It agrees with you
It will err on the side of keeping you happy. Left alone that produces confident agreement with whatever you brought in, so we teach people to instruct it not to, in their own words, in every serious piece of work.
Your team already uses ChatGPT. Are they using it well?
Most teams I train are already using ChatGPT before I arrive. They have formed a view of what it does, and that view was formed on their first few tries, before anyone showed them how. This is the session that goes back over that ground properly.
The short answer
ChatGPT training is worth doing because the gap between using it and using it well is far wider than most people realise, and nothing about the tool tells you that. Usually the limit is not the tool, it is how it has been set up and what it has been given. We run sessions on your team’s own work, teach the habits that stop it agreeing with everything and inventing the rest, and end with each person owning one job that now runs through it.
Sessions run in person around Sydney, where half the value is a room comparing what everyone just typed, and online for teams elsewhere in Australia.
If you are still deciding whether ChatGPT is the right tool for your business in the first place, that question is worth reading first: ChatGPT for business
What one client said about the session
“We were already using ChatGPT, but Paul showed us how to use that and other AI tools far more effectively across the business. The session was practical, tailored to the way our business operates, and gave us workflows we could implement straight away, along with a clear vision for where AI can add value in the future.”
Hayden Williams, Managing Director, Livery DoleWhy most teams decide it is not for them
Before a session starts, most people in the room have already tried ChatGPT and already decided what they think of it. That verdict is usually the problem, and it is worth understanding how they got to it. Somebody tries it, nobody has shown them how, the answer comes back thin, and they conclude it is not for them. Reasonable. Also expensive, because everything since has been done the long way.
The second pattern is assuming a newer model fixes it. It does not. Used the same way, a better model gives you a better version of the same weak result, which is why the sessions start with what you give it rather than which version you are on.
It sounds most confident when it is wrong
That is the genuinely dangerous part, and it is the reverse of how people read confidence in a colleague. Anything you are going to rely on gets checked against the source, and we practise that in the room rather than saying it once.
Everyone stops at the first rung
Ask a question, get an answer, start again tomorrow. The step that changes the economics is a saved workspace for a job you repeat, holding your instructions and your material so nobody is rebuilding context every morning.
It is only as good as what you feed it
If the underlying material is thin or scattered, the output is too. It is an amplifier, not a substitute, and a session that ignores that is teaching people to be disappointed more efficiently.
What gets built in the room
Not exercises. The actual job somebody does every week, working by the end of the session. A few we have built with clients:
- A vehicle dealership group
A listing turned into a full sales pack: the selling features, who the buyer is, how it sits against the competition, website copy, and a reply ready to send. Then that output fed into a second workspace that produced the social posts, the email and a posting plan.
- A children’s education business
A weekly content engine: the word of the week and the puzzles built around it, with the checks written in so it audits its own work for age level before it reaches a parent.
- A professional services firm
A client report drafted from their own past reports, so the voice and format came from work they had already approved rather than from a generic template.
The full training service is AI training and enablement
How an engagement runs is the Korbai Method
What we tell people not to do
As much of the session is spent on restraint as on capability, because the mistakes are more expensive than the missed opportunities.
Do not hand it everything
You would not hire someone on their first morning and give them the P and L. Same judgement applies here, and the rules are in our own AI use policy.
Do not send anything unread
Everyone has now received the deck that was obviously pasted out of a chatbot and never checked. Being that business costs more than the time it saved.
Do not buy before you look
Find out what is already in your stack first. More than one client has discovered mid-session that they were already paying for something they had never opened.
Do not automate what needs judgement
We draft, we sort, we prepare. A person decides and a person approves anything going to a client, and where the stakes are genuinely low we will say so.
Most teams are not on one tool
ChatGPT is usually the one people brought in themselves, so it tends to sit alongside whatever the business already pays for. Where that is the case a session covers both rather than arguing for one, because the habits carry across anyway.
Questions we get asked
Our team already uses ChatGPT. What would training add?
That is the most common starting point, and it is usually the one with the most left on the table. Almost everyone we meet is standing on the first rung, which is chat: one question, one answer, start again tomorrow. The session moves people up to giving it the context and instructions it needs, and to a saved workspace for a job they do every week rather than a fresh conversation every time.
People here have tried it and were not impressed. Is that fixable?
Usually, because the problem is almost never the tool. What happens is you try it, nobody has shown you how to use it, the output is poor, you form an opinion and you file it under not for me. The opinion is reasonable and the conclusion is wrong, and the cost is that everything since then has been done the slow way. That cycle is what the session is built to interrupt.
How do you stop people trusting output that is wrong?
By teaching what it does when it does not know, which is sound more confident, not less. Two habits do most of the work: tell it not to simply agree with you, and check anything you are going to rely on against the source. They sound trivial. Written into how somebody actually works, they make a profound difference to the quality of what comes out.
We do not want to be the business obviously sending AI-written material. How do you handle that?
By taking it seriously, because it is a real risk to your name and clients notice. Everyone has now received the presentation that was clearly pasted out of a chatbot and not even read. The answer is not to ban the tool, it is to fix the two things underneath: give it your own material to work from so it sounds like you, and keep a person approving anything that goes out.
Do we need paid accounts before you can train us?
No. Sessions work on either, and we cover what changes when you upgrade so you can decide afterwards rather than guessing first. Plenty of what makes the biggest difference costs nothing: structure, context, and having somewhere to keep the work rather than starting again each morning.
What comes out of a session?
One job per person that now runs through ChatGPT, built in the room on their own work, with a named next step and a date. Plus the material behind it: the instructions, the saved workspace, and the rules about what does not go in. If nothing operates differently a month later, the training did not land, and that is the measure we hold ourselves to.
ChatGPT is a trademark of OpenAI. Korbai is independent and not affiliated with, endorsed by or a partner of OpenAI.
Get started
Get more out of the tool you already pay for.
Book a discovery call and tell us what your team uses it for today. If the honest answer is that you need better data before better prompts, we will tell you that.