Using AI in Everyday Work: 100 Realistic Scenarios for Professionals Who Work at a Computer

In AI training, generic examples such as "it can summarize your emails" do not readily connect with the work participants do at their own desks. Research also points to this gap. In BCG’s 2025 survey of more than 10,600 employees, only about a third of employees say they have received sufficient training; regular use is significantly higher among employees who have received at least five hours of training and have access to in-person training and coaching. [1] In McKinsey’s 2025 research, organizations reporting high value from AI are 2.8 times more likely than others to say they have fundamentally redesigned their workflows. [2] In an experiment conducted by Dell’Acqua and colleagues with BCG consultants, AI helped participants complete tasks within its capability boundary 25.1% faster, while reducing the likelihood of a correct solution by 19% on a task outside that boundary. [3] The European Commission also explains that, to meet the literacy obligation under the EU AI Act, training should be tailored to the role, risk and context of use, and that simply having people read instructions for use will be insufficient in most cases. [4]

This is why the guide focuses on recurring tasks rather than tools. The first seven sections cover email, meetings, documents and presentations, spreadsheets and data, research, sales and customer service, and HR, finance, legal and operations tasks. The final three sections are devoted to integrations: Google Workspace and Gemini, Claude with MCP connections, and project management with Manus. The appendix summarizes productivity research and risks.

Each scenario follows the same seven fields. Level and tools specifies the required product and licensing requirements, while Who and in what situation describes a fictional person and task that reflect working life in Türkiye. How it is done today shows the existing manual workflow, while How to do it with AI presents the new workflow step by step, using menu names from official help documentation. The expressions inside {curly braces} in the Example prompt field are placeholders for participants to fill in with their own data. The Evidence and real-world example field provides a published organizational case study where one exists; otherwise, it states this explicitly and points to the official product documentation on which the scenario is based. The Points to watch field brings together licensing, language, personal data and verification limitations in one place.

The people and companies are fictional; product capabilities, menu paths, limitations and case study figures are linked to their sources. Menu names follow the English interface shown in official documentation; they may appear differently in the Turkish interface. Features vary by plan, license, administrator setting and region, and the information is based on official pages accessed as of November 2026.

Turkish language support varies from tool to tool, and even between features within the same tool. Turkish is included in Microsoft Copilot’s list of supported languages for work and school accounts [5] and can be selected as a target language for live translated captions in Teams. [6] By contrast, Turkish is not on the official language list for Google Meet’s "Take notes for me" feature [7], presentation generation with Gemini in Google Slides is listed as limited to English [8], Microsoft Copilot video recap supports only English [9], and Turkish does not appear on the list of supported languages for the AI Builder invoice model. [10] This is why it is important for participants to run a small test in Turkish in their own accounts before the training.

When using this guide in training, you can ask participants first to write down three recurring tasks from their own working week, then select the closest scenario from the index and rewrite the "Who and in what situation" field to fit their own work. The example prompt should be tested with fictional or anonymized data rather than actual customer or employee data. KVKK recommends preventing the uncontrolled transfer of personal and sensitive data to publicly available tools when using generative AI tools in the workplace, and having humans review the outputs. [11] NIST also recommends evaluating systems under conditions similar to real-world use and verifying the sources in the output. [12] The verification step in the final field of each scenario should therefore be treated as part of the exercise.