AI in Action: How to Apply Artificial Intelligence at Work
- Published
- Updated
Artificial Intelligence can save time, support creation and help organise information. That potential only becomes valuable when a person knows which task to choose, what context to provide and how to validate the result.
This shift from curiosity to application was the purpose of AI in Action: Strategies and Tools for the Digital Future, designed with ISQ Academy. The two online mornings were predominantly practical, with exercises based on real professional situations.
What does applying AI at work mean?
Applying AI is not opening a chatbot and asking it to "do this for me". It is redesigning part of the work to combine:
- the person's knowledge;
- the organisation's data and rules;
- the tool's capability;
- human review and decision-making.
A well-chosen use case has a known input, an expected output and criteria for assessing quality. Preparing an executive summary from an approved report is easier to control than asking AI to decide the company's strategy.
A five-step method
1. Choose a concrete task
Look for work that is repeated, time-consuming and easy to review: preparing drafts, comparing documents, organising notes or transforming information into a different format.
Describe the task with a verb and an object: "summarise this report", "classify these requests" or "prepare the agenda for this meeting".
2. Define the output
Specify the audience, goal and format. A useful output might be a table of decisions and owners, an email under 150 words or three scenarios with advantages, risks and assumptions.
Without an explicit criterion, speed does not mean quality.
3. Provide context and limits
Supply only the necessary information. State which sources may be used, which facts cannot be changed and when AI should admit that it does not have enough information.
Before uploading professional data, check internal rules, account settings and the tool's processing terms.
4. Review critically
Verify names, numbers, dates, quotations and inferences. Ask:
- Does the answer use only the available information?
- Is there an unsupported claim?
- Is the tone appropriate?
- Is an important perspective missing?
- Should someone affected by the decision be involved?
5. Measure and document
Record time before and after, the number of corrections and the situations in which the process failed. Save the approved prompt, a sample output and the review rules.
This documentation makes it possible to repeat what worked and teach others without relying on the memory of the person who started.
Different tools serve different functions
During the course, we explored assistants including ChatGPT, Microsoft Copilot, Notion AI and Perplexity. Selection should not begin with brand popularity, but with the task:
- General-purpose assistant: writing, synthesis and idea exploration;
- Work-integrated assistant: support in documents, email and meetings;
- AI-assisted research: discovery and organisation of sources;
- Specialist tool: performance of a particular function;
- Automation: movement of information and actions between systems.
A company may use several categories, but it needs to reduce duplication and establish which tool is authorised for each type of information.
Example: preparing a weekly meeting
A simple process can receive the week's notes, project list and pending decisions. AI organises:
- subjects requiring a decision;
- blockers;
- owners;
- suggested time per subject;
- unanswered questions.
The owner reviews the agenda, corrects priorities and sends the final version. After the meeting, the same structure can support follow-up without inventing decisions that were not recorded.
From training to capability
A training session opens possibilities. Capability appears when people apply the method, share examples, receive feedback and update their practice as the tools change.
The article AI at work: the problem is application expands on the criteria for selecting and integrating a first use case.
For continued learning, SuperHumano Academy brings together courses, resources and practical experiences throughout the year. Organisations can explore AI training for companies and role-based programmes in Training and Capability Building.
Frequently asked questions
What is the first task I should test with AI?
A frequent, low-risk task that is easy to verify. Summarising non-sensitive information or creating a first draft are good starting points.
How do I know whether AI is really saving time?
Measure the complete process, including preparation, review and correction. Compare quality as well as speed.
Do I need to learn every tool?
No. Understanding instruction, validation, data and safety principles is more durable than mastering dozens of interfaces.
Context note: this article was published to introduce the 3 and 4 June 2025 edition at ISQ Academy. That edition has ended; the text preserves the method behind the course and points to current options.