Knowledge base
AI, explained the way I explain it in the room.
No jargon. Each idea is tied to something you already know from running a business or a household: think of AI as a capable new employee you are learning to brief.
The terms, in plain words
The twelve words you will hear most, each with the comparison I use in workshops.
AI
Software that can understand instructions and produce text, images or analysis. Think: a new employee who has read an enormous amount.
Chatbot
A tool you talk to by typing, such as ChatGPT, Claude or Gemini. It answers in text. Think: your text-based assistant.
LLM (large language model)
The technology behind chatbots. You do not need to understand it deeply. Think: the brain of your AI employee.
Prompt
The instruction or question you type. Think: your brief to the employee.
Hallucination
When AI confidently gives wrong or made-up information. Think: a confident junior who is sometimes wrong and never says "I don't know".
Context
What the AI remembers during the current conversation only. Think: briefing someone at the start of a meeting. After the meeting they forget, unless told again.
Image generation
Creating pictures from a text description. Think: a graphic designer on demand.
AI agent
AI that completes a task with several steps and less supervision. Think: a trained employee who handles a whole process on their own.
Automation
Connecting AI to your tools so tasks happen without you typing each time. Think: a system that keeps running after everyone has gone home.
No-code tool
Software for building apps or chatbots by clicking, not programming. Think: furniture from a kit. No carpentry needed.
System prompt (instructions)
The standing rules you set for how an AI tool should always behave. Think: the employee handbook.
Connector
A link that lets AI read from, or act in, tools you already use, such as your email or files. Think: giving the employee a login.
Writing a prompt that works
A generic prompt gets a generic answer. Brief AI the way you would brief a new hire on their first day.
- Say who it is writing for. "For my saree shop's website", "to a supplier we have worked with for ten years".
- Say exactly what you want. An email, a list, a table, a reply. How long.
- Give it your facts. Names, dates, your policy, what happened. It cannot guess them.
- Set the tone and the limits. Polite but firm. Under 80 words. No exaggeration.
Then read what comes back and tell it what to change. The second version is almost always better than the first.
A weak prompt
Write an email to a customer about a delay.
The same request, briefed properly
I run a furniture workshop. Write a short, polite email to a customer, Mrs. Banerjee, whose dining table will be delivered on the 24th instead of the 17th because the polish needs longer to cure in this humidity. Apologise once, offer free delivery, under 90 words, no excuses.
Three golden rules
Every session ends on these. They matter more than any trick.
Privacy
Never paste customer phone numbers, financial details or confidential data into free AI tools.
Verify
Always double-check numbers, dates and anything legal or tax-related. AI can be confidently wrong.
Context wins
The more you tell AI about your own business, the better the result.
Prompts you can copy today
Replace the parts in square brackets with your own details. Workshop participants take home the full cheat sheet.
A difficult email or message
I need to write a [polite but firm] message to a [customer / supplier / landlord] about [situation]. The message should [goal]. Keep it professional, under [80] words.
Meeting notes into an action list
Here are my meeting notes: [paste notes]. Turn this into an action list with three columns: task, person responsible, deadline.
An SOP for new staff
Create a step-by-step SOP for [task]. Here is roughly how it works: [explain in two or three sentences]. Make it clear and numbered, for a new staff member.
What customers are really saying
Here are recent customer reviews: [paste reviews]. Tell me: 1) what customers love, 2) the top two or three complaints, 3) one action I could take this week.
Common questions
Do I need to know coding to use AI at work?
No. Everything in the workshops is done by typing ordinary sentences. If you can write a WhatsApp message, you can brief an AI tool.
ChatGPT, Claude or Gemini: which should I use?
They are more alike than different, and each has tasks it handles better. The honest way to find out is to give all three the same request and compare, which is exactly what we do in the session. Start with whichever one you already have access to.
Is it safe to put my business information into AI?
Treat a free AI tool like a helpful stranger: fine for drafting and thinking, not for customer phone numbers, bank details or confidential documents. Paid business plans come with stronger privacy terms, and it is worth reading them before you rely on one.
Can I trust what AI tells me?
Trust it like a capable junior: useful first drafts, occasional confident mistakes. Check every number, date, name and legal or tax point before it leaves your hands.
Will AI replace my staff?
In small businesses it mostly takes over the repetitive parts of a job: the same reply typed forty times, the report rebuilt every month. That usually frees people for the work customers notice.
What is the difference between a workshop and consulting?
In a workshop you learn to use AI yourself. In consulting I build a specific thing for your business and hand it over working. Many people do the first and then ask for the second.