KARL
HARTLEY.
AI researcher and implementation lead. I turn vocational standards and assessment into data that can be measured, then show organisations what to automate, what to keep human, and where AI adoption quietly fails.
UPCOMING
Taking AI for TVET quality systems
to Shanghai.
I am part of the SkillsColab team AI for TVET Quality Systems, presenting at the WorldSkills Conference 2026 with Eve Price of Solutionary. The theme is "Shaping a skills-driven future". Our case: vocational qualifications go stale because the machinery that maintains them is slow, manual and expensive. AI can rebuild that machinery, if you define what good looks like first and measure everything against it.
- check_circleConverting legacy occupational standards into contemporary competency frameworks
- check_circleGenerating assessment instruments that stay aligned to the standard
- check_circleChecking qualification architecture at national scale, not sample by sample
- check_circleKeeping humans over the loop: where expert judgement stays and where it is wasted
Research first.
Then change how organisations operate.
I am a Tairāwhiti (Gisborne) born technology and automation specialist who now calls Kirikiriroa (Hamilton) home. For more than 20 years I have designed systems, processes and learning for organisations across Aotearoa and beyond. I now run hotloop.ai, an AI process improvement consultancy where I map processes to determine what should be automated and how — whether that is RPA for repetitive tasks, AI for classification and decision-making, or custom scripting for scale.
Before Hotloop I spent five years leading an enterprise-wide CRM transformation that reshaped an entire customer journey. The system was being built in an agile manner while we rolled it out — so I had to design for uncertainty, building same-day feedback loops between developers and the frontline and tracking post-go-live performance to verify improvements landed. That enterprise thinking is where I operate best.
I then co-founded Epic Learning and spent 18 months leading its AI transformation — restructuring end-to-end processes, maintaining data integrity with AI in the loop, and training a team of 12+ on new operating models. As Principal Investigator for ConCOVE Tūhura I led New Zealand's largest study on AI-generated assessments, producing the country's first ethical framework for AI in vocational education — governance work that applies to every organisation scaling AI.
"AI can help solve problems. But if the structure and guardrails aren't orchestrated correctly, you end up with black box situations and no idea how to fix them. I design the structure that keeps humans in control while AI does the heavy lifting."
What I'm working on
A live snapshot of active projects, updated as things move.
WorldSkills Conference 2026
SkillsColab team: AI for TVET Quality Systems, with Eve Price. Shanghai, 23–25 September.
NZ standards intelligence
Every current NZQA unit and skill standard (16,279) extracted into a queryable API. The base layer for measuring quality at national scale.
Australian VET standards API
Units of competency from training.gov.au as structured data: elements, performance criteria, evidence and assessment conditions.
AI assessment suite
A drafted end-to-end process for generating, checking and moderating assessment against a standard. Built as reference for the RMIT work.
RMIT VET-AI workstreams
New research IP with RMIT on AI for Australian vocational quality systems, built on the NZ scaffolding. Planned and reviewed, build pending.
TDA Convention 2026
AI-generated assessments in TAFE delivery, with Eve Price. Brisbane Convention Centre, 6–7 May.
What is available
Through hotloop.ai I run short strategic AI missions for organisations that need a clear answer before they spend. Open to a senior role leading AI adoption for the right organisation.
START A CONVERSATION arrow_forwardAI opportunity mission. Map your processes, find where AI pays, and rank the pathways by value and risk.
What good looks like. Best practice and worked examples from national-scale research, so you start from evidence, not vendor slides.
Implementation process. The rollout plan, the governance and the feedback loops that tell you whether it worked.
Hidden loops. Why AI adoption fails quietly: unmonitored workflows, drifting outputs and the checks nobody owns. What to watch for.
Things I've built
Live tools and products that apply the same systems thinking I bring to client work.
ainews.hotloop.ai
Daily AI intelligence brief. Automated news aggregation, curated daily. No noise — just signal.
NZQA standards API
All 16,279 current NZ unit and skill standards as structured data. Lookup by ID, level, field or credits.
TGA standards API
Australian units of competency from the National Training Register, with elements, criteria and evidence.
assess.hotloop.ai
Generate, check and moderate assessment against a standard. A working demo of the method.
demos.hotloop.ai
Interactive prototypes and proof-of-concepts. See the methodology in action before committing to a build.
AI Automation & Implementation
Mapping processes, building automation, and designing the governance that keeps AI working for humans.
Process mapping & automation. Identifying what to automate and how — RPA for repetitive tasks, AI for classification and decision-making, custom scripting for scale.
Enterprise implementation. 20+ years of enterprise-scale rollouts with agile delivery, same-day feedback loops, and post-go-live performance tracking.
Human Over The Loop. Designing the governance and guardrails that keep humans in control. Structure first, then tools. Not the other way around.
Research-backed methodology. Frameworks tested at national scale — ethical AI governance, Te Tiriti compliance, and data sovereignty.
AI Generated Assessments
Where the Human Over The Loop method was built and tested. The same engine now reads every NZ standard and is the evidence base for the WorldSkills and RMIT work.
Leading New Zealand's largest study on AI in vocational education
Principal Investigator for a comprehensive study examining whether AI can create assessments that pass New Zealand's national moderation system. Built the Python automation that produced 1,000+ assessments — then developed the ethical and governance frameworks that every organisation scaling AI needs.
- check_circle Ethical framework responding to Te Tiriti o Waitangi
- check_circle 5-phase methodology for AI-generated assessment at scale
- check_circle Governance model validated by subject matter experts
Presentations & Panels
WorldSkills Conference 2026
UPCOMINGAI for TVET Quality Systems, SkillsColab team with Eve Price (Solutionary)
TDA Convention 2026
AI-Generated Assessments: Enhancing personalisation and quality in TAFE delivery (with Eve Price)
NZATD Webinar
AI Workflows and Copilot Studio
NZVETRF 2025
AI generated assessments for vocational education (with Eve Price)
AVETRA Conference
AI generated assessments for VET
ConCOVE Tūhura Conference
Future proofing assessment: AI generated design
No Frills National VET Research
AI generated assessments and inclusive VET pathways
NZATO Digital Learning Showcase
Shadow AI: the secret power shaping learning and development
TEDxRuakura Salon
Speaker
Writing & Features
Vocational Voices
NCVER Podcast: Inclusive VET systems & AI assessment.
Shadow AI
Article: How people quietly adopt AI tools at work.
LinkedIn Analysis
Regular insights on AI, automation & workflows.
WorldSkills SkillsColab
Team page: AI for TVET Quality Systems, Conference 2026.
Selected Projects
AI Process Automation
AI process improvement consultancy. Built Python automation that produced 1,000+ research assessments. AI agents in Copilot Studio, connected workflows across Anthropic and OpenAI backends. Voice tools, assessment dashboards, and content pipelines running in production.
CRM Transformation
Led training and implementation for an enterprise-wide CRM upgrade that reshaped the end-to-end customer journey. Agile delivery, same-day feedback loops, post-go-live tracking.
Standards Intelligence
16,279 NZ standards and a growing Australian corpus made machine-readable, then read end to end to find where qualifications are under-pitched, inconsistent or unmarkable.
AI Transformation
18 months embedding AI across a remote team of 12+. Restructured end-to-end processes, maintained data integrity, trained the team on new operating models.
AI Generated Assessment + State Compliance
Led design of a generative AI assessment engine that overlays state-based construction legislation onto assessment questions. Enabled providers to generate legally relevant variants instantly.
NZF Learn Platform
Developed learning dashboard and e-learning content for national coach and referee education. 4,000+ active users.
AI Activator
Developed introduction to AI module for the national AI Activator programme.