It's late 2025. AI is everywhere. In every app, every business conversation and every article. Look at you now, reading yet another article about AI…
And your team? Do you know how often they’re using AI? It’s probably more than you think. So the question you need to be asking isn’t whether your business should be using AI, it’s whether AI is working for you or against you.
The problem: AI black hole
For many businesses, AI isn't delivering. Teams are spending hours in back and forth loops trying to get it to solve one problem, ending up with worse work than if they'd just done it themselves. You've seen it. The document that took two days instead of two hours, full of robotic phrasing and zero substance. The image that looks almost right, but somehow completely wrong. The Copilot session that turned a simple spreadsheet into three hours of wrestling with overcomplicated formulas that still don't work.
When organisations apply AI haphazardly, with no vision, no direction, no training — they end up working for AI. Productivity vanishes into a black hole of endless prompts and disappointing outputs.
The difference: it's about people
Research from Google shows 88% of organisations that adopted AI early are seeing positive ROI. But here's what matters more: they're being deliberate about implementation.
The businesses getting real value aren't using better AI tools. They're leading their people differently. They've created a clear vision of what AI can bring. They have structured practices around when and how to use it. They've trained their teams to actually get the most from the tools.
The businesses that aren't? Same situation, different choices.
He aha te mea nui o te ao? He tangata, he tangata, he tangata What is the most important thing in the world? It is people, it is people, it is people.
People need leadership that empowers them with vision, practice and training to get real value from AI. That's the actual difference.
How we're making AI work at Hemisphere
We've used AI for over two years. Early on, we realised that empowering our people with AI meant more than just giving them logins. To make AI actually transformational – for both our team and our clients – we needed to think bigger. So we built a framework: vision, practice, capability.
VISION: HI + AI (HUMAN INTELLIGENCE + ARTIFICIAL INTELLIGENCE)
We blend two sides of thinking. Left brain: data and analysis. Right brain: creativity and intuition. That's Human Intelligence, which provides strategic thinking, cultural understanding, and critical judgement. The things that only come from people, that no algorithm can replicate.
Our vision is to amplify that thinking with AI. Human intelligence handles judgment and meaning. AI handles speed and scale. AI expands what we can explore; people decide what matters.
PRACTICE: WHEN TO USE AI (AND WHEN TO STOP)
Having a vision is the start. Making it real requires clear guidelines around when and how to use AI. Our practice evolves as we learn. You can't stay still in this space — something new comes along almost every day.
Choose the tools deliberately. We use Claude and Midjourney because we've assessed them, trained on them, and know them well. Everyone on our team gets consistent results from the same tools.
Know what AI does well. Generating frameworks. Researching topics. Editing for clarity. Restructuring information.
Know what AI doesn't do. Critical thinking. Pushing back on bad ideas. Feed AI a bad idea and it’ll respond with “this is great” when nothing could be further from the truth.
Grasping cultural context. Understanding te reo Māori or tikanga. Good luck getting general purpose AI tools to nail te reo — we've had clients try — and then come back to us to get it right. Any cultural work needs a person with cultural and ancestral intelligence. Getting this wrong isn't just inaccurate. It's disrespectful. Nothing will work here unless it’s led and shaped by indigenous knowledge.
Use AI in two areas. Ideation — when you're stuck, exploring angles, breaking through blocks. Editing — tightening language, restructuring, adapting content for different audiences.
Treat AI like a new team member and apply quality control. You wouldn't hand a brief to someone fresh and assume perfection arrives. You'd check their work, give feedback, refine things together. AI deserves the same scrutiny. Nothing leaves our office without significant human review and refinement.
Before anything goes out the door, ask three questions:
Build collective knowledge. When someone discovers an effective prompt, they add it to our library. No one reinvents the wheel. We've also built projects loaded with Hemisphere's guidelines, brand voice, and context that anyone can use. Everyone starts with the right context already loaded. As AI evolves, we update these resources—what works today might be outdated very quickly. We're constantly learning, sharing, and adapting as a team.
CAPABILITY: PROMPT WELL. TRAIN CONSTANTLY.
Vision and practice don't work without people who can execute. The difference between useful output and AI slop comes down to the prompt. Garbage in = garbage out.
At Hemisphere, we use what we call the Integrated Search Framework. It's built on a simple premise: the fundamentals of good marketing haven't changed, but the distribution channels have multiplied.
We train our team using RISEN: a framework that structures prompts for better results.
Role: tell it what role it's taking
Instructions: explain what you want
Steps: specify the structure
End goal: define the purpose
Narrowing: set constraints like word count or tone
Bad prompt: "Research customer buying motivations for me."
Good prompt: "You're a market research assistant helping an advertising agency. I need a research summary on customer buying drivers, comparing rational factors (price, features, specifications) versus emotional factors (brand perception, lifestyle alignment, status). Focus specifically on the New Zealand market using data from 2023-2025. I need this to inform our creative strategy for an upcoming campaign. Structure it as: 1) Key statistics on each driver type with percentages, 2) NZ-specific consumer behaviour patterns, 3) How these drivers differ by age demographic. Critical requirement: Every fact, figure, and claim must include a verifiable source with a clickable link. If data isn't available or reliable, flag the gap rather than speculating. Limit to 600 words so I can quickly extract insights and share relevant sections with the team and client as supporting evidence for our creative recommendations."
See the difference? Specificity gets better results. The first prompt gets you generic waffle. The second gets you actionable research with sources you can actually use. AI quickly handles work that would've blown your research budget. You get a resource that strengthens both your strategy and your client justification.
We also set up team profiles with our writing style, brand guidelines, and typical project types. When someone starts a new conversation, they can quickly give it Hemisphere context rather than starting from scratch. This one-time investment pays back every single day. (Turns out, teaching AI about yourself once beats re-explaining it 47 times a day!)
The path forward
Kāhore taku toa i te toa takitahi, he toa takitini. Success is not the work of one, but the work of many.
Vision. Practice. Capability. That's how you make AI work for your business and your people.
The best time to start was 2 years ago. The next best time is today.
Want to have a chat about making AI work for your business? We can help you build the vision, practice, and capability that gets real value from your team. Let's kōrero.
Craig Adolph is Head of Digital at Hemisphere in Tauranga, leading digital transformation initiatives and regional expansion across the Bay of Plenty and Waikato.
.svg)
.svg)
.svg)
