Guide · AI strategy and advisory
What you should get from AI advice, what a useful roadmap looks like on one page, when you need one, and what it should cost in New Zealand.
An AI advisor works out where AI will pay off in your business, where it will not, and what to do first. That is the whole job. Anything that does not lead to one of those three answers is padding.
In practice it is less glamorous than the title suggests. Most of the time goes on watching how a piece of work gets done today: who touches it, how often, where it waits, and what happens when the input is wrong. I ask why each step exists. The two answers I hear most are "because that's how we've always done it" and "because that's how the software works", and neither is a reason to automate the step. Often it is a reason to remove it.
Then the candidates get sorted. Some jobs need AI because someone has to read an email or a document and make a call. A lot need plain rules-based automation, which is cheaper and more reliable. Some should be left alone because the volume is too low to bother. Being told which is which, with a number against each, is what you are paying for.
For a small or mid-sized business, a useful AI roadmap fits on one page. If it runs to forty pages, nobody will open it twice. These are the five parts I would expect to see.
| Part | What it says | Example |
|---|---|---|
| Projects, in order | Three to five candidates, each with a rough cost, a rough saving and a named owner inside the business. | 1. Supplier invoices into Xero: about $3,000, saves about six hours a week, owned by the office manager. |
| What not to do yet | The ideas that were looked at and parked, with the reason. | A customer chatbot: too few enquiries a week to pay back. Look again when volume doubles. |
| Tool and data rules | Which assistant staff use, what must never go into it, and where data is kept. | Company accounts only, no client financials in prompts, anything built is hosted in Sydney. |
| How it will be measured | The baseline for the first project, taken before anything is built. | Invoices take 11 minutes each today, 140 a month. Measured again six weeks after go-live. |
| Review date | When the list gets looked at again. | Six months. The tools will have changed by then, and so will your priorities. |
The figures in that table are made up to show the shape. Yours come from measuring your own process, which is step two of my 8-step implementation guide.
Most businesses under about 50 staff need a first project and a one-page plan, not a strategy document. You learn more from one small thing running in production than from three months of planning, and the plan gets better once you have seen what the first project actually did.
There are cases where the written plan should come first:
Good advice starts with questions about your week and ends with something you could hand to a different supplier. If the plan only works with the adviser's own product, what you were given was a quote.
I have written the longer version separately: how to check an AI consultant is legit before you hire them.
For a small business, working out what to do first should cost nothing. That is what a free first call is for. My own approach: discovery is always free, and you leave with a recommendation you can act on, whether or not you hire me to build any of it. Nothing is billed until you have agreed a fixed price in writing.
A fuller roadmap is real work, and it is fair to pay for it. That is the case when several teams are involved, people need to be interviewed, there are workshops to run, or a board wants a paper. I quote that as a fixed price after the first call, or it becomes part of a fractional CTO retainer where the roadmap is kept up to date month by month.
Three things to be wary of when you are quoted for advice:
Build the first item on the list, small, and measure it against the baseline before starting the second. A first project should be working in two to three weeks and running against real data in test mode before it touches anything live.
Two things are worth writing into the plan from day one. The first is governance: who signs off what the system does, and what it must hand to a person. That is far easier to design in than to bolt on later. The second is portability. The best model for a job in March is often not the best one by September, so whatever gets built should let you swap the model underneath with a configuration change, not a rebuild.
Then look at the page again in six months. Cross off what is done, re-order the rest, and add what you have learned.
An AI advisor looks at how work is done in your business today, works out which jobs are worth automating and which are not, and writes down what to do first, what it should cost and how you will know it worked. A good one also tells you when the answer is a process fix or a tool you already pay for.
For a business under about 50 staff it should take days, not months: one or two conversations, a look at the process and the numbers, and a written page. Larger organisations with several teams and regulated data need longer, usually a few weeks of interviews and workshops.
Usually not a long one. Most small and medium businesses get further with one well-chosen first project and a short written plan for what comes next. Larger organisations with many teams already using AI benefit from written priorities and a clear policy on tools and data first.
For a small business, working out what to do first should cost nothing: it is what a free first call is for. With me, discovery is always free and nothing is billed until you have agreed a fixed price in writing. A fuller roadmap across several teams, with interviews and workshops, is fixed-price work quoted after that first call, or part of a fractional CTO retainer.
Mostly the label. Both should tell you where AI will and will not pay off. The more useful question is whether the person giving the advice can also build what they recommend, because a plan written by someone who has never shipped one tends to be hard to deliver.
Bring the plan you have, or just the list of things that are slow, to a free 30-minute call. You will get an honest read on what to do first and what to leave alone. Discovery is always free.
Book a free call