AI is arriving in law firms as digital colleagues. Copilot today, agents tomorrow. The risk is not the AI itself. It is the sprawl: permissions, data exposure, unmanaged agents, and governance that cannot keep up.

As organisations move quickly from experimentation into assistants and now agents, the challenge has shifted from adoption to control. The question is no longer if you deploy AI, but how you run it safely over time.

This is not a typical security webinar. It is a fast‑paced Legal AI Risk Simulation, built around three realistic incident scenarios reflecting the failure modes IT and security leaders are already seeing.

Each scenario breaks down the failure, the risk, and the control‑plane fix, giving you a practical view of the minimum baseline needed to prevent issues before they escalate.

Key takeaways

By the end of this session, you will be able to:

• Identify where AI introduces risk across permissions, data access, and agent use, and how to reduce exposure
• Understand the governance required to run Copilot and agents safely at scale
• Apply an Enable, Measure, Protect model to make AI adoption both secure and measurable
• See how the Microsoft security stack fits together for AI: Entra, Intune, Purview, Defender, Sentinel
• Take a “run it well over time” approach, not just initial deployment

Why this matters

Legal firms are moving quickly to adopt AI, but most are still early in understanding the risks that come with it.

Unlike previous technology rollouts, AI operates across existing data, permissions, and workflows. That means issues in identity, access, and governance are amplified, not isolated.

A single gap, whether in permissions, unmanaged agents, or oversight, can lead to unintended data exposure or loss of control.

Firms that succeed will be those that treat governance as an enabler of adoption, allowing teams to move faster with confidence, not slower due to risk.

Agenda

• Safe by design vs safe by default – Why legal cannot reuse standard AI rollout playbooks
• Legal AI risk scenarios – Three real‑world failure modes and the control‑plane fixes behind them
• The control plane for agents – Governance across the Microsoft stack and how the pieces fit together
• Governance as an accelerator – Applying the Enable, Measure, Protect model in practice
• Next steps and readiness view – Understanding your exposure and how to move forward

Meet the Speakers...

Dan Robbins

Legal & Accountancy Senior Enterprise Sales Consultant at ANS

Dan Robbins is a Senior Enterprise Sales Consultant at ANS, specialising in helping legal, accountancy and recruitment firms de-risk and accelerate strategic transformation through Microsoft cloud and AI. With over nine years of experience across Microsoft technologies, Dan focuses on aligning technology to business outcomes, bringing clarity to organisations navigating the rapidly evolving AI landscape. He works closely with leadership teams to uncover core operational and commercial challenges, building practical, evidence-led roadmaps that drive measurable value, reduce risk, and transform processes.

Jason Earnshaw

Practice Lead: Low code at ANS

With over a decade of experience in Power Platform solutions, Jason is an ANS Low Code Manager passionate about helping organisations build a low-code ecosystem to empower their people to modernise how they work. As a previous Principal Architect at ANS and a Customer Success / Digital Execution Manager at Mendix, Jason strongly focuses on Low Code delivery, transformation execution and adoption, working alongside some of the largest companies and understanding what it takes to succeed with Low Code.

Mark Johnson

Head of Pre-Sales: Security at ANS

Mark has over 30 years’ experience in the IT and networking industry. With a background managing highly secure mobile networks for the Armed forces, to owning his own business, Mark joined ANS 8 years ago to take responsibility for building and delivering ANS’ suite of connectivity & Security solutions which include Azure Sentinel, SD WAN, SASE, Cloud and Multi-cloud architectures.