Question 1

What is Harkness AI?

Accepted Answer

Harkness AI is an AI consultancy founded by Blake Harkness, based in Christchurch, New Zealand. We help Kiwi businesses harness the power of artificial intelligence through practical, results-driven solutions - from strategy and training to custom AI development and fractional AI support.

Question 2

Do you work with businesses outside Christchurch?

Accepted Answer

Yes. While we are based in Christchurch, we work with businesses across New Zealand - including Auckland, Wellington, Hamilton, Dunedin, and regional centres. Most consulting and development work is done remotely, and we travel for in-person workshops when it adds value.

Question 3

What industries do you specialise in?

Accepted Answer

We have hands-on experience across marketing, education, engineering, real estate, professional services, and community organisations. Because AI is a horizontal technology, our approach adapts to any sector - we focus on understanding your specific workflows and finding where AI delivers the most value.

Question 4

What makes Harkness AI different from other AI consultants?

Accepted Answer

Blake combines hands-on AI engineering with business consulting - meaning you get one person who can both advise on strategy and build the solution. We focus on practical, measurable outcomes rather than hype, and we are deeply embedded in the New Zealand market. You work directly with Blake, not a rotating team of juniors.

Question 5

Who will I actually work with?

Accepted Answer

You work directly with Blake on strategy, workshops, and senior engineering decisions. For larger builds, Blake brings in a small, vetted network of NZ-based AI engineers under his oversight. You always have one point of contact and no account-manager middle layer.

Question 6

What services does Harkness AI offer?

Accepted Answer

We offer AI strategy and education, fractional AI support, embedded AI engineer placements (a dedicated engineer from our team 2-3 days a week with Blake providing oversight), custom AI development, and AI talent recruitment. Each service can be tailored to your business.

Question 7

What is fractional AI support?

Accepted Answer

Fractional AI support gives your business access to on-demand AI expertise without the cost of a full-time hire. Think of it as having a dedicated AI specialist on your team part-time - we help you identify opportunities, implement solutions, and continuously improve your AI capabilities.

Question 8

Can you train my team to use AI?

Accepted Answer

Yes, AI training is one of our core offerings. We run workshops and hands-on sessions tailored to your team and industry - from introductory AI literacy for non-technical staff through to advanced prompt engineering and tool-specific training. Sessions can be in-person across NZ or delivered remotely.

Question 9

Do you build custom AI products from scratch?

Accepted Answer

Yes. We design and build custom AI tools - chatbots, internal copilots, document processors, RAG systems, workflow automations, and bespoke agents. We prefer to start with a small proof-of-value build so you can see results before committing to a larger programme.

Question 10

Do you offer ongoing support after a project ships?

Accepted Answer

Yes. We offer flexible ongoing support through our fractional model. After delivery we can stay on retainer to monitor performance, make improvements, train new staff, and identify the next opportunities. Many NZ clients start with a project and transition to ongoing support.

Question 11

Can you help us hire AI talent?

Accepted Answer

Yes. Through our AI talent service we help you recruit permanent AI engineers, ML practitioners, and AI-literate product people. Blake vets candidates technically and culturally, and can also advise on role scoping, compensation, and onboarding.

Question 12

What is a Claude operating system?

Accepted Answer

It's Claude set up to run inside your business - connected to the tools you use every day, trained on your context and brand voice, with custom skills, automations, and scheduled agents handling routine work. Instead of a chatbot you copy-paste into, it works like a team member who can actually do things in your systems.

Question 13

Which tools can you connect Claude to?

Accepted Answer

Most of the tools businesses run on - Gmail, Outlook, Google Drive, Calendar, Notion, Xero, HubSpot, Slack, Microsoft Teams and many more. There are 100+ connections available, plus custom connections for in-house or niche tools. If your team uses it every day, we can usually wire Claude into it.

Question 14

How long does setup take?

Accepted Answer

Most setups go live within a couple of weeks. We start with a discovery call to map your tools and the busywork worth automating, then build and connect everything, then hand it over running. The timeline depends on how many tools and workflows are involved.

Question 15

Do we need to be technical to use it?

Accepted Answer

No. The whole point is that it runs without anyone writing code. We handle the build and the connections, then hand over something your team uses in plain English. You also get notes and ongoing support so nobody is left guessing.

Question 16

Is our business data safe?

Accepted Answer

Yes. Claude only connects to the tools and data you approve, and access is scoped to what each workflow actually needs. Anthropic does not train its models on your business data through these connections. We set permissions up conservatively and walk your team through exactly what is connected.

Question 17

What does it cost?

Accepted Answer

Every business runs on a different stack, so setups are scoped per business rather than sold at a fixed price. Book a call and we'll map out what you need and give you a clear quote.

Question 18

How much does AI consulting cost?

Accepted Answer

Cost varies with the scope and complexity of the engagement. We offer project-based pricing and monthly retainer options to suit different needs. The quickest way to get a realistic estimate is to book a free discovery call and talk through what you want to achieve.

Question 19

Do you offer fixed-price projects or only retainers?

Accepted Answer

Both. For clearly scoped builds we will quote a fixed price. For exploratory or ongoing work we use a monthly retainer or time-and-materials so you only pay for what you actually need. We pick the model that best protects your budget.

Question 20

What is the typical project timeline?

Accepted Answer

Most projects run 2 to 12 weeks depending on complexity. A focused chatbot or automation might be 2-4 weeks, while a comprehensive AI solution with custom integrations can run 8-12 weeks. We provide a clear timeline estimate during scoping.

Question 21

Is the discovery call really free?

Accepted Answer

Yes - the initial 30-minute discovery call is free with no obligation. We use it to understand your business, answer questions, and see whether we are the right fit. You leave the call with at least one concrete idea you can action yourself.

Question 22

Can you integrate AI with our existing systems?

Accepted Answer

Yes - integration is one of our core strengths. We build API integrations, custom connectors, and middleware so AI capabilities work seamlessly with your existing software, databases, and workflows. Our goal is to enhance your current systems, not replace them.

Question 23

Which AI models and tools do you use?

Accepted Answer

We are model-agnostic. Depending on the task we use frontier models from Anthropic (Claude), OpenAI (GPT), Google (Gemini), and open-source models like Llama and Qwen where appropriate. We recommend the right model for the job based on accuracy, cost, privacy, and latency.

Question 24

How do you keep AI solutions secure?

Accepted Answer

Security is a priority in every project. We follow data-privacy best practices, design secure architectures, use least-privilege access, and ensure compliance with the New Zealand Privacy Act 2020. We also help you establish clear data governance so sensitive information is handled appropriately.

Question 25

Will my data be used to train AI models?

Accepted Answer

No. We use enterprise and API tiers of major AI providers where inputs are not used for model training by default, and we document exactly where your data goes. For sensitive workloads we can design fully private architectures using self-hosted or on-premise models.

Question 26

Who owns the code and IP we build together?

Accepted Answer

You do. By default, all code, prompts, documentation, and assets we build for you are owned by your business once the engagement is paid for. We retain the right to use general techniques and non-identifying learnings, but never your data, branding, or bespoke logic.

Question 27

Will you sign an NDA?

Accepted Answer

Yes - happy to. We regularly sign NDAs before discovery calls when the conversation involves sensitive information. If you do not have a template, we can provide a simple mutual NDA.

Question 28

How do you measure success and ROI?

Accepted Answer

We agree on success metrics during scoping - typically a mix of time saved, revenue or leads generated, error reduction, or user adoption. We track these through the engagement and report back in plain language so you can see the value being delivered.

Question 29

What happens if the AI makes a mistake?

Accepted Answer

Every production AI system we build has guardrails, monitoring, and a human fallback path for edge cases. We scope AI to tasks where mistakes are recoverable, log outputs for review, and build in escalation to humans for high-stakes decisions. No AI replaces accountable humans in our work.

Question 30

Do you comply with the NZ Privacy Act?

Accepted Answer

Yes. We design our solutions to meet the requirements of the New Zealand Privacy Act 2020, including lawful data collection, storage location, access control, and breach response. For organisations with additional obligations (health, finance, government), we can adapt our controls accordingly.

Question 31

How do I get started?

Accepted Answer

Book a free 30-minute discovery call through our website. We will discuss your business, understand your challenges, and identify where AI could make the biggest impact. From there we provide a tailored proposal with clear next steps.

Question 32

What happens after I book the discovery call?

Accepted Answer

You receive a short intake form so the call is focused from minute one. On the call we map your top workflows, identify AI opportunities, and talk through risks. Within a few working days you get a written summary and, if there is a good fit, a proposal.

Question 33

Do I need technical knowledge to work with you?

Accepted Answer

Not at all. We handle the technical side and explain everything in plain language. Your job is to bring the domain expertise about your business - ours is to translate that into effective AI solutions.

Question 34

What should I prepare for the first call?

Accepted Answer

Come with a rough sense of the workflows you find repetitive or time-consuming, any AI tools your team has already tried, and one or two outcomes you would like to see in the next 90 days. You do not need a brief or a budget - that is what the call is for.

Question 35

Do you need a computer science degree to work in AI in New Zealand?

Accepted Answer

No - you do not need a computer science degree to work in AI. Most of the technical skills are learned on the job, because formal education has not kept pace with how fast the field moves. Blake Harkness runs Harkness AI with a mechatronics engineering background, not a CS degree. What matters more is systems thinking: anyone who can map a clear workflow or flowchart can now build working tools, because AI coding assistants turn that design into software. Across New Zealand, employers are hiring for AI capability, curiosity and a habit of shipping small projects far more than for any specific qualification.

Question 36

What AI skills are New Zealand employers actually hiring for?

Accepted Answer

New Zealand employers are hiring for four core AI skills. First, generative AI: prompting well, building custom GPTs, and using assistants like Claude and ChatGPT day to day. Second, automation: connecting tools with platforms like n8n, Make and Zapier and calling AI APIs to remove repetitive work. Third, building things: shipping simple web apps and internal tools, which no longer requires a CS degree. Fourth, and the most underrated, communication: being able to explain an AI solution to non-technical colleagues and scope it to a real business problem. The people who pair one technical skill with strong communication get hired fastest.

Question 37

How big is the AI skills gap in New Zealand?

Accepted Answer

The AI skills gap in New Zealand is real and widening. Blake Harkness, who founded the Young Kiwis in AI community (now 200+ members across 11+ cities), has businesses reaching out most weeks looking for young people who can help them with AI and automation - and there simply are not enough people with hands-on experience to meet that demand. For job seekers this is good news: the bar is lower than most expect. If you can use AI tools to genuinely save a business time or money, and communicate clearly, you are already ahead of most applicants. The fastest way to stand out is to build and ship two or three small projects rather than wait for a formal qualification.

Question 38

Should businesses use AI to cut costs or grow revenue?

Accepted Answer

Businesses get far more from AI when they treat it as an amplifier rather than a cost-cutter. The bigger opportunity is using AI to augment your existing team so the same people produce dramatically more - better proposals, faster responses, more leads worked - not simply replacing roles to trim payroll. Companies that focus only on cutting headcount tend to miss the revenue upside and often degrade the customer experience. In Blake Harkness's work with New Zealand businesses, the wins that stick are the ones where AI removes repetitive admin so staff can spend more time on high-value, revenue-generating work. Start by asking where AI could help your team earn more, not just where it could let you spend less.

Question 39

How much should a small New Zealand business budget for AI tools?

Accepted Answer

A practical starting budget is one good paid plan for each person who will use AI daily - typically NZ$20 to NZ$60 per seat each month - plus a small allowance for automation tools. As a transparent benchmark, Blake Harkness spends around NZ$690 a month on the full stack he runs his business on, with one tool (Claude) accounting for roughly NZ$200 of that and replacing several others. The key rule for any business: always use paid plans with model training turned off, never free tools, whenever company or client data is involved. Judge the spend against time saved - even a few hours back per person each week usually pays for the whole stack many times over.

Question 40

What should a business's first AI project be?

Accepted Answer

Your first AI project should be small, repetitive and low-risk - something your team already does often where mistakes are easy to catch. A simple approach works well: audit your week to find the most repetitive workflows, pick one, test an AI solution for about two weeks, then expand once it proves its value. Good first candidates are drafting routine emails and proposals, summarising documents, triaging enquiries, or extracting data from forms. Resist the urge to start with a big, flashy build. A focused proof-of-value that saves real hours builds trust and tells you exactly where to invest next.

Question 41

Is it safe to put business data into ChatGPT or Claude?

Accepted Answer

Yes - it is safe to use ChatGPT or Claude with business data if you set them up correctly. Use a paid or enterprise tier, turn off model training in the settings, and keep genuinely sensitive client data out of any free tool, because free tiers can use what you type to train future models. A simple traffic-light rule helps: green for low-risk internal tasks, amber for anything customer-related that needs care, and red for highly sensitive data that should stay in private or self-hosted systems. Harkness AI designs solutions to meet the New Zealand Privacy Act 2020, with clear control over where data is stored and who can access it.

Question 42

Is ChatGPT or Claude better for business use?

Accepted Answer

For most New Zealand business use the better tool depends on the task, but Blake Harkness favours Claude for day-to-day work because of its MCP integrations - connections that let the AI work directly inside your files, databases and software instead of forcing constant copy-paste between apps. That tighter integration is why he cancelled a two-year ChatGPT subscription. That said, Harkness AI stays model-agnostic: ChatGPT, Claude and Gemini each have strengths, and the right choice comes down to accuracy on your specific task, cost, data privacy and speed. The practical advice is to pick one as your daily driver, learn it deeply, and only switch a task to another model when it clearly does that job better.

Question 43

How much does it cost to build an AI automation or agent for a NZ business?

Accepted Answer

Building a custom AI automation or agent for a New Zealand business typically ranges from a few thousand dollars for a focused automation up to NZ$50,000 or more for a comprehensive system with custom integrations. A single well-scoped workflow - say an AI agent that qualifies inbound leads, or a tool that drafts a routine document - usually sits at the lower end and can be live in two to four weeks. Harkness AI recommends starting with a small proof-of-value build rather than a large programme: you see real results, keep risk low, and learn exactly where the next investment should go before committing more budget.

Question 44

What is speed to lead and why does it matter for sales?

Accepted Answer

Speed to lead is how fast a business responds to a new enquiry, and it is one of the biggest sales levers most New Zealand businesses ignore. A three-year study of 15,000 leads across six companies found that reaching a lead within the first hour makes you about 10 times more likely to have a real conversation, and reaching them within five minutes makes you about 21 times more likely to qualify them as a genuine opportunity. The problem is that few teams can answer every enquiry instantly. That is why Blake Harkness builds AI voice agents - including ones with a natural New Zealand accent on a real NZ phone number - that respond to every lead within seconds, day or night, so none slip through the cracks.

Question 45

How do New Zealand businesses get found in AI search like ChatGPT and Perplexity?

Accepted Answer

To get found in AI search tools like ChatGPT, Claude and Perplexity, New Zealand businesses need Generative Engine Optimisation (GEO), which is different from traditional SEO. SEO aims to rank a page in a list of links; GEO aims to get your content quoted directly inside an AI answer. The tactics that work: publish clear, structured content that answers real questions the way people actually ask them, add FAQ and structured-data markup so machines can parse it, and - most importantly - include unique, first-hand information that exists nowhere else, such as your own data, results and expert opinions. AI models cite the original source of useful facts, so the goal is to be that source. This page is built exactly that way.

Question 46

What did Blake Harkness learn about AI from the New Zealand delegation to China?

Accepted Answer

On an AI delegation through China, Blake Harkness saw artificial intelligence woven into everyday infrastructure at a scale that is still rare in the West. Highlights included Hangzhou's City Brain running legal AI at around 90% accuracy, an agentic hospital at Tsinghua University training doctors on tens of thousands of simulated cases, a major bank running more than 400 internal AI assistants, and a humanoid robot capable of doing the switching work on an electrical substation. The bigger lesson for New Zealand businesses was less about any single robot and more about iteration speed: China's edge now is how fast it tests, fails and rebuilds. The takeaway is to start adopting AI in real workflows now rather than waiting for the technology to settle.

Frequently Asked Questions

About Harkness AI

Services

Pricing & Timelines

Technical

Process & Trust

Getting Started

AI Insights

Still have questions?