ChatGPT
Projects can keep chats, files and project instructions together. Memory, project sharing, tool access and workspace controls vary by account type and configuration.

The biggest improvement often does not come from switching AI model. It comes from learning how to work with the AI assistant you already have.
This is a short, practical learning course for ordinary business users. Pick your current level, try one useful technique at a time and build towards safer, more repeatable work.
Find your starting pointYour learning journey: six short stages, plus a habit that applies throughout: verify what matters.
Want the wider business perspective?An provides much of the underlying intelligence. The around it determines much of your day-to-day experience, including files, web research, , projects, notebooks, connectors, scheduled work, agents and security controls.
Two AI assistants can use similarly capable models but provide very different experiences because of the tools and information surrounding them.
Choose the description closest to your normal behaviour. There is no wrong answer. This simply highlights the next useful skill.
If you told a colleague, “Prepare me for tomorrow's meeting”, they would need more information. Who are you meeting? What happened previously? What do you want to achieve? AI benefits from the same useful context.
A reasonable starting prompt
“Help me prepare for a meeting.”
There is nothing wrong with this prompt. AI simply has very little information to work with.
A more useful instruction
“I'm meeting an existing client tomorrow. They are a 120-person professional services organisation we have supported for three years. I want to understand their priorities for the next 12 months rather than simply discuss today's IT issues. Help me prepare as an experienced account manager. Give me five questions to ask followed by a suggested 45-minute agenda. Keep the language conversational and non-technical.”
Goal
Prepare for the meeting.
Context
Existing client, 120-person professional services business.
Objective
Understand future priorities.
Perspective
Experienced account manager.
Output
Five questions and an agenda.
Audience / style
Conversational and non-technical.
Many people use AI like a search engine: question, answer, finished. A more useful pattern is question, answer, feedback, improvement, challenge, final result. Every useful correction gives the assistant another clue about what “good” looks like to you.
A simple conversation
Would you question a colleague? Question AI too. If a colleague gave you a surprising figure or recommendation, you might ask where it came from, what alternatives they considered and whether the evidence is current. Use AI in the same way.
A helpful conversation can take ten minutes to build and then disappear into your chat history. When a process works, save the learning as well as the answer.
Interactive reusable prompt examples
Imagine working with somebody who forgot the client, project, objectives and preferred style every time you spoke. For ongoing work, giving AI appropriate persistent context can make the assistance more specific and consistent.
ChatGPT
Projects can keep chats, files and project instructions together. Memory, project sharing, tool access and workspace controls vary by account type and configuration.
Claude
Projects can hold focused chats, a project knowledge base and instructions. Sharing, enhanced retrieval and the newer project experience differ by plan and staged rollout.
Microsoft 365 Copilot
Copilot Notebooks can bring selected chats, files, pages, meeting notes and links together. Licensing, SharePoint or OneDrive requirements and gradual feature availability matter.
Look for the capability, not necessarily the same feature name.
A focused example: Client ABC
Useful approved background
A meaningful request
“Prepare me for tomorrow's Client ABC meeting.” The same request can now be much more useful because the work has an agreed, relevant source pack.
Ask whether you are using your organisation's approved account, whether the content is confidential or personal, how that specific service handles data and whether a third party is involved. Account type and organisational configuration matter. Do not make blanket assumptions about all AI services.
Read: Data Loss Prevention & AI ReadinessA is a controlled link between an AI assistant and another approved source. It can make information easier to find, summarise or use. It does not mean an assistant should have unlimited access.
Calendar
Documents
CRM
Business systems
Current product reality: Microsoft 365 Copilot can use Microsoft and approved external content through configured connectors, subject to tenant and source permissions. ChatGPT connected apps vary by plan, region, workspace, role, model and app approvals. Claude connector and tool capabilities change by plan, workspace and rollout. The capability may exist without being enabled in your account.
The progression is gradual. You can start with a simple question and move towards a repeatable workflow only when the task, sources, output and review step are understood.
Skills and reusable workflows
helps AI understand what it should know. A or reusable workflow helps it understand how a particular job should be performed. For example, a monthly expenditure report can consistently group information, compare it to budget, flag material differences and draft an executive summary.
Agents and recurring work
A general assistant can help with many things. An is more like an assistant given a particular job, such as a policy helper, sales assistant or onboarding assistant. It may combine instructions, selected knowledge, connected systems and tools.
Feature availability changes quickly. Microsoft 365 Copilot agents, ChatGPT scheduled and event-triggered work, and Claude scheduled tasks may depend on a licence, subscription, workspace, administrator control, connected source, model, plan or staged rollout. Confirm the actual capability and its permissions in the approved account before designing a workflow around it.
You have completed the six-stage learning journey. Keep improving useful work and verifying what matters.
Generative AI is very good at creating plausible, fluent language. That can make an answer feel authoritative, even when it has misunderstood context, made an assumption, used outdated information or invented an incorrect detail. A convincing answer is not proof of a correct answer.
Asking AI to act as an employment lawyer, accountant, doctor, cybersecurity consultant or financial adviser can help establish perspective, terminology, issues to consider and style. It does not make the assistant professionally qualified. Use appropriate professional review for material decisions.
Add useful context to one prompt today.
Ask AI to interview you before answering an important question.
Challenge one answer rather than accepting it immediately.
Wavex has helped organisations adopt changing technology since 1998. Modern AI adoption similarly needs the right combination of technology, security, data, governance, training, useful use cases and iteration. The aim is not to force every employee into one tool. It is to help people improve real work safely.
Start with one ordinary task. Give the assistant better information, challenge the output, save what works and keep people responsible for what matters.
Speak with a sector specialist. We aim to reply within 1 hour. No sales pitch - just an honest conversation about your IT challenges and how we can help.