Skip to content
All resources
Play For COO, Delivery lead

Your Team Needs a Tutor That Watches the Screen With Them

5 min read Published 31 Jan 2025 Video companion

Start here

Why most software training quietly fails

Most software training fails for the same reason. Someone watches a tutorial, then opens the actual tool, and nothing on screen looks quite like the video. The version is different. The menu is in a different place. The workflow they need does not match the example. So they close the tab, go back to doing things manually, and the training budget is quietly written off.

There is a better way, and it costs nothing to try. Google AI Studio’s real-time screen-sharing mode acts as a live tutor: it watches your screen, identifies the software you are using, and guides you through it step by step. It can troubleshoot errors, explain interface elements, and walk a complete beginner through a workflow it has never seen before.

The real problem

Generic tools give generic results

The villain here is not AI itself. It is the vendor noise around it — the promise that any tool, pointed at any workflow, produces useful output with minimal effort. Most firms have found that is not true. ChatGPT gives generic answers. Copilot summarises the wrong documents. Demos look polished; real briefs do not.

Training is the same. Generic tutorial videos are made for generic users. Your team works with specific software, in a specific way, on specific client workflows. A tutor that can see what they are actually looking at is a fundamentally different proposition.

You are not pasting screenshots or describing problems in text. You are having a conversation with something that has eyes on your actual workspace.

Open the mode

What Stream Real Time actually does

Google AI Studio is free — all you need is a Google account. The interface has several modes: prompt creation, starter app templates, and model fine-tuning. The one that changes how your team learns is Stream Real Time.

When you open it, you share your screen with the AI. It watches what you are doing, in context, and responds to questions about what it can see. In practice your team can ask:

  • “What are the main elements of this interface?” And get each one named and explained in place.
  • “Walk me through this task as a beginner.” And get grounded, step-by-step guidance.
  • “I have an error here — what is causing it?” And get a fix based on what is on screen right now, not a generic description of the software.

Run it

Three tools, one session

Here is what a single session looks like across different software packages.

  • OBS Studio (screen recording). The AI identified the main interface elements — scenes, sources, the audio mixer. Asked for three beginner tips for recording a YouTube video, it returned focused guidance: set up your scene sources correctly, configure your microphone before recording, and run a short test before committing to a full take. A grounded starting point that saves a beginner an hour of trial and error.
  • DaVinci Resolve (video editing). The AI recognised the software without being told. When the cursor pointed at the workspace tabs along the bottom — Media, Cut, Edit, Fusion, Colour, Fairlight, Deliver — it named each one and explained where it sits in a production workflow, then walked through adding the first piece of media to the Media Pool. Those tabs mean nothing to a new user without context; a tutor that can see which one you are hovering over guides you through the exact action.
  • n8n (workflow automation). The most relevant example for teams working toward the Agents layer of the Knowledge–Workflow–Agents (KWA) framework. n8n is an automation platform for building AI-powered workflows without writing code. The session started with a community template for generating Instagram content from trending data — the AI described the workflow diagram, explained what each node does, and outlined the setup steps.

Build it live

A working agent, from nothing

Then a fresh workflow was opened and the AI guided the build of a simple chatbot powered by Google Gemini, from scratch:

  1. Add a “When Chat Message Received” trigger node.
  2. Add an AI Agent node connected to the trigger.
  3. Configure the agent’s prompt and output settings.
  4. Connect the agent to the Chat output.
  5. Link the Google Gemini chat model as the AI backend.

When a configuration error appeared — a missing API key — the AI identified the problem, explained what was needed, and walked through generating a new key from the Google Cloud console. Once the key was added and the workflow tested, the chatbot responded correctly through the built-in chat interface. For a team member who has never opened n8n before, that changes what is possible in a single afternoon.

On privacy

What the tutor can and cannot see

One question worth addressing directly: does sharing your screen with an AI raise data concerns? In this context, the AI is observing the interface — the controls, menus, and workflow structure. It cannot modify anything on screen, access your files, or retain information between sessions. For teams working in sandboxed environments or on non-sensitive workflows, the privacy profile is straightforward.

That said, your firm should have a clear AI usage policy before rolling out any tool. If you do not have one, that is the first step in the Govern stage of the 5 Steps framework.

Why it matters

How wins actually spread

The 5 Steps for AI Leadership framework is sequential: Align, Activate, Amplify, Accelerate, Govern. Most firms are stuck between Activate and Amplify. Individual people have experimented and some have found genuine uses — but the wins are not spreading. Adoption is uneven because learning is uneven. The person who figures out a new tool is usually the one who was already comfortable experimenting. Everyone else waits for a course that may never arrive.

A real-time AI tutor removes that barrier. It meets each person where they are, in the tool they are actually using, on the task they are actually trying to complete. That is how wins spread — not through firm-wide training programmes, but through each person gaining enough confidence to take the next step. A tutor that can see what your team is working on, and guide them through it in real time, is one of the most direct ways to close the gap between “we bought the tool” and “our team actually uses it.”