Now enrolling·AI Foundations

Services

Strategy consultancies sell you analysis. We build your capability.

CoursesStage I

Build the human capability before you build the workflow.

Running an AI-native analytical workflow well requires fluency in three things at once: strategic analysis, agentic AI tradecraft, and workflow design. Courses are where the human capability gets built.

See full courses
PilotStage II

We run the methodology on your real problem. Your team is in the room.

Bring us a strategic question of your choice. We build a workflow, run it, present the outputs, and work with you on follow-up questions, alternative scenarios, and stress-testing the analysis. Typical turnaround is days to weeks, not months.

What you receive

An interactive analysis covering actors, drivers, systems maps, scenarios, and decision packages

Every reasoning step visible, every source traceable

Governance-ready output structured for boards, committees, and leadership teams

Your team in the room while we run it — first-hand exposure to how AI-native intelligence is produced

Tell us your question

Two example pilots, each led by the relevant domain principal

Strategic Intelligence Briefing

In the National Interest

Prepared for

Mallesons
Anthea Roberts

Pilot lead

Anthea Roberts

Founder, Director & CEO

Events analysed

0

Regulatory, political, economic, and technological

Strategic drivers

0

Forces shaping Australia’s regulatory landscape

Key actors

0

Institutions, industries, and stakeholder groups

Scenario futures

0

Divergent pathways stress-tested for impact

Tipping points

0

Thresholds that reshape the landscape

Read the full briefing

Investment Systems Mapping

Strategic Portfolio Analysis

70/30 Australian Pension Portfolio — 5–7 Year Horizon

Sue Brake

Pilot lead

Sue Brake

Chair & Investment Vertical Lead

The Question

What are the critical uncertainties affecting our portfolio's core exposures, and how should we position the portfolio to perform robustly across multiple plausible futures?

Macro drivers scanned

0

Political, economic, social, technological, legal, environmental

Causal connections mapped

0

How drivers reinforce, constrain, and cascade into each other

Scenario futures built

0

Divergent worlds stress-tested against the portfolio

Committee-ready paper

0

Recommendations, signposts, and no-regret moves

Read the full analysis

Augmentation, not just automation.

Generic AI helps your team draft and summarise. The Dragonfly methodology improves the quality of the thinking itself — surfacing competing views, stress-testing assumptions, and mapping how forces interact across domains.

Generic AI

Mirrors your assumptions

Output quality depends on your expertise

Single perspective, prompt-dependent

Optimises for plausible-sounding answers

No protection against fabrication

Dragonfly methodology

Challenges them systematically

Methodology is built in — 13 lens agents, 70+ analytical skills

Multi-lens synthesis across competing views

Explicit confidence levels, source attribution

Calibrated confidence levels and explicit uncertainty flags

Capability transfer partnershipStage III

A platform creates vendor dependency. A licence creates institutional capability.

Stage 3 is the partnership. Inside the partnership, three things happen in parallel — not three more stages, three concurrent layers running through the engagement.

Led by experts

Anthea and Sue lead every partnership personally.

The same principals who built the methodology guide every engagement — from kickoff through workflow co-design and analyst training.

Anthea Roberts

Anthea Roberts

Founder, Director & CEO

Professor of international law and global governance. Author of Six Faces of Globalization (FT and Fortune Best Book, 2021). League of Scholars’ world’s leading international law scholar (2019). Taught at ANU, Harvard, Columbia, and LSE.

Sue Brake

Sue Brake

Chair & Investment Vertical Lead

Former CIO of Australia’s Future Fund (grown from $70B to $300B). Pioneer of the Total Portfolio Approach — estimated to improve fund performance by 1.0–1.5% per annum over the long term.

Methodology licence

Your team gets access to the full agentic analytical system — agents, skills, workflows — and ongoing updates as we improve it. The methodology evolves with the underlying AI models. It is not a frozen snapshot.

Capability building

We upskill your analysts in three things simultaneously: strategic analysis methodology, agentic AI tradecraft (how to read an agent, write a skill, evaluate an output), and workflow design. The three are inseparable.

Workflow development

We co-build new analytical workflows alongside your team for the analyses you run most often — increasingly co-designed, eventually built by your team independently.

How one government partnership unfolded

From a single workflow to multi-year capability transfer.

Dragonfly was commissioned to build a single workflow on a strategic question of the client's choice. We developed the workflow, ran it, and presented the output.

The client asked for it to be re-run a week later. Then every three days. Then daily. That lived experience — having an analytical engine they could point at a real strategic question, see the output, iterate, re-run — convinced them this was a capability worth bringing in-house.

Talk about a partnership

What it became

Multi-year methodology licence — agents, skills, and workflows transferred into their environment, with ongoing updates

Two capability-building cohorts — around 25 analysts each, trained on strategic methodology, agentic AI tradecraft, and workflow design

An ongoing relationship co-building new workflows with their team as new analytical needs emerge

Read: Tools vs Technique — The Capability Gap
How we deploy

The methodology runs where you do.

The leverage we provide is at the methodology layer, not the harness or platform layer. We deploy on purpose-built agent harnesses so analysts work close to their actual environment.

Local-first, sovereign where required

Agent harnesses run locally on the analyst's machine. Government and regulated clients deploy the same methodology against models on sovereign infrastructure.

Harness- and model-portable

Agents and skills are not tied to one product. Compatible with leading local agent environments — Claude Code, Codex, Cowork — and frontier models. The methodology moves as the frontier moves.

Workflows and ad-hoc use

Run pre-defined analytical workflows for recurring questions, or invoke a single specialist agent on a single problem.

Dragonfly workflow

Frame the question. Challenge assumptions. Synthesise outputs.

    What
    distinguishes
    Dragonfly
    Thinking
    from
    conventional
    AI
    tools
    is
    its
    cognitive architecture
    ,
    which
    is
    layered
    over
    large
    language
    models.
    Rather
    than
    simply
    generating
    responses,
    the
    system
    guides
    users
    through
    a
    process
    of
    structured
    analysis
    and
    incorporates
    multiple
    perspectives.

    Development Policy Centre

    Independent review

    I’d
    used
    AI
    quite
    a
    lot
    but
    I
    had
    never seen a tool like Dragonfly before.
    It
    gave
    me
    something
    concrete
    I
    could
    show,
    adapt
    and
    use.

    Chloe Tallentire

    Department of Finance

    Dragonfly
    is
    the
    kind
    of
    powerful
    tool
    we
    need
    to
    build
    into
    business
    as
    usual.
    Rapid
    analysis
    in
    two hours that previously took three days.

    Bridi Rice

    CEO, Development Intelligence Lab