AI literacy for women in tech: 7 skills to practise this week

You do not need to build an AI model to ask useful questions about one. If an AI tool is helping your team write, analyse, code or make recommendations, you need enough understanding to assess its contribution.

AI literacy means understanding what a tool can do, where its limits matter and how to use it thoughtfully. It includes technical awareness, but also clear communication, evidence checking and knowing who makes the final decision.

For women entering tech or moving into product and delivery, these are useful skills to practise alongside your existing strengths. You can begin with a small task and a few deliberate checks.

Seven skills you can practise

1. Describe the task clearly

A prompt is the instruction you give an AI tool. Start with the work you need to do: the intended reader, the source material, the output and the boundaries.

Practise with a fictional meeting note. Ask for three decisions and their owners, using only information in the note. Tell the tool to mark missing owners as unknown. Check whether it followed the instruction before changing the prompt.

2. Separate fluent writing from evidence

An AI hallucination is an output that presents unsupported or incorrect information as if it were reliable. A polished explanation is not proof that its claims are true.

Take a short answer and underline its factual claims. Check each against the original material. If it cites a source, open that source and confirm that it supports the specific claim. An authentic link can still be irrelevant.

3. Recognise what information should stay private

Before entering work information, check your organisation’s approved tools and data rules. Replacing a person’s name does not necessarily remove identifying details from a document.

Practise by reviewing a fictional customer complaint. Highlight the details the task genuinely needs and the details it can do without. Do not upload real customer records to test the exercise.

4. Ask who might be poorly served

Bias can appear when a system’s outputs systematically favour or disadvantage particular people or situations. It can reflect data, design decisions or how a tool is used.

Imagine an assistant giving career advice. Test fictional profiles with different career breaks, working patterns and routes into tech. Look for unsupported assumptions. A small exercise can reveal a concern, but it cannot demonstrate that a system is fair overall.

5. Evaluate the output against a purpose

An evaluation is a structured check of whether a system performs the task you need. Define what a good result looks like before judging the answer.

For a meeting summary, check factual accuracy, coverage of decisions and handling of unknowns. Use the same checks on several examples. Include an awkward example, such as a discussion where no decision was reached, rather than testing only tidy notes.

6. Keep human accountability visible

Someone should know who reviews the output, who can approve the next action and what happens when the tool is wrong. The answer varies with the task’s consequences.

Practise by drawing a short workflow: source note, draft summary, human review, approved update. Name the person responsible at each stage. If the workflow includes sending messages or changing records, identify the permission required before that action happens.

7. Check whether AI actually helps

A fast first draft can create a slow correction process. Compare the whole task, including preparation, review and rework, with your usual approach.

Try a small, repeatable task using fictional material. Record the time spent, corrections needed and quality of the final result. Treat one attempt as an observation, not evidence that the tool will save the same time everywhere.

Put the skills together

Here is an illustrative team exercise: an AI tool turns workshop notes into proposed actions. Before using the result, the reviewer checks that every action has a source, tentative suggestions remain tentative and nobody has been assigned work without agreement.

If the tool invents a deadline, correct the draft and record the failure. If it leaves a deadline unknown, decide whether that is the right result. Completeness is useful only when the information is supported.

The NIST AI Risk Management Framework provides a voluntary reference for managing AI risks. Use it as a starting point for learning the questions organisations ask, rather than treating one completed exercise as a governance qualification.

Choose one skill for this week

Pick a task you already understand and practise evaluating its output. Keep a short note of the source, your checks, the errors and the human decision.

That note gives you something concrete to discuss with a colleague or mentor: how you assessed an AI-assisted task, what you found and what you would change next.

Further reading: NIST: AI Risk Management Framework.