An AI engineer in your team, without hiring one.
We scope the work with you first. Then a Harkness AI engineer joins your team one to three days a week - in your standups, in your tools, working through a build list we agreed up front. Blake Harkness leads the scoping and oversees the whole engagement.
Built for New Zealand businesses of roughly 20 to 100 staff with no AI function yet.
Who it’s for
You’ve got more AI work than a workshop can fix, and not enough of it to justify a full-time hire. That gap is exactly where this sits.
A good fit if
- You’re a business of roughly 20 to 100 staff, and nobody’s job is AI
- You’ve had the “we should be using AI” conversation and now there’s a list of things you want built
- The work touches several systems - your accounting, job management, CRM, whatever it may be
- You want the capability built inside your business, not handed back as a black box
Not the right fit if
- You’ve got one job to do - a single build or a Claude setup will get you there for less
- Your team mostly needs to learn the tools - training is the better start
- You need a senior specialist to own production machine learning - hire for that
How it works
Three stages. The first one is paid, because it’s the part that makes the rest work - the engineer turns up with a plan instead of spending a month working out your business.
- About two weeks
Paid scoping
Blake maps your systems, what can be connected and what can’t, and turns your ideas into a prioritised build list with a 90-day plan. You also get an honest call on whether you need an engineer at all - sometimes the answer is something smaller.
- 1-3 days a week
An engineer in your team
A named Harkness AI engineer works inside your team and your tools, shipping off the build list and reprioritising it with you as the business changes. Blake stays on as engagement lead, with regular check-ins and a monthly written report.
- When it ends
Handover or managed support
The engineer’s last couple of weeks go on documentation and handover. Then you run it yourselves, move onto lighter ongoing support, or hire full-time - and we can help with that hire too.
What you get
A named engineer
One person you know by name, in your standups and your chat channels, working in your systems rather than from somewhere else.
A build list from day one
The scope gives the engineer a prioritised queue, so week one is building, not discovery. It gets reprioritised with you as things change.
Blake as engagement lead
Blake runs the scoping, checks in with your sponsor regularly and writes a monthly report on what shipped and what’s next.
Connectors already built
We’ve built connectors for systems like Xero, simPRO and ServiceM8, so the engineer often starts with the plumbing already there.
Your team upskilled
The engineer works alongside your people, so the know-how stays in the business rather than leaving with a supplier.
A clean exit
Everything built for you lives in your systems and is documented at handover, whichever way you go next.
Who the engineers are
Straight answer: our engineers are early in their careers. They’re sharp and they ship, and they’re backed by three things a solo hire doesn’t get - a scoped build list, connectors we’ve already built, and Blake overseeing the engagement. You know who your engineer is before they start.
Blake has done this job himself. Part of his week is spent as a fractional AI engineer working inside client teams, including Patersons Land Professionals. That’s where the way we scope and run these engagements comes from. More about Blake.
What it looks like from the client’s side
McKenzie & Willis engaged us to build an internal AI quality assurance platform, and Alex, one of our engineers, did the work with their team. This is what their IT Manager wrote.
“We engaged Harkness AI to build an internal AI-driven quality assurance platform, and Alex has been excellent to work with throughout. He picked up a fairly niche operational problem quickly, turned it into a workable technical design, and delivered in steady increments rather than disappearing for weeks at a time. Communication was clear and honest, including when something turned out harder than first scoped. The platform is now in testing with our team and doing what we asked of it. Easy to recommend if you want practical AI implementation rather than theory.”
How it compares with hiring, an agency or a fractional CAIO
Each of these is the right answer for somebody. Here’s the short version, then where each one wins and where it falls short.
| Embedded AI engineer | Full-time hire | AI agency | Fractional CAIO or advisor | |
|---|---|---|---|---|
| How you pay | A paid scope, then a fixed monthly fee for an agreed number of days a week | Salary - often $130,000+ for an experienced AI engineer - plus recruitment and overheads | Project fee - larger builds often run into five or six figures | A monthly advisory retainer |
| Commitment | A three-month minimum, then month to month with one month's notice | Permanent employment | Fixed scope and contract per project | Monthly retainer |
| Who builds | Your embedded engineer, working through an agreed build list | Your hire - if you find the right person | The agency team, outside your business | Usually nobody - you get strategy and a roadmap, not builds |
| Speed to start | About two weeks of scoping, then the engineer starts | Often months, once recruitment is done | Weeks of scoping and contracting first | Quick to start |
| Knowledge retention | High - built inside your team and documented at handover | Highest - context and code live in-house permanently | Low - mostly leaves at handover unless you pay for ongoing support | The strategy stays, but little gets built |
| Best for | 20-100 staff businesses with a real build list and no AI function | Continuous, well-defined AI workload | One large build with a clear specification | Leadership teams that need strategy and governance before building |
Embedded engineer vs hiring an AI engineer
Hiring a full-time AI engineer is the strongest long-term option - when the workload justifies it. Everything lives in-house: the context, the code, the roadmap, and the person who owns them. If AI is core to your product, or your AI work genuinely fills a week every week, that’s where you should end up.
The catch is cost and timing. Experienced AI engineers in New Zealand often command $130,000 or more, good candidates are hard to find, and recruitment takes months. The bigger risk is hiring before the work is defined: you pay full-time rates while the business is still working out what to build.
A common path is to start with an embedded engineer, let the real workload show itself, then hire when the numbers make sense. We can help with that through our AI talent service.
Where a full-time hire wins
- Highest knowledge retention - context, code and capability live in-house permanently
- Full-week availability and ownership, including support for production systems
- Best long-term economics once there is genuinely a full week of AI work
Where a full-time hire falls short
- Salaries often exceed $130,000 in NZ, before recruitment costs and overheads
- Recruitment typically takes months in a market where AI engineers are scarce
- Expensive if the AI workload is not yet defined - you pay full-time rates during discovery
- You carry the risk of choosing the wrong person
Embedded engineer vs an AI agency
An AI agency gives you something one engineer can’t: parallel capacity. A team of engineers, a project manager, QA and documentation, all pointed at one build. For a large, clearly specified project - especially with formal procurement and fixed-scope contracts - an agency is often the right tool.
The trade-offs are cost and continuity. Agency pricing funds account management and overheads on top of engineering, the people who pitch aren’t always the people who build, and when the project ends most of what they learned about your systems walks out the door. If your need is a steady stream of smaller builds rather than one big one, an embedded engineer fits better.
Where an agency wins
- More hands - a team can deliver a large build faster than one engineer
- Structured delivery with project management, QA and documentation baked in
- Fits enterprise procurement: formal contracts, fixed scope and tender processes
Where an agency falls short
- Higher total cost - you fund account management and overheads, not just engineering
- The people who pitch are not always the people who build
- Knowledge mostly leaves at handover unless you pay for ongoing support
- Changing scope mid-project is slower and costlier than with an embedded engineer
Embedded engineer vs a fractional Chief AI Officer
A fractional Chief AI Officer or AI advisor works with your leadership on strategy, governance and a roadmap. That’s genuinely useful, especially before you’ve decided what to do. But most advisors don’t build, so you can end up with a good plan and still nobody to deliver it.
An embedded engineer is the other half. The scoping phase gives you the plan, and then someone in your team actually ships it, with Blake keeping an eye on the direction the way an advisor would.
Where a fractional CAIO wins
- Strong on strategy, governance and board-level conversations
- Useful before you know what to build, or when AI risk is the main question
- Light time commitment from your leadership team
Where a fractional CAIO falls short
- Most do not build - the roadmap still needs someone to deliver it
- Advice without delivery can stall once the workshop energy fades
- You may end up paying for an advisor and a builder separately
How to decide
Match your situation to the option, not the other way around:
- One tool, one job. Drafting, summarising or note-taking for individuals - a paid plan on a tool like Claude or ChatGPT, typically NZ$20-60 a seat each month, or a Claude setup, is probably all you need.
- A real build list, part-time workload. You want AI working across your systems but there isn’t a full week of AI work yet - an embedded AI engineer fits best.
- You don’t know what to build yet. Start with strategy - the scoping phase, a fractional advisor, or our consulting work.
- One large, well-specified build. Clear scope, formal procurement and the capability to own it afterwards - an agency is a strong option.
- Continuous workload or an AI-core product. Enough defined AI work to fill every week - hire full-time, and use an embedded engineer to bridge the gap while you recruit.
Frequently asked questions
Start with a scoping call
Tell us what you want built and which systems it touches. If an embedded engineer isn’t the right fit, you’ll hear that on the call - along with what is. Want the wider picture first? See how Harkness AI works with NZ businesses.
Last updated 18 September 2026 - launched the Embedded AI Engineer offer page, folding in the fractional vs hire vs agency comparison