Beyond the Chatbot: What It Actually Takes to Deploy AI Safely in Your Firm

General AI chatbots are everywhere in accounting firms today, approved or not, but they hit a wall fast: every new conversation means rebuilding context from scratch, and every upload is a judgment call about security. This session breaks down the real difference between general-purpose AI and a platform-native system built with firm data already in context, including what recent IRS guidance means for how firms should be reviewing AI-assisted work. We'll ground the discussion in a live look at Coworker, Canopy's native AI execution layer, and the human-in-the-loop framework that keeps your team in control of every output before it reaches a client.

Learning Objectives:

1. Identify the limitations of general-purpose AI tools when handling firm and client data.
2. Differentiate the security tradeoffs between connecting outside AI to client data via API versus a closed, platform-native system
3. Understand how recent IRS guidance on AI use affects review requirements for AI-assisted work.
4. Apply a human-in-the-loop review framework, from always-allow to review-required, to reduce risk in AI-assisted workflows.
5. Explore practical, firm-wide automation use cases, including file routing, scope creep detection, and client communication drafts.

Presenter and Moderator Bios

Presenter: Luke Frye, CPA, Accountant in Residence, Canopy

Luke Frye, CPA, is the Accountant in Residence at Canopy, where he brings a background in accounting tech startups and private practice to help firms build smarter, more scalable businesses. After growing his own bootstrapped firm to $500K ARR and helping platforms like Bench and Pilot reach thousands of clients, Luke now focuses on one mission: helping firms add $10 million in collective revenue this year. He bridges modern software and firm operations so practitioners can reclaim their time and build firms that are not just profitable, but truly valuable.