Role play context
Background and scenario framing for the Defence leadership meeting role play. The pre-meeting briefing (linked below) is the artefact I'd walk through during the meeting.
The meeting
First meeting with Australian Defence leadership
Mistral AI has secured a 30-minute introductory meeting with three senior Defence leaders. This is the first time Defence has engaged directly with Mistral. The meeting was facilitated through an existing relationship with the CIO's office. The objective is discovery — understand their priorities, demonstrate credibility, and propose a value discovery workshop as the next step.
The pre-meeting briefing is the customer-facing artefact I'd send to the CIO's office in advance and step through during the meeting. It's designed to demonstrate preparation, frame Mistral's value proposition around Defence priorities, and create a structured conversation rather than a generic pitch.
Defence personas
Prof Tanya Monro AC
Physicist and Chief Defence Scientist. Leads DSTG — Australia's defence science and technology organisation. Her mandate is asymmetric capability advantage through science and technology. She evaluates every technology proposal against scientific rigour and evidence. Published researcher with deep expertise in photonics and sensing. Appointed Companion of the Order of Australia.
Tom Hamilton
CDI since July 2024. Leads the Defence Intelligence Group (DIG) — responsible for intelligence collection, analysis, and dissemination across the ADF. Operates at the intersection of multiple classification levels and Five Eyes intelligence sharing. His analysts are drowning in data volume that exceeds human processing capacity.
Chris Crozier
Defence CIO since August 2023. Background in BHP and Orica — brings a commercial enterprise lens to Defence IT. Responsible for Defence's digital transformation, cloud migration (Azure-anchored), SAP ERP modernisation, and sovereign infrastructure strategy. The meeting gatekeeper and likely first point of ongoing engagement.
Dr Paul Robards
Owns the Defence Data Strategy — the strategic framework for how Defence collects, manages, and exploits data. DDG (Crozier's group) delivers the technology and infrastructure that implements this strategy. Robards is the bridge between strategic data intent and technical execution. Not in this first meeting, but a critical future stakeholder for any AI/ML engagement.
Pre-meeting briefing walkthrough
The Defence briefing is structured as a consultative conversation guide, not a slide deck. Each section is designed to open a discovery thread with a specific persona.
Executive summary
30-second positioning. Sovereign by architecture, domain-adapted for Defence, edge to enterprise. Three cards that establish Mistral's value proposition immediately — designed to land with Crozier (sovereignty), Monro (domain adaptation), and Hamilton (edge deployment).
Strategic alignment with Defence priorities
Six cards mapping Mistral capabilities to NDS priorities. Each card is a discovery hook — "Is this how you're thinking about [decision advantage / sovereignty / interoperability]?" Opens conversation rather than presenting at them.
Why domain adaptation matters
The core thesis. Contrasts the chatbot era with the agentic era, then proves the point with three case studies — Singapore HTX (direct Mistral partnership), US Army TRACLM (military domain LLM), and medical coding (RAG's sharpest failure). Designed for Monro — demonstrates that domain adaptation is not optional, it's structural.
The customisation ladder
Visual showing 8 levels of model customisation from prompt engineering to embodied AI. Designed for Monro — demonstrates technical depth and positions Mistral as the only provider covering the full spectrum. Key talking point: "Where on this ladder does Defence need to be?"
AUKUS alignment
Addresses the allied interoperability requirement head-on. Open-weight models solve the data sovereignty problem architecturally — each nation adapts the same base model without exposing classified training data. References Mistral's existing Five Eyes-adjacent partnerships in Singapore. Designed for Crozier and Hamilton.
Use case mapping
Five cards connecting specific Defence programs to Mistral capabilities — TAS (trust through domain adaptation), Ghost Bat/Ghost Shark (edge AI via Ministral), DAIRN (co-development partnerships), Ethics & Assurance (open-weight auditability), and DAIC milestones (military planning frameworks). Each card is a conversation starter for the relevant persona.
Success with Mistral
Testimonials from Singapore MINDEF, French AMIAD, and Luxembourg government. Social proof from defence and sovereign institutions. Builds confidence that this is proven, not experimental.
Proposed next steps
The close. Four-step value discovery workshop — not asking for a deal, asking for a structured half-day session. Lowers the commitment bar while creating a concrete next step with a defined timeline.
Questions
Provocative questions grouped by persona. Designed to shift from presentation mode to dialogue. Each question frames inaction as the risk and domain-adapted AI as the unlock.
How I'd run the 30 minutes
Set the frame
Thank them for the time. Establish that this is a consultative conversation, not a pitch. "We've done our homework on Defence's priorities — we'd like to check our understanding and hear directly from you."
Discovery by persona
Walk through the briefing sections, but pause on each to ask discovery questions. Monro: "Where on the customisation ladder does DSTG need to be?" Hamilton: "What's the biggest bottleneck in your intelligence processing pipeline?" Crozier: "How does Defence think about AI vendor risk in the sovereign architecture?"
Position Mistral against what we've heard
Reflect back their priorities and map Mistral's capabilities to what they've told us. Use case studies selectively — only the ones that are relevant to what they've said. This is where the Singapore/France references earn their weight.
Propose the next step
Value discovery workshop — a structured half-day session to map use cases, assess data readiness, identify the right starting point on the customisation ladder, and define a phased engagement plan.
Handle curveballs
Reserve time for objections and unexpected questions. Most likely: "Why not OpenAI?" / "We're already using Azure AI" / "How do you handle classification?" These are addressed in the briefing but may need live handling.