Write a message, reuse a method or prepare work with your tools: AI becomes useful when it meets a specific need. Based in Waremme, Sterk IT helps you understand assistants and agents, then choose an approach that suits your uses.
Is this for you?
For curious individuals, sole traders and small teams who want to understand the tools before making them part of their routine. No technical vocabulary is needed to get started.
UNDERSTAND IT VISUALLY
One request. Complementary roles.
Choose an example, then select an element in the diagram to understand its role. A clear request is often enough; methods and connections are added when needed.
One goal, different kinds of support.
YOUR REQUEST
“Plan a family weekend that suits our budget and the children’s ages.”
An assistant can also act as an agent.
When needed, it can draw on…
✓
You make the final decision.
You check dates and prices, then decide before making any booking.
Illustrated example · no connection to your accounts.
AI assistant
Your point of contact.
You talk to it to ask a question, understand a topic or prepare content. It also helps you clarify what you want to achieve.
In this example
It asks for dates, budget, interests and constraints. It restates your needs before proposing an answer.
An assistant can also use tools or act as an agent. These roles can work together.
Agent
The task coordinator.
It can break a goal into steps, choose from authorised tools and adapt what happens next based on the results.
In this example
It can compare activities, check the available information and adjust the programme based on your feedback.
You define its goal, its access and when it must ask for your approval. Its results still need checking.
Skill
A reusable method.
Like a task guide: a Skill brings together instructions, templates and resources, sometimes scripts, to guide the AI through a specific task.
In this example
A method keeps track of your criteria: budget, travel times, the children’s ages and breaks. You can reuse it for another trip.
A Skill guides how the work is done. It does not retrain the model or guarantee correct results.
MCP
The connection to tools.
MCP, or Model Context Protocol, is a standard that lets a compatible AI application communicate with tools or data sources: documents, a calendar, calculations…
In this example
If the application supports it, an authorised MCP connection can access your calendar to find free dates.
MCP defines how to communicate. Access, permissions and confirmations depend on the application and its configuration.
EXPECTED RESULT
A programme to review, with options and details to check.
Available features depend on the application and the access granted. Skills and MCP are optional. These examples include human approval before booking or sharing a document.
We start with a concrete task: preparing a draft, structuring an idea, rewording text or creating working material. The tool is chosen to suit the task and the constraints around your data.
02
Build a reusable method
A clear request, useful examples and a method adapted to your task. Depending on the application, a Skill can package these instructions and resources for reuse. We test the result with your examples.
03
Explore more advanced uses
For a specific need, we assess whether an agent or an MCP connection to your tools would help. Capabilities, access, costs and approval steps are clarified before any further project. Checking the answers remains essential.
From the first conversation to the handover.
Choose your starting point
Describe a task and what you would like to improve. We establish your level, goals and available tools before defining the support.
Practise with your examples
We work with non-sensitive or anonymised examples. You use the tool, compare results and discover how to refine an answer.
Keep a reusable method
You keep a method for asking, checking and deciding when to use AI. Any follow-up work or automation is assessed as a separate requirement.
Your practical questions.
Do I need an AI subscription first?
Not necessarily. Useful tools and access are specified according to your goals. Any subscription, its limits and its cost should be identified before choosing a solution.
Can we work on a business use case?
Yes, using a clearly scoped case and appropriate data. Your organisation’s rules, authorised access and confidentiality need to be considered before using an external service.
Are the answers reliable?
They need checking. AI can invent a fact, a source or a step. Coaching teaches you to verify what matters; it does not guarantee automatic accuracy or a quantified time saving.
Do I need an agent or MCP to get started?
No. A clear request is often enough. Reusable methods, agents and connections are chosen according to the task and what your application supports. The aim is a useful solution that you understand and can control.
LET’S TALK ABOUT YOUR NEEDS
A clear starting point. A suitable next step.
Describe your situation to Cédric. The scope, arrangements and quote are agreed before you commit.