- Is Harkness AI just one person?
- Blake does the strategy, the client work and the build. Behind that there is a small bench of contractors: a developer for larger builds, a marketing contractor, and specialists brought in when a project needs them. So it is one person you deal with and more than one person delivering. If you want a large agency with account managers, that is not this.
- You are 23. Why should we trust you with our systems?
- Fair question, and the honest answer is that you should judge the work rather than the birthday. Blake has a mechatronics engineering degree from Massey, spent a year driving AI adoption inside MainPower, and has since worked with more than 30 New Zealand businesses. He sits on Engineering New Zealand’s AI Advisory Committee, contributes to the AI Forum NZ construction working group, and has been a finalist for the Aotearoa AI Awards Rising AI Star in both 2025 and 2026. The case studies on this site name real organisations, and he will put you in touch with them.
- What happens if you get busy, or get hit by a bus?
- Everything is built in your accounts, on your infrastructure, with your logins. Nothing runs through a Harkness AI server, and nothing depends on Blake being reachable for it to keep working. Handover documentation is part of every build rather than an extra. If Blake disappeared tomorrow, your systems would keep running and another engineer could pick up the thread.
- Do you actually build it, or do you outsource it?
- Blake builds the AI work himself. Larger software projects bring in a developer he has worked with for years, and you will be told when that happens rather than finding out later. Nothing is sent offshore to a team you have not met.
- What if AI is not right for our business?
- Then you get told that on the call, for free, and you have lost half an hour. It happens. Some processes are too rare to automate, some are too judgement-heavy, and some businesses have a data problem to solve before AI is even a question. Blake has referred work to other people when it was not a fit or when he did not have capacity. A free call that ends in "do not build this" is a good outcome, not a wasted one.
- What does it cost?
- It depends on what you are doing, and anyone who quotes you before understanding your business is guessing. Engagements range from a fixed-scope setup through to an engineer embedded in your team for a day or more a week. The call is free and you will get a real number after it, in writing, with the scope attached.
- Where does our data actually go?
- Into the tools you already run, under your own accounts. Most setups are given read-only access to begin with, so the worst case is a wrong answer rather than a wrong entry. What each connection can and cannot reach is written down before it is switched on, and you can revoke any of it without asking. If you have a policy about where data may live, bring it to the first call.
- Why Claude rather than ChatGPT?
- Mostly because of what it connects to. For New Zealand businesses running Xero and similar tools, Claude currently has the better connector story, which is the difference between AI that answers questions and AI that does work. It is not loyalty. Where ChatGPT is the better fit, that is what gets recommended, and there is a full comparison in the articles.
- Will we end up dependent on you?
- The aim is the opposite. Training is part of most engagements because a system nobody in the building understands is a liability. You own the accounts, the code and the documentation. Plenty of clients have a setup built, learn to run it, and only come back when they want the next thing.