Let's start with the numbers, because they matter.

And the stakes of that gap are stark. The jobs women hold are three times more likely to be automated by AI. Yet women are the ones least likely to be upskilling to stay ahead of it.

This is not a story about women being less capable, less curious or less interested in technology. The women I know are brilliant, adaptable and more than capable of learning anything they put their minds to.

This is a story about access. About who gets invited into the conversation. About who the resources, the spaces and the language of AI were built for.

Why the Gap Exists

There are a few things happening at once that compound the problem.

The resources weren't built for us

Most AI learning content is written by and for software engineers, data scientists and tech founders. The examples use coding metaphors. The vocabulary assumes a technical background. Even the communities where AI is discussed are dominated by the same voices.

Women who are curious about AI — but don't identify as technical — hit this wall early and often conclude that AI is "not for them." It is. The resources just weren't designed with them in mind.

The encouragement gap is real

The jobs women hold are three times more likely to be automated by AI — yet women are using AI at a 25% lower rate than men. Fortune magazine called this "a paradox that could create a two-tiered AI economy." When women aren't encouraged to experiment with AI tools at work, and aren't represented in the rooms where AI strategy is being formed, the gap compounds.

The roles most at risk are disproportionately held by women

Research consistently shows that roles involving repetitive tasks, data entry, coordination and administration are most likely to be disrupted by AI automation. These roles are also disproportionately held by women — particularly women in lower-income brackets and women who have taken career breaks for caregiving.

The women who most need to understand AI are often the least likely to have been given the tools or encouragement to do so.

What Closing the Gap Actually Looks Like

This is not a problem that will solve itself. But it's also not as complicated to address as it might seem.

Start where you are

You do not need to understand how a large language model works to use one effectively. You do not need a computer science degree to build an AI workflow that saves you hours every week. The barrier to entry is lower than the AI industry would have you believe.

The most important thing you can do is start. Pick one task you do repeatedly and try using ChatGPT or another AI tool for it this week. That's it. One task.

Find your community

Learning is faster and more sustainable when you do it alongside people navigating the same questions. Find spaces where women are sharing what is actually working in their real jobs — not abstract AI theory, but practical applications in marketing, HR, finance, ops and beyond.

Claim your seat at the table

AI literacy is becoming a professional essential at the same speed that digital literacy did twenty years ago. The women who develop it now will have a genuine advantage in how their careers develop over the next decade.

This is not about becoming a tech expert. It is about being the person in the room who understands what AI can and cannot do — and who knows how to use it to do better work.

When most people think about "using AI at work," they picture opening ChatGPT and typing something in. But there's a good chance you've been using AI tools every single day without realising it.

Here's what's quietly running AI in the background of the tools you're probably already using.

The AI You're Already Using

Gmail and Outlook

Every time Gmail suggests a reply with Smart Reply, categorises your emails into Primary, Social and Promotions, or moves a suspicious email to Spam — that's AI. Outlook's Focused Inbox, predictive text and meeting scheduling suggestions all use machine learning to get better at predicting what you want.

LinkedIn

Your LinkedIn feed is entirely curated by AI. The job recommendations you see, the "People You May Know" suggestions, the posts that appear at the top of your feed — all determined by algorithms that learn from your behaviour over time. When LinkedIn shows you a job and says "you're a strong match," that's an AI making an assessment.

Salesforce and HubSpot

If you use a CRM, you're likely already working with AI features. Salesforce Einstein scores your leads, forecasts deal probability and surfaces which opportunities are most at risk. HubSpot's AI features draft emails, score contacts and predict which leads are most likely to convert.

Zoom and Microsoft Teams

The meeting transcription that appears after your Zoom call is AI. The background blur that makes your home office look presentable is AI. Teams' noise cancellation — which filters out the sound of your dog or your neighbour's lawnmower — is AI. Zoom's meeting summaries and action item extraction are AI.

Canva

Canva's Magic Write generates copy. Background Remover uses computer vision to isolate subjects in images. The new AI image generation features create visuals from text prompts. If you've used any of these, you've used AI.

Grammarly

Grammarly doesn't just check spelling. Its tone detection, rewriting suggestions and plagiarism checking are all powered by large language models — the same technology behind ChatGPT, just applied specifically to writing assistance.

Excel and Google Sheets

Excel's Ideas feature uses AI to analyse your data and surface insights. Google Sheets can now accept natural language queries — you type "show me total sales by region last quarter" and it builds the formula for you. Both tools are quietly becoming much more powerful through AI integration.

So What Does This Mean for You?

The question isn't whether AI will affect your work. It already is.

The more useful question is: are you the one directing it, or just the one it's happening to?

Every tool listed above has settings, features and capabilities that most users have never explored. The professionals who take the time to understand what AI can actually do — in the tools they're already paying for — will get significantly more value from the same software.

Start this week: pick one tool from the list above and spend 20 minutes exploring its AI features. You'll be surprised what's already there.

When most people hear "AI agent" they picture something from a sci-fi film — autonomous robots making decisions with no human oversight. The reality is far more practical and far more useful.

An AI agent is simply a system that works for you automatically. It's triggered by an event, completes a series of steps, and delivers an output — without you having to initiate each step manually. Think of it as the difference between asking a colleague to do one task and hiring someone who just handles a recurring process for you without being asked every time.

The best part: you do not need to write a single line of code to build one.

First: What's the Difference Between an AI Tool and an AI Agent?

When you open ChatGPT and type a question, you are using AI as a tool. You provide input, it produces output, you review it, the loop ends.

An AI agent is different. It:

• Receives a goal, not just a task
• Takes actions autonomously across multiple steps
• Can call other tools — sending emails, updating spreadsheets, posting to Slack
• Runs automatically when triggered, not just when you ask it to

The practical difference: asking ChatGPT to summarise a report is using a tool. Building a system that monitors your inbox, identifies reports, summarises them and sends you a daily digest without you touching it — that's an agent.

Step 1: Pick the Right Problem

The best first agent solves a problem that is:

• Repetitive — you do it more than once a week
• Rule-based — it follows a predictable pattern most of the time
• Measurable — you can tell whether the output is good or not

Good first agent targets include summarising weekly reports and emailing them to your team, turning meeting notes into structured action item lists, drafting responses to common customer enquiries, and screening job applications against a job description.

Not good first agent targets: anything requiring deep creative judgment, anything where a mistake has serious consequences, anything too unpredictable to define clear steps.

Step 2: Map Your Workflow Before You Touch Any Tool

This step takes 10 minutes and saves hours of frustration later. Write down the answers to these four questions:

Write these down before opening any tool. This clarity is what separates a workflow that works from one that doesn't.

Step 3: Use Zapier to Build It

For your first agent, start with Zapier. It connects over 6,000 apps — Gmail, Slack, HubSpot, Notion, Google Sheets and more — and lets you build automated workflows visually. There is a free tier and you can get your first automation running in under 20 minutes.

Here is a real example: a workflow that takes your meeting notes from Gmail and turns them into a structured action item list in Notion.

Step-by-step in Zapier:

1. Go to zapier.com and click Create Zap
2. Set your trigger: App = Gmail, Event = New Email Matching Search, Filter = subject contains "Meeting Notes"
3. Add an AI action: App = ChatGPT (or Zapier's built-in AI), paste your system prompt
4. Add your output action: App = Notion, Action = Create Database Item
5. Test with a real email, review the output, then switch it on

Step 4: Always Add a Human Review Gate

Before you trust any agent to run fully automatically, add one step: route the output to yourself for review first. In Zapier, this means adding an email step that sends you the draft output before it goes anywhere final.

Review every output for the first two weeks. Once you trust it, you can remove the gate. AI agents occasionally produce confidently wrong output — a review gate is how you catch it before it causes a problem.

This is not optional. It is how professional AI workflows are built.

Real-World Examples by Role

• Marketing: Agent monitors competitor websites weekly, AI summarises changes, sends digest to Slack every Monday
• HR: New application arrives, AI screens CV against job description, sends recruiter a structured summary
• Operations: New deal submitted in CRM, AI checks against pricing policy, flags exceptions for review
• Finance: New invoice received by email, AI extracts key data, creates entry in tracking spreadsheet

Every one of these can be built in Zapier with no coding experience. The hardest part is usually the first step: deciding which problem to solve.

Here is the most common mistake people make with AI tools: they type a vague question and get a vague answer, conclude that AI isn't that useful, and go back to doing things the old way.

The problem is almost never the AI. It's the brief.

Think about it this way: if you hired a brilliant new team member and on their first day you handed them a Post-it note saying "write me a report," you would not be surprised if what came back wasn't quite right. You'd expect to give a proper brief — the context, the audience, the format, what good looks like.

AI is exactly the same. The quality of what you get out is directly proportional to the quality of what you put in.

The Formula That Changes Everything

There is one framework that will immediately make your AI prompts better. Once you see it, you will use it everywhere.

Let's break down each part.

Role — tell the AI who to be

Starting your prompt with "You are a [role]" immediately shapes the tone, vocabulary and perspective of the response. "You are a senior HR business partner" produces very different output from "You are a first-year psychology student" — even for the same question.

Context — set the scene

What situation is this for? Who is the audience? What does the AI need to know to do a good job? The more relevant context you provide, the more tailored the output will be.

Task — be precise about what you want

Vague tasks produce vague results. "Write something about this" is not a task. "Write a 200-word LinkedIn post that leads with a provocative question, summarises the key insight, and ends with a call to action" is a task.

Format — specify how you want it structured

Do you want bullet points or paragraphs? A table or a list? Under 150 words or a full page? Specifying format saves you from having to edit the output into the right shape after the fact.

Weak vs Strong: See the Difference

Real Prompts by Industry — Copy and Adapt These

Marketing & Communications

HR & People

Finance & Administration

Operations & RevOps

Three More Tips That Make a Real Difference

Tell it what you don't want

Negative instructions are just as powerful as positive ones. "Don't use corporate jargon," "don't be apologetic," "don't use bullet points" — these shape the output as much as telling the AI what to do.

Feed it examples of your own writing

If you want AI to match your voice, give it examples. Paste in two or three pieces of your own writing and say "match this tone." It will mirror your style far more accurately than if you just describe it.

Treat it as a conversation, not a one-shot request

If the first output isn't quite right, don't start again. Tell it what to change: "make it shorter," "make the tone warmer," "lead with the second paragraph instead." Iteration is faster than starting from scratch every time.

If you have used Claude or ChatGPT and thought "this is useful, but it would be so much better if it could just see my actual files and emails" — you are about to get very excited.

MCP — Model Context Protocol — is the technology that makes that possible. And while the name sounds technical, what it actually does is surprisingly straightforward once you understand the idea.

What is MCP in Plain English?

Until recently, every AI tool was isolated. You could ask Claude a question, but it could only answer based on what you typed in. If you wanted it to help with a document, you had to copy and paste the whole thing. If you wanted it to check your calendar, you had to type out your schedule. Every interaction started from scratch.

MCP changes that. It is a universal standard — introduced by Anthropic in November 2024 and now adopted by OpenAI, Google and thousands of developers — that lets AI tools connect directly to the apps and data sources you already use.

Instead of copy-pasting your Google Doc into Claude, Claude can read it directly. Instead of describing what is in your Slack channel, Claude can search it. Instead of typing out your calendar, Claude can check it. MCP is what makes AI feel genuinely connected to your actual work, rather than isolated in a chat window.

Which Tools Can Claude Connect to via MCP?

The list is growing rapidly, but right now Claude can connect to the tools most professionals use every day. Here are the most useful ones for non-technical professionals:

Google Drive and Google Docs

Claude can read, search and summarise documents in your Google Drive without you copying anything. Ask it to summarise last quarter's strategy doc, find all files related to a specific client, or compare two versions of a document — and it does it directly from your Drive.

Gmail

Claude can search your inbox, read email threads and draft replies based on the actual content of your emails. Instead of describing a conversation to Claude and asking for help responding, you can ask it to read the thread and draft a reply.

Google Calendar

Claude can check your schedule, identify free time and help you plan your week. Ask "what does my Thursday look like?" or "find me a 30-minute slot this week for a focus block" and it works from your actual calendar.

Slack

Claude can search your Slack channels, read threads and summarise conversations. If you have been away and need to catch up on a channel, or want to find a decision that was made in a thread three months ago, Claude can search it for you.

Notion

Claude can read and write to your Notion pages and databases. Ask it to update a project page, add items to a database or draft content directly into a Notion document.

HubSpot and Salesforce

Claude can access your CRM records, summarise contact histories and help draft follow-up communications based on actual deal data.

How Do You Actually Set It Up?

There are two ways to connect tools to Claude using MCP, depending on your technical comfort level.

Option 1: Claude Connectors (No technical knowledge needed)

The easiest way is through Claude's built-in Connectors — available on the Claude desktop app. Think of it as an app store for MCP connections. You click to connect a service, log in with your account, and Claude is connected. No configuration files, no setup steps.

To access this: download the Claude desktop app, open Settings, and look for the Integrations or Connectors section. Connect the apps you want — Google Drive, Gmail, Calendar, Slack — and you are done.

Option 2: Manual MCP server setup (Requires some technical steps)

For tools not yet available as built-in Connectors, you can connect them manually using community-built MCP servers. This requires editing a configuration file and running a few commands — it is manageable if you are comfortable with basic computer tasks, but it is not quite point-and-click yet.

Why This Matters for Your Work

The practical shift MCP enables is significant. Instead of AI being a separate tool you switch to, it becomes something woven into the tools you already live in.

The professionals getting the most value from AI right now are not the ones who have learned the most prompts. They are the ones who have connected AI to their actual work context — their documents, their emails, their calendars — so that every interaction starts from a place of real understanding rather than a blank page.

MCP is what makes that possible. And the good news is that the barrier to accessing it is lower than you might think. If you have the Claude desktop app, you can connect your first tool today in under five minutes.

One Thing Worth Knowing

When you connect tools to Claude via MCP, Claude can read and in some cases write to those tools. Before connecting anything, it is worth understanding what access you are granting. For most professional use cases this is straightforward — you are connecting your own accounts, not your employer's systems. But if you are using Claude in a workplace context, check your organisation's AI acceptable use policy before connecting company data sources.

The general principle: connect your own accounts freely, and connect employer systems thoughtfully.