunch.

What we have achieved: The design has been fully approved, and all pages are technically implemented. This means three of four project phases are now complete. We are currently working on loading speed optimization to ensure your new website runs fast and reliably. Early tests are already showing significantly faster load times compared to the current website.

Where we need your help: To finalize the homepage and the About Us page, we still need the final image assets (team photos and header image). As soon as the images reach us, we can complete these two pages. Ideally, we would receive the files by the end of this week so we can stay on schedule.

Timeline: Our go-live date will likely shift by about one week, as the image integration is still pending. Once the material is available, we can confirm the exact date. All other work is progressing on schedule.

Next steps: In the coming week, we will complete the performance optimization and begin final testing across different devices and browsers. In parallel, we are preparing the domain switchover to ensure a smooth go-live.

If you have any questions, please do not hesitate to reach out.

Best regards

Final step: Replace [CONTACT] with “Müller” and [COMPANY] with “Müller GmbH.” The AI translates “Blocker: image assets overdue” into a polite request, frames the delay as a fact rather than an accusation, and structures the update so the client immediately understands what is going well and what they need to do.

The Next Step: Building Your Own Templates

The four examples reveal a pattern: the effort is not in the AI itself, but in crafting the right prompt once. Once you have a solid prompt for complaint emails, you can reuse it every time. Once you have defined a meeting minutes template, you just plug in the bullet points.

The logical next step: create fixed templates and system prompts for the three to five most common text types in your business. Invest once, save time permanently.

This is exactly what I help with in my AI coaching: together, we identify the writing tasks that cost you the most time and develop reusable prompts tailored to your company, your tone, and your requirements. No generic tips, but templates you can put to work the very next morning.

All names of individuals and companies used in this article are fictitious. Any resemblance to real persons or businesses is purely coincidental and unintentional. The examples are provided solely for illustrative purposes.