AI Starter Guide
10 concrete ways to use AI productively starting tomorrow
Specifically for small and medium-sized businesses
Jörg Amelunxen · Software Architect & AI Coach for SMBs
Why this guide?
You know AI is important. But there is a gap between the hype on LinkedIn and your day-to-day reality. On one side, spectacular demos from corporations with million-dollar budgets. On the other, your reality: a small team, a full order book, no time for experiments.
73%
of SMBs in Germany still don't have an AI strategy.
Source: Bitkom, 2024
The good news: you don't need an IT department or programming skills to use AI productively. The tools are here, they are affordable, and many of them you can try out today.
In this guide I will show you 10 concrete ways to use AI in your everyday work. No theory, no buzzwords. Just step-by-step instructions that work tomorrow.
For each way you get: the problem it solves. The solution in one sentence. Concrete steps to follow. A tool recommendation. And a realistic estimate of how much time you will save.
By the way: AI is not a replacement for your expertise. It is a tool that takes over the routine work so you can focus on what really matters.
Before you start: Protect your data
One golden rule applies to all 10 ways in this guide: Think before you enter data into an AI tool. AI providers differ significantly in what happens with your inputs. Some use them for training, some don't. Some store them, some delete them immediately.
The simplest safeguard: Replace names and sensitive details before you paste text into the AI. "Acme Corp" becomes "Company A", "Jane Smith" becomes "Person 1". It takes 30 seconds and reduces the risk enormously.
At the end of this guide you will find a complete GDPR checklist with the key rules for using AI in your business.
Summarize and prioritize emails
The Problem
More than two hours per day in your inbox. You scroll, sort, and hunt for the one important message buried between newsletters and CC chains.
The Solution
AI-powered email triage: the AI reads your emails, summarizes each in one sentence, and flags what truly needs your attention.
How it works
- Gmail: Gemini is already built in and can summarize emails directly and suggest replies
- Outlook: Microsoft Copilot natively summarizes email threads and drafts replies
- Alternatively: open Claude, ChatGPT, or Gemini in a browser and paste an email thread
- Prompt: "Summarize this email conversation in 3 bullet points. What is the core question? What is expected of me?"
- Use the result as a basis for your reply
Warning: Data Sensitivity
Company emails almost always contain sensitive or personal data. Before you process emails in AI tools, check which plan you are on and what the provider does with your data. Free tiers and business tiers have very different terms. When in doubt, replace names and company names beforehand (see "Before you start") or use self-hosted models.
Insider Tip
Copying emails out of your inbox is often harder than expected. Many email clients block selecting multiple messages. The trick: select the emails you want and click "Forward". Most clients will create a new message containing the full content of all selected emails, which you can then copy easily.
Tool Tip
Gmail + Gemini, Outlook + Copilot, Claude
Gmail and Outlook already have AI built in. For everything else: copy and paste into the AI tool of your choice.
Time Saved
30 to 60 minutes per day
Proposals and copy in minutes instead of hours
The Problem
You write every proposal from scratch. Every product description, every client presentation costs you half a day.
The Solution
AI as a writing assistant: you provide bullet points and context, the AI drafts a professional version that you only need to refine.
How it works
- Collect 2 to 3 of your best previous proposals as references
- Create a prompt with context: client, project, scope, budget range
- Prompt: "Draft a proposal in the style of the reference. Client: [name], Project: [description], Scope: [details]"
- Refine the draft with your domain expertise
- Save successful prompts as templates
Warning: Data Sensitivity
Proposals contain client names, project details, prices, and often confidential information. These are sensitive business data. Before using AI, check what agreement you have with your AI provider and whether a Data Processing Agreement (DPA) is in place. When in doubt, replace client names with placeholders before pasting text into the AI.
Insider Tip
Mistral LeChat (lechat.mistral.ai) is a European alternative hosted on EU servers and subject to European data protection law. For proposals, contracts, and other business documents, that can be a decisive advantage.
Tool Tip
Claude, ChatGPT, Mistral LeChat
Claude is especially well suited for longer, structured text. Use the Projects feature to store your templates permanently.
Time Saved
2 to 4 hours per proposal
Meeting notes that write themselves
The Problem
Nobody wants to take minutes. Details get lost, action items are forgotten, and three weeks later no one remembers the agreement.
The Solution
Automatic transcription and summarization: the AI listens, takes notes, and delivers a structured protocol with action items after the meeting.
How it works
- Choose a transcription tool (see tip below)
- Inform all participants that the meeting will be recorded (GDPR!)
- Start the recording at the beginning of the meeting
- After the meeting: let the AI extract a summary and to-dos
- Share the notes directly in your team chat
Warning: Data Sensitivity
Meeting recordings are personal data. You need the explicit consent of all participants before you record. Clarify in advance where the recordings are stored and who has access. Many transcription tools store data on US servers.
Insider Tip
You don't have to integrate a tool into your meeting right away. Start simple: record the meeting with your phone's voice memo app and upload the file to an AI tool for transcription afterwards. That way you keep full control over the recording.
Tool Tip
Otter.ai, Fireflies.ai, tl;dv
tl;dv offers a free tier and integrates directly with Google Meet and Zoom. For Teams users: Microsoft Copilot can do this natively.
Time Saved
1 to 2 hours per week
Market research at the push of a button
The Problem
Hours of Google research for competitors, trends, and pricing. At the end you have 30 tabs open and no clear overview.
The Solution
Structured AI research: you ask a specific question, and the AI delivers a structured analysis with sources.
How it works
- Formulate your research question as specifically as possible
- Prompt: "Analyze the top 5 competitors in the [industry] space in [region]. Compare: pricing model, target audience, strengths/weaknesses. Cite sources."
- Have the AI create a comparison table
- Verify the key facts manually (AI can hallucinate!)
- Use the result as a foundation for strategic decisions
Insider Tip
Perplexity offers a free tier that is more than sufficient for most research tasks. The advantage over ChatGPT: every statement is automatically backed by sources you can check directly. This significantly reduces the risk of hallucinations.
Tool Tip
Perplexity, ChatGPT with web search, Gemini
Perplexity is built specifically for research and automatically provides source citations. Ideal for market analyses.
Time Saved
3 to 5 hours per research task
Social media content without an agency
The Problem
No budget for an agency, no talent for copywriting. Your last LinkedIn post is three months old.
The Solution
AI-powered content creation: you bring the expertise and the AI turns it into posts that sound professional and match your voice.
How it works
- Define your brand voice: 3 adjectives that describe your tone
- Collect 5 of your own posts or texts as a style reference
- Prompt: "Write a LinkedIn post about [topic]. Tone: [your 3 adjectives]. Target audience: [description]. Length: 150 to 200 words."
- Edit the output and add your personal touch
- Create a content plan: 2 posts per week for one month
Insider Tip
With ChatGPT you can create a "Custom GPT", with Claude a "Project". In both cases you store your brand voice, target audience, and sample texts once. After that, a short prompt like "Write a post about [topic]" is enough and the AI hits your tone right away.
Tool Tip
ChatGPT, Claude, Jasper
Create a Custom GPT or Claude Project with your brand voice and sample posts. That way every new post will be more consistent.
Time Saved
2 to 3 hours per week
Answer customer questions automatically
The Problem
The same 20 questions keep coming in. By email, by phone, by chat. Your team spends hours on repetition.
The Solution
AI-based FAQ and reply templates: the AI knows your most common questions and generates context-aware answers that you only need to send.
How it works
- Collect the 20 most frequent customer questions from the last 3 months
- Write an ideal answer for each question
- Upload them as a knowledge base into an AI tool
- Prompt: "Answer the following customer inquiry based on our FAQ: [insert question]"
- Review the answer and adjust the tone before sending
Warning: Data Sensitivity
Customer inquiries often contain names, order numbers, and personal details. For the FAQ knowledge base, use only generic questions and answers, not real customer data. If you paste customer inquiries into the AI, anonymize them first.
Insider Tip
Don't start with a chatbot. First, create an internal tool for your team: a Custom GPT or Claude Project that knows your FAQ and generates draft replies. A human always sends the final answer. That saves time and avoids mistakes.
Tool Tip
ChatGPT Custom GPTs, Claude Projects, Intercom Fin
To start, a Custom GPT with your FAQ is enough. If you have a live chat, you can use Intercom Fin or Tidio AI as a bot.
Time Saved
3 to 5 hours per week
Analyze spreadsheet data intelligently
The Problem
Spreadsheets full of data but no time for analysis. Pivot tables feel like hieroglyphics, and there is no budget for a data analyst.
The Solution
Load data into AI and detect patterns: you upload your spreadsheet, describe what you want to know, and get analysis plus visualizations.
How it works
- Export your data as CSV or XLSX (without sensitive customer data!)
- Upload the file to ChatGPT (Plus) or Claude
- Prompt: "Analyze this sales data. Show: top 3 products, revenue trend over the last 12 months, seasonal patterns. Create a chart."
- Ask follow-up questions: "What stands out?" or "Do you see any anomalies?"
- Export the results for your next team meeting
Warning: Data Sensitivity
Business data such as revenue, margins, and customer statistics are confidential. Before uploading, remove columns with customer names, email addresses, or other personal data. Often, anonymized data (Customer 1, Customer 2) is perfectly sufficient for the analysis.
Insider Tip
ChatGPT (Plus/Team) can run Python code in the background. That means you can not only get analyses but also finished charts and even processed spreadsheet files for download. Just ask: "Create a bar chart and give me the processed data as an Excel download."
Tool Tip
ChatGPT (Plus/Team), Claude
ChatGPT can execute Python code and generate charts directly. Claude can analyze large files and describe patterns in text form.
Time Saved
2 to 4 hours per analysis
Optimize job postings and applications
The Problem
The talent shortage hits SMBs especially hard. Your job postings sound like everyone else's, and the applications pile up unsorted.
The Solution
AI-optimized job ads and structured applicant evaluation: the AI helps you write postings that stand out and evaluate applications systematically.
How it works
- Collect your most successful job postings as a reference
- Prompt: "Write a job posting for [position]. Company: [brief description]. What makes us special: [USPs]. Tone: authentic, not corporate."
- Have the AI adapt the posting for different platforms (LinkedIn, Indeed, your own website)
- For applications: "Compare these 5 applications against the criteria: [must-have requirements]. Create a ranking."
- Use the shortlist as a starting point, not a final decision
Warning: Data Sensitivity
Applicant data is particularly sensitive personal data. Never upload complete applications to public AI tools. If you use AI for pre-screening, anonymize the documents (remove names, photos, addresses) and make sure a human always makes the final decision. This is not only GDPR-relevant but also required by anti-discrimination laws.
Insider Tip
Use AI especially for the job ad itself, not just for screening applications. Have the AI generate different tones (casual vs. professional) and A/B test them on different platforms. An authentic posting that sounds like a real company often delivers more than a higher salary.
Tool Tip
ChatGPT, Claude, Personio (with AI features)
Important: let the AI pre-screen, but always make the final decision yourself. Automated applicant rejections only with human review (in compliance with anti-discrimination laws).
Time Saved
3 to 6 hours per job posting
Translations for international clients
The Problem
DeepL is often not enough. Technical terms get mistranslated, context is lost, and the translation sounds wooden.
The Solution
Context-aware AI translation: you provide not just the text but also the context, the target audience, and the terminology. The AI delivers translations that read naturally.
How it works
- Create a glossary of your most important technical terms with the desired translation
- Prompt: "Translate the following text into [language]. Context: [industry/topic]. Target audience: [description]. Use these technical terms: [glossary]. Tone: professional, not stiff."
- Have the AI suggest alternative phrasings for critical passages
- For recurring texts: save the prompt as a template
- For important documents: have a second AI review the translation
Warning: Data Sensitivity
For contracts, terms of service, or legally binding documents: AI translations are a good starting point but do not replace a professional specialist translation. When legal validity is required, have a native speaker or specialist translator finalize the AI translation.
Insider Tip
Use two AIs as a cross-check: have one tool create the translation and another review it. For example: DeepL translates, Claude reviews and refines. For contracts or legally relevant texts, this double check is invaluable.
Tool Tip
Claude, ChatGPT, DeepL Write
Claude is strong at context-aware translations. For quick standard translations, DeepL remains excellent. The best combination: DeepL for the raw text, Claude for the fine-tuning.
Time Saved
1 to 3 hours per translation
Document processes before the knowledge walks out the door
The Problem
Everything is stored in the heads of a few employees. When someone gets sick or leaves, critical knowledge is lost.
The Solution
AI-powered process documentation: you describe or narrate a process, and the AI structures it into clear, reusable documentation.
How it works
- Identify the 5 most critical processes that exist only in people's heads
- Record a conversation with the knowledge holder (with their consent)
- Transcribe the conversation (see Way 3)
- Prompt: "Create process documentation from this transcript. Structure: objective, prerequisites, steps (numbered), common mistakes, contact person. Target audience: new employees."
- Have the knowledge holder review and supplement the documentation
Warning: Data Sensitivity
Internal process documentation can contain trade secrets, proprietary know-how, and confidential procedures. Think carefully about which details you enter into an external AI tool. For especially sensitive processes: describe the process abstractly and add the concrete details (system names, credentials, contact persons) manually afterwards.
Insider Tip
You don't need a formal interview. Just ask the knowledge holder to explain the process out loud while you record a voice memo. People explain things verbally much more completely and naturally than in writing. The AI then turns the spoken knowledge into clean documentation.
Tool Tip
Claude, ChatGPT, Notion AI
Claude can process very long transcripts (up to 200,000 characters). Ideal for detailed process interviews.
Time Saved
4 to 8 hours per process
GDPR Checklist for AI Use
AI tools are powerful, but you bear the responsibility for the data you enter. You should follow these five points from the very start:
No personal data in public AI tools
Customer names, addresses, emails, or contract data do not belong in ChatGPT & Co. Anonymize or pseudonymize before entering data into an AI.
Check the Data Processing Agreement (DPA)
If you use AI tools professionally, you need a DPA with the provider. OpenAI, Anthropic, and Google offer these for their business plans.
Opt out of training data
Many AI providers use your inputs for training. You can disable this in the settings. For sensitive data, this is mandatory.
Create a team policy
Define clearly: which data may go into AI tools? Which tools are permitted? A simple document with 5 rules is enough. For example: "No customer data, no contracts, no internal financial data."
Transparency toward customers
If AI-generated content goes to customers (e.g. automated replies), inform them about it. This builds trust and may be legally required depending on the context.
Your next step
You now have 10 concrete approaches for using AI in your business. Each one can save you hours per week and improve the quality of your work.
But every business is different. Which of these ways will have the biggest impact for you depends on your processes, your team, and your goals.
60-min walkthrough • €99
60 minutes that sort things out
In 60 minutes we will figure out together where AI has the biggest impact for you. No sales pitch, just real sorting work: a concrete analysis of your situation and a plan for the next steps. €99 as a low-friction filter, so the slot actually goes to you.
Jörg Amelunxen
Software Architect & AI Coach for SMBs
I develop AI strategies that deliver results for SMBs from day one. No buzzword consulting, just systems that work tomorrow.