Applied AI agency
An applied AI agency that ships working systems, not pilots
We help teams work fluently with AI, then build and run the systems that do the work — Educate, Build, Run. Our method draws on Anthropic's AI Fluency framework (A/A/A × 4D): Automation, Augmentation, Agency across Delegation, Description, Discernment, Diligence. We're model-agnostic — we build on your provider or one we host, and recommend the model that fits the job.
Our Mission
To graduate Australian organisations through Anthropic’s AI Fluency framework (Automation, Augmentation, Agency) without selling Agency to a buyer who is ready only for Automation. The framework is the spine. The engagement is the practice.
A team, not a solo consultant in a logo
Engineering, change management, health, design, operations, HR — assembled so the four competencies are covered by people who have practised them. Description and Discernment need engineers. Diligence needs change managers and clinicians. Delegation needs leadership. We brought the people.
10 team members across leadership, engineering, health, design, and operations
Leadership
William Prosek
Founder & Director
Eight years in deep reinforcement learning, computer vision, and applied ML before generative AI was a category. The depth shows up on the Description and Discernment side of the work — knowing when an output is right, when a model is bluffing, and when a fine-tune is the wrong answer.
William founded Adaptation AI to fix a specific problem: Australian buyers being sold “AI strategy” by firms that had never shipped an agent. Adaptation AI exists to be the firm that has actually run the framework on its own work first.
Rhett Picken
Chief Strategy Officer
Over 15 years of experience in strategy, technology adoption, and digital change. Drives market positioning, technology roadmaps, and strategic growth initiatives.
Ramona Prosek
Chief Operating Officer & CFO
Experienced director and entrepreneur with a strong background in business operations, practice management, and financial administration across multiple industries.
The Team
Real fluency across the four 4D competencies means staffing them — not assumed. Engineering for description and discernment. Change management and clinical experience for diligence. Operations and HR for the delegation calls leaders have to make.
Victor
Engineering
Methodical, outcome-driven engineer with an exceptional ability to architect outcomes others wouldn’t think possible.
Kokopelli
Design
Sharp creative eye with a natural sense of balance and composition. Always across the latest in design thinking and visual trends.
Kate
Change Management
Seasoned change management specialist with years of experience guiding government and enterprise organisations across Australia through complex transitions.
Stuart
Human Resources
Psychology-trained people specialist who brings deep insight into human behaviour, team dynamics, and organisational culture to every engagement.
Brock
Support
Resourceful problem-solver who can work through just about anything on the fly. Versatile, reliable, and always ready to deliver.
Jarvis
Personal Assistant
The person who makes everything happen. Keeps the team coordinated, unblocked, and moving forward across every project.
Jamie
Health
Medical professional with years of clinical and consulting experience in health technology. Brings evidence-based rigour to our healthcare AI decision-making and systems design.
Why we exist
Anthropic publishes a framework for working fluently with AI, and it’s open: AI Fluency — three modes of engagement (Automation, Augmentation, Agency) crossed with four competencies (the 4Ds: Delegation, Description, Discernment, Diligence). It’s good. Almost no one in the Australian market is using it.
We watched the alternative play out — clients buying “AI strategy” from firms that had never shipped an agent, then writing off the budget when the pilot stalled. Most of those projects failed for the same reason: someone sold Agency to a buyer ready only for Automation, with no Delegation map and no Diligence plan.
Adaptation AI was built in Adelaide to put that framework to work from day one. Every service maps to a square on the matrix. Every proposal cites the framework by name. Every engagement begins with a free 30-minute scope-out call that places you on the grid before we sell you anything else.
What we build
The systems we’re equipped to build, each exercising a different square of the A/A/A × 4D matrix. Capability demonstrated in our own stack; client case studies are published as they clear review.
Multi-Agent Orchestration
Autonomous multi-agent architectures with hub-and-spoke coordination across distributed environments
Healthcare AI Platforms
Clinical decision-support and patient-management tooling, built to healthcare privacy and governance requirements
Agent Tooling & Extensions
Custom plugins, skills, and testing orchestrators for team-specific agent workflows
Financial Analysis Agents
Autonomous multi-agent systems for real-time financial analysis and market intelligence
Custom Model Training
Fine-tuning pipelines and training infrastructure for domain-specific small language models
Multi-Cloud DevOps
Deploy across AWS Bedrock, Google Vertex AI, and Microsoft Azure AI Foundry with session management, fleet orchestration, and model-agnostic agentic testing infrastructure.
Why Australian, why Adelaide
Diligence is the fourth D, and it gets local fast — Privacy Act, IRAP Protected, state-level health and education rules. We architect data sovereignty from the first design (AWS Sydney, Google Cloud Australia, Azure Australia East), not as a compliance wrapper bolted on at procurement.
The applied-AI opportunity in this country has a narrow window before the global firms catch up — and we’re using it.
Adelaide is home base. Defence tech, university research, and South Australian government make it the right city for the work. Remote delivery covers every other state.
How we work — the four Ds
These aren’t our values. They’re Anthropic’s competencies. We adopted them because they’re the right way to do this work, and we want every client to recognise the language when they read the framework themselves.
Delegation before deployment
We tell you what to give the AI and what to keep human before we tell you what to build. Half the “AI projects” we’re asked to scope shouldn’t be AI projects yet — and we’ll say so.
Description, written down
Skills, system prompts, and evals in version control. If a result can’t be reproduced from the brief, the brief was wrong. We rewrite the brief, not the model.
Discernment, on the record
Every output that leaves an Adaptation AI engagement has been reviewed by a named human. Observability for the autonomous parts. Audit trail for everything else.
Diligence, taken seriously
Creation diligence (which AI, used how), transparency diligence (who knows it was used), deployment diligence (who vouches for the output). We sign our work — figuratively and, when the deployment matters, literally.
Want us to tell you which rung you're on?
Ask about the Build de-risk on-ramp: we scope and build the first slice, and you pay only if it meets the acceptance criteria you set. 30-min call books straight to a real engineer.