---
title: "Learning to Work with AI — Thoughtful Robots"
description: "People need more than a tool. They need to know how to instruct AI, what context it needs, which knowledge it should use, and how to judge what comes back."
source: "https://thoughtfulrobots.ai/articles/learning-to-work-with-ai-revised"
---

# Learning to work *with* AI

People need more than a tool. They need to know how to instruct AI, what context it needs, which knowledge it should use, and how to judge what comes back.

This article takes a closer look at the **Workshops & Training** part of our approach to organisational transformation with AI.

As AI becomes more capable, organisations are exploring more ways to use it across everyday work. But making these capabilities available to people is only part of the change. People also need to learn **how to work effectively with AI as part of their everyday work**.

That means more than learning how to use a particular AI tool. People need to understand how to give AI the right instructions and context, how it can work with organisational knowledge and systems, how to evaluate what it produces, and where human judgement needs to remain.

As AI begins to take different forms within the workplace — from assistants that help with individual tasks to systems that can access knowledge, use tools and take actions — people also need enough understanding of these capabilities to make good decisions about **where and how they should be used**.

Learning to work with AI is therefore about developing both the **practical skills and technical understanding** needed to use it effectively, and eventually help shape how it becomes part of the way work gets done.

## Beyond Learning an AI Tool

Let's return to the B2B software company from the previous article. It is exploring an AI assistant to help its support team find and use product knowledge when responding to customers.

A support executive can quickly learn how to open the assistant and ask it a question. But that alone doesn't mean they can use it effectively. Suppose a customer reports an unusual product issue. What should the executive tell the AI? What information about the customer or product does it need? Which internal knowledge should it use? And when an answer comes back, how does the executive know whether it is good enough to act on?

These are not really questions about **how to operate the tool**. They are questions about **how to work with AI while doing the job**. And answering them requires understanding not only how to interact with AI, but also **what is happening behind that interaction and what the technology can enable**.

## Building the Foundations for Working with AI

Let's start with the interaction itself. When our support executive asks the AI assistant for help, the **prompt** tells the model what they want it to do. But the prompt alone is rarely the whole picture.

The model also needs **context** — information relevant to the task. In our example, that might include the customer's question, the product they are using, previous troubleshooting steps, or other details about the situation.

Then there is **organisational knowledge**. The answer may depend on product documentation, support policies, troubleshooting guides or previous cases. Understanding the difference between what the model already knows, what information is provided as context, and what company knowledge it needs access to is an important part of understanding how these systems work.

And whatever the model produces still needs to be **evaluated**. Is the answer accurate? Did it use the right information? Is anything important missing? Is it appropriate to send to the customer?

So even in this relatively simple interaction, the support executive is beginning to understand several foundations of working with AI: **how to instruct it, what context it needs, what knowledge it should work with, and how to evaluate what comes back**.

These foundations become even more important as AI moves beyond answering a prompt and starts to **connect to systems, use tools and take actions**.

## Understanding the AI Building Blocks

Our support assistant is a useful starting point, but AI can become part of work in many different ways. To make good decisions about how to use it, people need a working understanding of the **building blocks behind these different possibilities**.

It starts with the **LLM, prompts and context**. The model provides the underlying intelligence, while prompts tell it what we want it to do and context gives it the information it needs for the task. When that information lives across company documents and knowledge bases, approaches such as **RAG (Retrieval-Augmented Generation)** can help the model retrieve and work with relevant organisational knowledge.

But useful context does not live only in documents. It may sit inside a CRM, support platform, project management system or another application. **Connectors** can make information from these systems available to AI, while **MCP (Model Context Protocol)** provides a standard way for AI applications to connect to external data, tools and services.

AI can also interact with applications in other ways. **WebMCP**, for example, allows websites to expose structured capabilities that AI agents can use, rather than requiring them to understand a website only through its visual interface. **Plugins and Agent Skills** can package specific tools, instructions and repeatable capabilities that extend what an AI system can do.

These building blocks can then come together in different kinds of AI experiences. A **copilot** may assist someone while they remain closely involved in the work. An **agent** can be given tools and greater responsibility to carry out a sequence of steps towards an outcome. As we move along that spectrum, questions around **permissions, guardrails, evaluation and human oversight** become increasingly important.

For employees, the goal is not necessarily to learn how to build every one of these technologies. It is to develop enough working knowledge to understand **what they do, how they fit together, what they make possible, and what changes when AI is given greater access or autonomy**.

This is an important part of how we approach **Workshops & Training at Thoughtful Robots**. We help teams build a working understanding of these AI building blocks and, wherever possible, explore them through the work they already do. The level of technical depth may differ by role, but the goal is the same: to help people **expand their capabilities and become more confident in understanding, using and shaping how AI fits into their work**.

## Different Roles, Different Ways of Working with AI

Not everyone in an organisation needs the same depth of understanding. What matters is developing the **level of AI knowledge and capability that is useful for the role**.

In our B2B software company, a **support executive** might need to understand how AI uses context and company knowledge, what information it can access, how to evaluate its responses, and when human judgement is required.

A **product manager** may need to go further — understanding what capabilities such as RAG, connectors, MCP and agents make possible, so they can make better decisions about where AI belongs in the product or workflow.

An **engineer** may need much greater technical depth: how these capabilities are implemented, evaluated, secured and integrated with existing systems.

For a **leader**, the emphasis may be different again. They need enough understanding to ask the right questions about where AI can create value, what level of autonomy is appropriate, what risks need to be managed, and what capabilities the organisation needs to build.

The objective is therefore not to give everyone the same AI training. It is to build a **shared foundation across the organisation, with the depth and application shaped around the work different people are responsible for**.

## From Using AI to Shaping How Work Gets Done

As people build their understanding of AI, they can begin to look beyond individual interactions and see **how these capabilities could change the larger workflow**.

Our support team may recognise that AI could retrieve the relevant product knowledge, bring in information from another system, help investigate the issue, prepare a response, or potentially carry out certain actions. They can also identify which parts should remain with people, where approval is required, and what information or safeguards need to be in place.

At that point, people are no longer only learning **how to use AI within an existing way of working**. They are developing the understanding needed to **help decide how the work itself should change when AI becomes part of it**.

That is ultimately the capability we want to build: not simply better AI users, but people who can **thoughtfully shape how AI becomes part of the way work gets done**.

## What Comes Next

Learning to work with AI gives people the capability to help shape how AI should become part of their work. The next step is making that way of working easier to repeat — connecting AI to the right organisational knowledge and systems, putting the appropriate access, controls and evaluation in place, and embedding it into the tools people already use.

Figuring out where AI belongs — in your product or how your organisation works? **Let’s work through it together.**

### Getting ready for AI: *knowing where to begin*

How an organisation understands its workflows, identifies and prioritises AI opportunities, assesses readiness, and decides what to build, buy, wait on, or stop.

### Installing a way of working with AI — *built to adapt*

How organisations choose AI tools and vendors, install governed AI-enabled workflows, and maintain them as capabilities that adapt.