The Nexairi Dispatch · Wednesday, April 29, 2026 · Issue #8
AICPA just made AI literacy a CPA requirement
The AICPA Skills Accelerator is the first formal AI curriculum for CPAs — here's what's in it.
Good morning, friends. The AICPA launched its first AI training program for CPAs this week, the same week audit committees started asking auditors to explain AI governance — a job nobody assigned them. This issue runs deep on what that shift means for the profession. Also: auditors are seeing AI overspending show up in going-concern analyses for the first time — that's new territory.
📚 ACCOUNTING — AICPA just launched AI training for CPAs
What happened: The AICPA and CIMA launched the AI Skills Accelerator, a formal training program covering AI tools, practical workflows and governance for accounting professionals. It's the first structured AI curriculum the profession has produced at the national level.
Why it matters: Most CPAs have been building AI knowledge on their own. A formal program from the AICPA signals that AI literacy is now a professional competency — not a personal initiative. Firms that delay will fall behind clients already adopting AI tools faster than their advisors can explain them.
What to watch: Watch whether state CPA societies attach CPE credits to the program. If they do, adoption will accelerate fast.
🏛️ AUDIT — Audit committees are asking auditors to explain AI
What happened: PCAOB findings show audit committee chairs are going to their outside auditors as the primary source for AI governance guidance. The pattern is consistent: boards don't know who else to ask, so they're asking the auditors.
Why it matters: Auditors are being pulled into an advisory role they weren't trained for. For firms with audit clients, this is both a business opportunity and a liability risk if they can't answer competently. The PCAOB is watching how firms respond.
What to watch: Expect formal PCAOB guidance on auditor AI advisory responsibilities within the next year. The time to build an internal framework is now, before the guidance forces one.
⚡ FINANCE OPS — Ten finance workflows AI can cut in half
What happened: A new Nexairi guide maps 10 finance workflows — accounts payable, month-end close, expense management, cash flow forecasting and more — where AI automation can recover 10-15 hours per week without adding headcount.
Why it matters: Most finance teams are still running these workflows manually. The tools to automate them exist, cost less than a single hire and are ready to deploy. The holdback is usually knowing where to start — this guide removes that excuse.
What to watch: AI-native AP and close tools are converging on a subscription price any 3-person finance team can afford. The gap between early adopters and everyone else will show up in close-time benchmarks before year-end.
💼 BUSINESS RISK — AI spending is now a going-concern flag
What happened: Auditors are questioning whether aggressive AI investment threatens clients' long-term financial viability. Companies spending heavily on AI infrastructure without clear ROI are appearing in going-concern discussions for the first time.
Why it matters: Going-concern is a formal audit warning — it tells investors a business may not survive another year. Attaching it to AI spending is a significant shift. CFOs who can't tie AI costs to measurable outcomes will face scrutiny at the board level, not just internally.
What to watch: Watch for the first public going-concern disclosure that names AI spending as the primary factor. It will set a precedent for how boards frame AI budget decisions going forward.
Outside Nexairi
Musk seeks $134B and OpenAI's nonprofit status back — MIT Technology Review
The lawsuit over OpenAI's restructuring as a for-profit is heading to trial. A ruling either way will shape how the next generation of AI labs structure themselves.
OpenAI models and Codex now run inside AWS — OpenAI
GPT models, Codex and Managed Agents are now accessible directly within AWS infrastructure. Enterprises already on AWS can deploy OpenAI tools without leaving their existing cloud environment.
ChatGPT Enterprise clears federal security approval — OpenAI
OpenAI achieved FedRAMP Moderate authorization, opening ChatGPT Enterprise and the API to U.S. federal agencies. Government AI adoption just got a cleaner runway.
Google and Pentagon ink an 'any lawful use' AI deal — The Verge
Google reportedly agreed to supply AI to the Pentagon for any lawful purpose — a broader mandate than previous defense contracts and a signal that big tech AI is moving deeper into national security.
OpenAI releases Symphony, an open-source agent orchestration spec — OpenAI
Symphony transforms issue trackers into autonomous agent systems, turning software engineering workflows into something Codex can manage end to end.
Tool Worth Knowing: Kōan
Kōan is an observability platform for AI agents — it shows every tool call, decision and reasoning step your agents make in real time. If you're deploying AI agents in a finance or accounting workflow, Kōan is the audit trail you didn't know you needed.
Deeper Read
The missing step between hype and profit — MIT Technology Review
AI companies have built the technology and made the promises — but lack a clear path to monetization in real enterprise workflows. Worth reading before your next AI vendor pitch.
Rebuilding the data stack for AI — MIT Technology Review
Enterprise AI needs unified, governed data infrastructure — not fragmented systems — to produce trustworthy outputs. Directly relevant if you're advising clients on AI readiness.
Three reasons DeepSeek V4 matters — MIT Technology Review
Frontier-level performance at a fraction of competitor costs, plus China's push toward chip independence — the DeepSeek V4 story has more pricing implications than most coverage acknowledges.
Quick Hits
- CEO Financial Dashboard — 6 weekly metrics every CEO should track
- Is Your CPA Firm's Data Ready for AI?
- Small CPA Firms Now Get Benchmarking Data Worth Thousands
- OpenAI–Microsoft Exclusivity Ends, Multi-Cloud Era Begins
- NVIDIA Nemotron 3 Nano Omni — multimodal long-context model for agents
- Microsoft VibeVoice — open-source frontier voice AI
- Devin for Terminal — CLI agent that keeps working when you close your laptop