AI is now built into software that CPA firms already use, and the pressure to have a position on it keeps rising. For most small and midsize firms the question is not whether these tools matter, but how to engage with them on terms that fit a practice built on trust, confidentiality, and professional judgment. This course makes an evidence-based case that professional skepticism is an asset in technology decisions, provided it is aimed at the right target. Drawing on research in trust calibration, human and AI decision making, cognitive offloading, and professional identity, it examines why experienced professionals discount sound algorithmic advice, what that discounting costs in accuracy, and where reasonable caution turns into accidental delay. Participants locate their firm on the AI Trust and Usage Quadrant, apply five trust questions to vendor claims, and review bounded first steps designed to protect independent judgment during a pilot. The course is vendor neutral and recommends no product. This event may be a rebroadcast of a live event and the instructor will be available to answer your questions during the event.
Learning Objectives
After attending this presentation, you will be able to...
- Distinguish calibrated trust from overtrust and undertrust, and identify the professional judgment risks each one creates.
- Identify a firm's current position on the AI Trust and Usage Quadrant and the governance implications of unstructured tool use already occurring inside the firm.
- Apply five vendor trust questions covering permission, uncertainty, audit trail, explainability, and change control to evaluate an AI-enabled product.
- Select bounded first steps and rollout rules that protect independent professional judgment during a technology pilot.
Major Topics
The major topics that will be covered in this course include:
- Why one visible machine error collapses trust while identical human error does not, and what that asymmetry costs firms in both directions
- Where small and midsize firms actually stand on AI adoption, and why a firm that has not adopted broadly is not automatically behind
- Calibrated trust: the difference between misuse (overtrust), disuse (undertrust), and trust matched to what a tool actually does well
- The five-stop adoption spectrum: paralysis, accidental delay, informed caution, responsible piloting, and reckless adoption
- The AI Trust and Usage Quadrant: four firm postures and the next step each one implies
- Governed versus unstructured AI use, and why staff already bringing their own tools makes abstention a decision rather than an option
- What the evidence says about human and AI combinations, and the two task questions that predict whether a tool will help or hurt
- Reading vendor claims: the gap between a demo and production, and why fluent explanations increase acceptance of wrong answers
- Five trust questions to ask any AI vendor: permission, uncertainty, audit trail, explainability, and change control
- The NIST AI Risk Management Framework trustworthiness characteristics as a technical backbone
- Three vulnerabilities that current firm financial performance can conceal: client expectations, talent, and deferred learning
- Cognitive offloading and skill retention: what firms lose by waiting, and the conditions under which offloading frees capacity instead
- Five actions a firm can take this week, and four evidence-based rollout rules for a first pilot
