The wrong question
Most owners I talk to open with the same question. Which AI tool should I buy.
That is the wrong question. The right question is whether you are making more money. Activity is not results, and a full dashboard of AI tools can still lose you money.
If you have automated the wrong thing faster, you have just got to broke quicker. We wasted the best part of a year inside OxygenIT learning that, and the only thing it bought us was a longer list of subscriptions.
The right first question
Before you spend a dollar on AI, answer three questions.
- What is the actual constraint on the business right now. Demand, delivery, margin, or the owner’s time.
- Will this AI move that constraint, or just make something faster that does not need to be faster.
- Does the client care about this change, or is it just an internal toy.
If you cannot answer those three clearly, do not buy the tool yet.
The third one catches more bad purchases than the other two combined. A great many AI projects make an internal process feel more modern without changing anything the client experiences. That is not worthless, but it is not a return either, and it should not be funded as though it were.
What we did inside OxygenIT
We pointed AI at our own operation first, before we sold a word of advice to a client. The results were these. We cut the number of support tickets by 50 percent. We cut the time to resolve by 50 percent.
That was not a marketing claim. That was two years of internal work, with the documentation to back it, dual ISO 27001 and ISO 42001 certification, and the track record to prove it.
A 50 percent reduction in our cost to serve funded up to 50 percent off IT service cost for our clients. The honest word is “up to”, because clients on cyber and other lines see a different net number. If a provider quotes you a flat headline saving with no qualification, ask them which lines it does not apply to.
The part most firms skip
We went back to our existing clients under agreement and offered them the same reduction, not just new prospects.
That is the sentence worth taking from this piece, because it is the one that cost us money in the short term and made us money in the long term.
It is also the decision most firms talk themselves out of. The argument against it is always the same: the existing clients were happy at the old price, so why leave money on the table. It is a reasonable argument and I think it is wrong, for reasons that show up two or three years later rather than this quarter.
The bank analogy
Banks run a “new customers only” pricing model. The best rate goes to the person who has been loyal for five minutes, and the person who has been there fifteen years quietly pays more for the same product. It is a trust killer. Everybody knows it is happening and everybody resents it.
We run the opposite model. If we can lower our cost to serve, our existing clients get the saving too.
The result is in the numbers. 98 percent retention from 2022 to 2025. Clients who have been with us for two decades. Loyalty is not bought with marketing. It is bought with the same deal for the same work, whether you started last week or ten years ago.
There is a hard-nosed version of this too, if fairness does not move you. Retention is cheaper than acquisition. A client who does not leave is a client you never have to replace before you can grow at all.
How to find your own constraint
Map the work that is currently on your desk. Sort it into three buckets.
- Work only you can do. Keep it.
- Work a competent team member could do with the right system. Hand it off.
- Work that should never need a human again. Automate it, with governance.
If most of your week is in bucket three, you have your AI target. If most of your week is in bucket one, you have a constraint problem, not a tool problem, and no amount of software will touch it.
Most owners find the exercise uncomfortable, because bucket two turns out to be the biggest of the three. That is not an AI finding. That is a delegation finding, and it is worth knowing before you spend anything.
The standard for this series
Every week is something we have run inside our own firm or inside a real New Zealand client firm. No theory I have not run myself. No invented case studies.
What to do next
If you want to work out where AI would actually move your numbers, book a 15 minute call with me. If it is not a fit, I will say so. The AI Readiness Audit is the lighter option.
Week 2 of The AI Divide lands next week. Same firm, two futures.
