I spent more hours than I planned this weekend going through everything that happened in AI in March. Not skimming headlines. Actually reading the announcements, the research papers, and the regulatory filings.
My honest takeaway: this was not a normal month.
The pace of releases from Anthropic alone would have been newsworthy in any prior year. Add a federal court ruling, three major regulators moving simultaneously in different directions, and Goldman Sachs publishing research that quietly confirmed what most CFOs have been privately thinking about AI ROI, and you have a month that deserves more than a quick scroll.
This edition covers five clusters of news from March. Each one has a direct implication for how you run your finance function. The free section gives you the full picture. The paid section tells you exactly what to do.
I am starting something new this week. One short poll per edition, results shared the following week.
Why? Because I know you are curious what other finance leaders are thinking. We are a community of 1,000+ CFOs and finance professionals. That is a meaningful signal. What we know, what we use, what we are struggling with, that is worth sharing.
Answer below and I’ll share the results next Tuesday
What is the biggest barrier to effective AI implementation in your finance team today?
March News Lineup
1. Anthropic had a remarkable March
No other AI company moved as fast as Anthropic in March. The volume of releases was unprecedented, and several of them matter directly for finance professionals.
Here is what shipped:
Memory for all users including free tier, with a ChatGPT and Gemini import tool (read more)
Claude Marketplace launched with tools from Harvey, Snowflake, Replit, and others, zero commission, applies toward existing Anthropic spend (read more)
Claude for Excel and PowerPoint upgraded with shared context across both apps, plus integrations with S&P Global, LSEG, Moody's, FactSet, Daloopa, and PitchBook (read more)
Interactive visualizations, charts, diagrams, and widgets built inline, available on all plans (read more)
1M token context window at standard pricing, no surcharge (read more)
Claude Dispatch, assign tasks from your phone, get results on your desktop (read more)
Computer Use in research preview, Claude can operate a Mac desktop autonomously (read more)
Microsoft 365 Copilot integration, Claude now available inside M365 apps (read more)
Claude Partner Network launched with $100M commitment; Accenture training 30,000 professionals on Claude (read more)
Mythos model accidentally leaked, described as a step change above Opus, currently in early-access testing (read more)
So what: Anthropic is no longer competing on individual features. It is building an ecosystem. The financial data integrations, the Excel add-in, the enterprise partnerships, these are moves aimed directly at the finance function. That said, Claude experienced at least five outages in March, including two significant ones in the final week. The capability is real. So is the reliability concern.
2. The AI-in-Excel war is officially on
Finance teams live in Excel. That has not changed. What has changed is what Excel can do when AI sits inside it.
In March, three tools shipped major finance-specific Excel capabilities in the same month:
ChatGPT for Excel launched with GPT-5.4, direct integrations with FactSet, LSEG, S&P Global, Moody's, and Daloopa, and the ability to build financial models via natural language (read more)
Claude for Excel upgraded with cross-app context sharing with PowerPoint, plus the same institutional data vendor integrations (read more)
Microsoft Copilot in Excel now edits local workbooks without requiring cloud storage, and pulls context automatically from emails, meetings, and files (read more)
Google Gemini 3.1 shipped with significantly improved Sheets analysis capabilities (read more)
So what: The outputs are cleaner than they were six months ago. The formulas are reliable. This is not a demo anymore. One limitation: none of these tools currently carry memory across sessions. Every time you open a new session, you start from scratch. That matters for recurring workflows. But for ad hoc analysis, model building, and one-off tasks, the capability is there and worth your time.
3. Regulators moved, in three different directions
Three major regulatory developments landed in March, and they are pulling in different directions.
The EU Council and Parliament aligned on extending high-risk AI compliance deadlines from August 2026 to December 2027. Credit scoring, insurance underwriting, and fraud detection systems get more time. However, the original August 2026 deadline remains the legal default until the Digital Omnibus is formally finalized. Penalties remain up to 7% of global turnover. (read more)
The White House released its National AI Policy Framework recommending federal preemption of state-level AI laws and a single national standard. Nonbinding for now, but signals the direction. (read more)
The FTC issued enforceable guidance on AI washing and algorithmic discrimination under existing authority. No new legislation needed. They can move now, with penalties above $53,000 per violation. (read more)
UK regulators, CMA, FCA, ICO, and Ofcom, jointly stated that businesses are legally responsible for their AI agents' actions toward consumers. No grace period. Effective immediately. (read more)
So what: The EU gave you more time. The FTC did not. The UK removed the grace period entirely. If you operate across jurisdictions, you cannot apply one compliance posture to all three.
4. Your ERP probably already has agentic AI
The major ERP and FP&A platforms shipped production-ready AI agents for finance workflows in March. Most finance teams do not know these features exist in their current licenses.
SAP shipped a Cash Collection Agent, Accounting Accruals Agent, Expense Audit Agent, and Receipt Analysis Agent as generally available features (read more)
Oracle Analytics Cloud launched AI Data Agents that deliver decision-ready answers from Fusion ERP data (read more)
Workday launched a natural language interface via its Sana partnership, covering Workday data and connected systems (read more)
Mastercard launched a Virtual CFO agent for small businesses, drawing on transaction network data (read more)
So what? ERP vendors are building this into the platform, whether you asked for it or not. The question is no longer whether AI is coming to your finance systems. It is whether you know what is already there. It is also the safest and fastest way for finance pros to get access to AI agents - so go ahead and check what your ERP can do now.
5. The research said the quiet part out loud
Three studies published or widely cited in March arrived at the same uncomfortable conclusion.
OpenAI's own research confirmed that hallucinations are mathematically inevitable. Not a fixable bug. A permanent feature of how these models work. Reasoning models hallucinated more, not less, than simpler ones. (read more)
Goldman Sachs found no meaningful economy-wide AI productivity impact. The 30% gains exist in two places only: software coding and customer service. Everything else is flat. (read more)
Gartner surveyed CFOs and found that only 7% report strong AI impact, while nearly 60% are planning to increase AI spending by 10% or more this year. (read more)
So what: The investment is running ahead of the results. The tools are outpacing the governance and the skills. Most CFOs already sense this. The research just confirmed it.
The news tells you what happened in March. Five clusters, ten jurisdictions, dozens of product releases.
But knowing what happened is not the same as knowing what to do with it.
In the paid section, we go through each cluster with a direct CFO action. What to do about Anthropic's pace and its reliability gaps. How to approach the Excel war without defaulting to whatever came bundled in your license. What the regulatory trifecta actually requires from you right now. How to find out what AI is already running inside your ERP. And an honest two-part framework for the hardest question in finance AI right now: why is your spending ahead of your results, and what do you do about that.
Everything in the paid section is something I would tell a client in a working session.
Closing Thoughts
March was a lot. If you are feeling behind, that is not a sign you are doing something wrong. It is a sign that the pace is genuinely accelerating.
The most useful thing I can tell you is this: you do not need to track everything. You need to track the things that affect your function and have a position on each one. That is what this newsletter is for.
See you next Tuesday.
We Want Your Feedback!
This newsletter is for you, and we want to make it as valuable as possible. Please reply to this email with your questions, comments, or topics you'd like to see covered in future issues. Your input shapes our content!
Want to dive deeper into balanced AI adoption for your finance team? Or do you want to hire an AI-powered CFO? Book a consultation!
Did you find this newsletter helpful? Forward it to a colleague who might benefit!
Until next Tuesday, keep balancing!
Anna Tiomina
AI-Powered CFO
1