Skip to main content
Business AI
How to Use AI at Work
Twelve practical AI use cases — with exact prompts — that knowledge workers are using right now to reclaim hours every week.
The opportunity
2.5hrs
Average time saved per knowledge worker per day when using AI tools, per McKinsey productivity survey 2024 [McKinsey]
37%
Of employees now use AI for work tasks, up from 4% in 2022 — adoption has accelerated faster than any previous enterprise technology [Gallup]
66%
Of workers say they wouldn't tell their boss they used AI to complete a work task — pointing to both adoption and a confidence gap [Microsoft]
The workers getting the most from AI at work share one trait: they've stopped thinking of AI as a search engine and started treating it as a junior colleague with very specific strengths. Once the basics feel routine, the next step is learning to automate tasks with AI rather than just prompt them one-off. You wouldn't ask a junior colleague to fact-check your legal contracts without supervision — but you'd absolutely ask them to draft a first version of a report or summarise a meeting transcript.
These 12 use cases are categorised by the type of work they affect. Each includes a proven prompt structure you can adapt immediately — and if you want to go deeper, start with our guide to prompt engineering.
Writing and communication
01
Draft emails and responses
Save 45min/day
Paste a received email, describe the context, and ask for a response in your voice. Particularly effective for difficult emails — declining requests, handling complaints, negotiating terms. Edit the draft rather than writing from scratch.
Prompt structure
Here's an email I received: [paste email]. I need to [decline/respond to/ask for X]. My relationship with this person is [context]. Write a response that is [professional/warm/direct] and under 150 words.
ChatGPTClaudeGeminiCopilot in Outlook
02
Summarise long documents
Save 2hrs/week
Paste a report, PDF, or article and ask for a structured summary. Specify the audience and what they care about for much better outputs. Particularly effective for board reports, research papers, lengthy contracts.
Prompt structure
Summarise this document in 5 bullet points for a [C-suite executive / non-technical audience / sales team]. Focus on [key decisions needed / business impact / timeline and milestones]. Highlight any risks or open questions. [paste document]
ClaudeChatGPTNotebookLM
03
Write reports and proposals
Save 3hrs/project
Provide the key points you know, the audience, the purpose, and any data you have. AI generates the first draft — you edit, not write. Particularly effective for status updates, project proposals, business cases.
Prompt structure
Write a [2-page business case / 1-page project update / executive summary] for [audience]. Key points to cover: [list]. Data I have: [list figures]. Tone: [formal/direct]. Format: [sections with headers / bullet points / prose].
ClaudeChatGPTCopilot in Word
Research and analysis
04
Research a topic quickly
Save 1hr/task
Use Perplexity for research tasks that need current, cited information. It retrieves from the live web and shows sources. Better than asking ChatGPT about things that may have changed after its training cutoff.
Prompt structure
Give me an overview of [topic] including: current market size, 3 key trends in 2025–2026, major players, and recent news. Include sources. I need this for [context — competitive analysis / investor pitch / team briefing].
PerplexityChatGPT with searchGemini with search
05
Analyse data in spreadsheets
Save 2hrs/task
Upload a CSV or paste spreadsheet data and ask for analysis in plain English. Identify patterns, outliers, correlations, and summary statistics without needing to know Excel formulas. Claude handles tables particularly well.
Prompt structure
Here's data from [source]: [paste data]. Tell me: 1) The top 3 patterns you see, 2) Any outliers or anomalies, 3) What this suggests about [business question]. Also suggest 2 charts that would illustrate this data effectively.
ClaudeChatGPT with Code InterpreterCopilot in Excel
06
Competitive research
Save 3hrs/month
Build systematic competitor intelligence by asking AI to analyse public information: pricing pages, job postings (reveal strategic priorities), LinkedIn posts from founders, product changelogs.
Prompt structure
I'm researching [competitor name]. Based on their public communications, job postings, and pricing, identify: their apparent strategic priorities, how they position vs [your company], any recent product or pricing changes, and gaps or weaknesses I could use in sales conversations.
PerplexityClaudeChatGPT
Meetings and productivity
07
Meeting preparation
Save 30min/meeting
Paste the meeting agenda, attendee backgrounds (from LinkedIn), and your goals. Get a prep brief with suggested talking points, questions to ask, and potential objections to handle.
Prompt structure
I have a [30-min sales call / executive review / negotiation] with [role/company]. Agenda: [list]. My goals: [list]. Key background on them: [notes]. Give me 5 talking points, 3 questions to ask, and 2 potential objections with responses.
ClaudeChatGPT
08
Transcribe and summarise meetings
Save 45min/meeting
Use a transcription tool (Otter.ai, Fireflies, or Zoom's built-in AI) to get the transcript, then ask an LLM to extract decisions, action items, and key points. Eliminates manual note-taking entirely.
Prompt structure
Here's the transcript of a [meeting type] meeting: [paste transcript]. Extract: 1) All decisions made, 2) Action items with owner names, 3) Open questions that need follow-up, 4) A 3-sentence summary for people who didn't attend.
Otter.aiFirefliesClaudeCopilot in Teams
09
Create templates and processes
Save 2hrs/quarter
Describe a repetitive process you do manually and ask AI to create a template, checklist, or SOP. Once created, you never have to recreate it from scratch. Particularly effective for onboarding docs, project kickoff checklists, review frameworks.
Prompt structure
Create a [template/checklist/SOP] for [process]. It should cover [key steps]. Format it so someone new to the role can follow it without additional guidance. Include [any specific requirements — time estimates, responsibilities, sign-offs].
ClaudeChatGPT
Getting better results
The three things that improve every AI output
1. Specify the audience — "write for a CFO who doesn't have technical context" produces different (better) results than just "write". 2. Specify the format — "use 5 bullet points with 1 sentence each" gives you something you can actually use. 3. Give it examples — "match the tone of this paragraph [example]" is more reliable than adjectives like "professional" or "concise".
The workers getting the most from AI don't have special technical skills — they've built habits around prompt quality. They know to give context, specify format, and edit AI drafts rather than accepting first outputs. And as the research in how much AI actually increases productivity shows, those habits are where the real gains come from. If you use AI at work and aren't happy with results, the prompt is usually the problem before the model is.
For a deeper guide to writing effective prompts, see the Veltrix Prompt Library — 50+ templates tested across business use cases.
FAQ
Is it safe to paste work documents into ChatGPT?
It depends on your company's data policies and what the document contains. OpenAI's Enterprise plan and Claude's Team/Enterprise plans offer data privacy guarantees where your inputs aren't used for training. Check your organisation's AI usage policy before pasting sensitive documents, client data, or proprietary information into consumer AI tools.
Do I need to tell my employer I'm using AI?
Check your company policy first. Many organisations have or are developing AI usage policies. From an ethical standpoint, if your deliverable needs to represent your expertise (a client report, a strategic recommendation), you should be applying sufficient review and editorial judgement to own the output — not just submitting AI-generated text unchanged.
Which AI tool is best for work?
For most knowledge workers: Claude 3.5 Sonnet handles long documents and nuanced writing best. ChatGPT with the GPT-4o model is the most capable general tool. Gemini integrates best with Google Workspace. Microsoft Copilot integrates best with Office 365. Most heavy users have two: one for deep analysis work (Claude), one for quick tasks (ChatGPT or Gemini). See the full AI Tools directory for detailed comparisons.
Sources
[McKinsey] McKinsey Global Institute — "The economic potential of generative AI" (2024)
[Gallup] Gallup — "Americans' Use of AI in the Workplace" (2025)
[Microsoft] Microsoft Work Trend Index — "Will AI Fix Work?" (2024)
04 — Don't watch from the outside
Veltrix Collective
Stay ahead of
the curve
the curve
Weekly briefings on AI tools, adoption trends, and what actually matters for practitioners. No hype. Just signal. Join readers navigating the shift.
Weekly, every Tuesday · No spam · Privacy policy · Unsubscribe anytime
Written by Luke Madden, founder of Veltrix Collective. Data synthesis and analysis by Vel.