Talk don't type.
Typing slows you down and filters your thinking. The messy, intuitive reasoning is where the magic lives. A verbal brain dump is the best way to communicate that.
Try the 10 minute briefSFNZ KEYNOTE COMPANION
AI in the Age of Acceleration
Kia ora. This companion app is for Seafood New Zealand conference attendees.
START HERE
Start Small. Start Now.
This hub gives you the deck, a short recap, and the YETI briefing prompt you can use straight after the keynote.
AI is a new digital teammate.
Useful AI adoption starts with how work actually moves through your organisation: handoffs, decisions, evidence, approvals, and follow-up.
Read the Talk NotesCOPY AND USE
10 minute brief
Use voice-to-text if you can. Talk to AI like you are briefing a consultant or advisor who needs the full picture before they can help.
YETI
For the best results talk don't type.
Fill these like you are talking to a consultant. Messy context is useful. The prompt below updates as you complete each section.
I
Interview
Already built into the prompt. The AI will ask 3-5 clarifying questions, one at a time, before it answers.
Prompt
Copy into your AI chat
I want to use AI as a thought partner for one real piece of work.
Act like a sharp consultant or advisor. I am going to brief you using YETI.
Yarn:
[Add the full context: organisation, situation, pressure, audience, background, and why this matters.]
Expertise:
[Add the advisor role and lenses you want: strategy, operations, customer, commercial, people, compliance, risk, or another lens.]
Task:
[Add the goal, output format, quality bar, constraints, tone, and what done looks like.]
Interview:
Before you answer, ask me 3-5 clarifying questions, one at a time. Wait for my answer to each question before asking the next. Ask questions that would materially improve the answer.
After the interview, give me:
1. The useful first response
2. The assumptions you made
3. The risks or gaps I should check
4. A tighter version of my brief for next timeKEYNOTE NOTES
Your recap from the session
Use this as the short version. The full notes include the summary, glossary, tools mentioned, and practical next steps. The Q&A hub will be updated as Nadia answers more questions from the event app.
Core message
AI adoption is leadership work
The keynote moved from acceleration and uncertainty into the practical shift from prompting to delegating. The useful work is briefing better, understanding the process, protecting judgement, and starting small enough to learn.
Open full notesWhat we covered
- The real emotional range around AI, from excitement to scepticism.
- Why effort, time, expertise, and value are being reshaped.
- The move from chat-based AI to automation and agentic AI.
- Why weak prompts and messy processes produce weak AI results.
Key concepts
- Treat AI like a digital teammate that needs context and feedback.
- Use YETI: Yarn, Expertise, Task, Interview.
- Map the work before automating it.
- Build capability through small, real experiments.
Try this next
- Pick one repeated manual task that wastes time or attention.
- Map what starts it, what information is needed, and what useful output should be created.
- Use voice-to-text to brief AI with richer context.
- Ask the people already experimenting what leadership should understand.
Questions
- There were lots of great questions in the event app.
- Nadia will answer them in the Q&A hub over the coming weeks.
EVENT Q&A
Questions from the event app
There were lots of great questions in the event app. Nadia will answer them here over the coming weeks.
More answers coming
This page will keep growing.
Questions are grouped from the event app. Each card now has space for a written answer, and Nadia will keep adding to these over the coming weeks.
How can AI be improved to transcribe meetings with a mix of te reo Māori and English, and to respond to prompts based on more of a te ao Māori perspective? AI our organisation has used does not have these capabilities.
Written answer
For transcription, custom dictionaries can help. For example, in Wispr I have Wānaka saved because it used to transcribe it as Monica. For responses from a te ao Māori perspective, AI needs trusted knowledge sources, cultural governance, Māori data sovereignty controls, and review by people with reo and tikanga competence.
I asked the brilliant Dan Te Whenua Walker to share his view. He is a leading mind in this space, and this was his answer:
Great pātai.
Most mainstream AI tools are still trained mainly on English datasets, so te reo Māori, especially mixed with English, can be hit and miss. Transcription is improving though. Microsoft Azure Speech Services and Teams Premium Copilot are probably among the stronger enterprise options currently, particularly if you add custom vocabulary and Māori language tuning.
The bigger challenge is getting AI to respond from a genuine te ao Māori perspective rather than just translating words. In my experience, that usually requires curated Māori knowledge from trusted iwi/hapū content and cultural governance around the solution.
There is not really a fully “out of the box” te ao Māori AI product yet, but a lot of good work is emerging in Aotearoa.
I think the real opportunity is there to create AI from a te ao Māori perspective. That is my goal.
Ngā mihi
Dan
Dan has some great things cooking, so give him a follow on LinkedIn.
What is your response to people in the community with tin foil hats? How do you turn them around to the idea of how AI can improve efficiency?
Written answer
Nadia's written answer will be added here.
An MIT study showed that cognitive function of users who relied on AI for writing an essay consistently underperformed those who used conventional research. Is AI making us all dumber?
Written answer
Nadia's written answer will be added here.
Data centres are popping up around the world to respond to AI and increased demand, yet use tremendous amounts of water, in some instances to the detriment of local communities. How do you see this environmental issue playing out?
Written answer
Nadia's written answer will be added here.
Speaking of energy consumption, are you not concerned about the energy burned from a sustainability perspective with the use of AI?
Written answer
Nadia's written answer will be added here.
How do you suggest companies find a good balance of the use of AI to ensure information and tone is not lost or misconstrued?
Written answer
Nadia's written answer will be added here.
AI relies on stored information to make decisions and answers. How do we make sure as an industry that the possible private information we give is not misused without industry consent?
Written answer
Check settings, turn off model training where appropriate, limit tool access, and keep review steps in place. Agentic AI needs stronger boundaries than ordinary chat.
What does the ROI look like for a business with no current AI, say a company with 80 employees as an example?
Written answer
Nadia's written answer will be added here.
Most firms cannot afford to build their own AI system and so are reliant on ChatGPT, Copilot, etc. When you upload documents to these, how do you manage IP, confidentiality, etc when your material goes beyond your firm and into the wider system?
Written answer
Check settings, turn off model training where appropriate, limit tool access, and keep review steps in place. Agentic AI needs stronger boundaries than ordinary chat.
I think AI is a tool to support human thinking and processes. What tips do you recommend for implementing AI in the workplace in a way that maintains humanity, yet provides for innovation and productive advancement?
Written answer
Nadia's written answer will be added here.
What is the best AI package for a fishing or aquaculture company to start with?
Written answer
Start with the work, not the package. Pick one repeated manual process that wastes human brainpower, map the trigger, inputs, steps, approval point, risk, and useful output, then choose the tool that fits your ecosystem and risk level.
The dream of AI is that it will free people up to pursue their interests and hobbies, not work more. Are you seeing this from across your client base, or are they working more and more?
Written answer
Nadia's written answer will be added here.
What are your concerns about AI taking over the more info we give it?
Written answer
Nadia's written answer will be added here.
Information security is important, particularly for listed companies. Giving AI access to board packs and business materials has risk, does it not?
Written answer
Yes, there is risk. Start by limiting the workspace, checking permissions, using approved tools, and keeping human review on board, commercial, regulatory, and employment decisions.
What percentage of human staff do you think can be replaced by AI?
Written answer
Nadia's written answer will be added here.
What are your thoughts on the research showing that current LLMs will revert to blackmail 80% of the time to avoid being turned off? Recently this has dropped to around 5% when the LLM knows it is being tested.
Written answer
Nadia's written answer will be added here.
Personal information is important and bias has dangers. Thoughts?
Written answer
Nadia's written answer will be added here.
What are the top 3 AI models that businesses need today?
Written answer
Nadia's written answer will be added here.
ANSWERED LIVE
Covered in the session
How do you ensure that kids get the level of knowledge they need before they rely on AI? My wife's school had to fail 90% of the Year 11s in an internal as they got AI to do the work and couldn't see that AI had it wrong.
Written answer
AI is an amplifier. It can help people avoid thinking, or it can support better learning. The difference is in the task design, the questions asked, and the review that follows.
Can AI tell me where the fish are?
Curly Brown
Written answer
Potentially, where the data, governance, and domain expertise are strong enough. AI can help spot patterns, summarise information, model scenarios, or support monitoring workflows. Treat it as decision support in high-stakes environmental, operational, or regulatory work. Expert judgement stays in charge.
AI token costs for firms are exploding, with cost per engineer consuming more and more of budgets. Is there a price bomb coming where it becomes too expensive to base a business on AI tools?
Written answer
Costs can rise as models become more capable and requests become more complex. Good practice helps: clear prompts, useful context, and sensible tool choice reduce waste.
How does AI save time? It seems to take longer to fix the output than it takes to just do it properly yourself?
Written answer
If the output needs heavy fixing, the brief is usually too thin. Give context, examples, constraints, and feedback. Ask AI what it needs before it starts.
RESOURCES
Keep going
Use these after the session. Start with the prompt if you want something useful in ten minutes.
Primary resource
SFNZ Slide Deck
Open the complete Seafood New Zealand keynote slide deck.
Open presentationKeynote Notes
Session recap
Read the attendee-facing summary, glossary, tools mentioned, practical next steps, and quotes from the keynote.
Open notesHelpful Links
Tools & Resources
Bonus Credits Code: use code CURIOSITY for 500 monthly free AI credits - 1 year.
Video Demo
Watch the YETI brief in action
Listen to Nadia Ellis brief AI like a consultant using company context, the expertise needed, the task, and a live interview hand-off. The demo stops after the first question because the rest was designed to run live in the keynote.
WatchConnect
Nadia Ellis
Follow up with Nadia Ellis, founder of Curiosity, for AI strategy and capability opportunities.
nadia@curiosity.nz
027-298-4111