The finance systems & AI brief for controllers and CFOs

TOGETHER WITH CFOLAYER

Bespoke consulting for finance tech stack strategy, selection, & implementation

Get a personalized roadmap for your finance tech stack. CFOLAYER partners with CFOs and controllers to design, select, and implement the right systems for your scale and business model - eliminating vendor confusion and slow time-to-value.

Mercury launches Command, an AI layer for running your business banking (3 minute read)
Mercury launched Command, an AI assistant built into its banking platform that lets customers check their cash position, change auto-transfer rules, categorize transactions, and create and send invoices from a single natural-language chat, with every action requiring explicit confirmation before it runs. What this means for your stack: if your company banks with Mercury, Command could absorb routine treasury and bookkeeping clicks, but because it acts on live account data and can move money, confirm the approval controls and per-user permissions before anyone uses it for payments.

CFOs keep raising tech spending despite economic worry (2 minute read)
A Grant Thornton survey found finance chiefs are still increasing technology budgets even amid economic uncertainty, with AI the main driver of the new spending.

BCG says vibe coding is coming to finance, and to set guardrails first (3 minute read)
BCG predicts that "vibe coding" (building software by describing what you want to an AI rather than writing code) will spread into finance teams, letting them spin up their own forecasting, anomaly-detection, and document-review apps, but warns it can create "shadow code," security holes, and audit gaps without governance. What this means for your stack: set guardrails before analysts start shipping AI-built tools, keep these apps sitting on top of your core accounting and planning systems rather than replacing them, and require documentation and review so you do not trade shadow spreadsheets for shadow code.

AI is eroding trust, and finance teams can rebuild it (4 minute read)
This opinion piece argues that finance and accounting teams are well placed to restore confidence in AI-influenced numbers by pairing AI tools with human judgment, stronger oversight, and practical upskilling rather than blind adoption.

Most midsized companies now use AI for FP&A (2 minute read)
A Vena Solutions survey found a majority of midsized companies now use AI in financial planning and analysis (FP&A), in many cases because the C-suite mandated that finance apply it to internal processes.

Companies are saving cash with AI, just less than they hoped (3 minute read)
Firms report real cost savings from AI but below what they projected, and the gap between expected and actual return could widen as they spend more on complex new tools.

FASB clarifies what companies must disclose in quarterly reports (3 minute read)
The FASB (the US accounting rule-maker) issued guidance (ASU 2025-11, an accounting standards update) that standardizes interim (quarterly) disclosure requirements, adding a principle to help identify which events need disclosure and clarifying the form and content of those disclosures for both public and private companies. What this means for your stack: review your quarterly close and disclosure checklists against the new principle so your interim reporting templates capture the events it now flags.

Using AI agents for QA, and what it means for testing finance implementations (8 minute read)
Antirez (the creator of Redis) argues that large language models open a strictly more powerful way to automate quality assurance, not just to generate code. The approach: give an AI agent a markdown file with structured test instructions, environment context, and a list of QA activities, then let it self-direct the pass, including inspecting recent changes first to find what is most likely to have regressed.

How this applies to finance system implementations and regression testing. The analogy holds cleanly in several areas:

Integration regression testing. After a Workday or NetSuite patch or config change, you are doing essentially what antirez describes: checking that downstream outputs (GL posting, intercompany eliminations, consolidation roll-ups) still behave correctly. An agent given a sandbox and a structured test plan could run those checks without a human clicking through screens.

UAT coverage gaps. Covering every line of code does not mean covering every possible state. Finance user acceptance testing (UAT) has the same problem: teams test the happy paths, miss edge cases (multi-currency transactions, period-boundary postings, subledger-to-GL timing), and ship anyway because the go-live date is fixed.

Regression on config changes. The specific technique of asking the agent to first inspect what changed, then prioritize testing around it, applies directly. Push a new Adaptive Planning model version or modify a NetSuite workflow rule, and an agent can diff the config, identify downstream dependencies, and focus its test pass there.

"Psychological QA." The approach can also move into the softer side of quality: asking the agent to flag things that look surprising, undocumented, or sloppy from the user's point of view. For finance systems this is the "does this make sense to a controller" layer, testing that report labels are correct, that rounding behaves as expected, and that error messages are actionable. That review almost never happens in a structured way.

Where it is harder in finance. The friction points are sandbox fidelity and data sensitivity. Antirez controls his open-source software end to end. Finance sandboxes are often stale or not representative of production configs, and running an agent against real transaction data creates compliance exposure. You would need clean test data sets and a well-maintained refresh cadence, which most organizations do not have. The more immediate win is probably using this pattern to generate and maintain the test plan itself, structured markdown checklists a human or agent can execute, rather than full autonomous test execution. That is a lower-risk entry point with immediate return.

Apptio expands its cloud-spend tracking tool (2 min read) — a FinOps update aimed at giving finance better visibility into cloud costs.

UK Treasury wavers on a £1.7B ERP program it agreed to fund (3 min read) — a cautionary tale on large public-sector ERP governance.

What CFOs would (and would not) trust an agentic AI clone to do (2 min read) — finance leaders weigh which tasks an AI version of themselves could handle.

Worth a second look

Ramp launches an AI system for the accounting close and raises $750M — Ramp's move into close, reconciliation, and journal entries puts it head to head with the close-software vendors.