The finance systems & AI brief for controllers and CFOs

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.

🏛️ Systems & Stack

A Deloitte report finds software vendors increasingly charging companies extra to access their own data as AI use grows, a practice called data tollgating that can blindside budgets built before these fees existed. Deloitte recommends auditing data rights before contract renewals rather than waiting until negotiations begin. What this means for your stack: review your SaaS (software delivered over the internet) contracts now for data access clauses before locking in new AI tools, since these rights are hard to win back once surrendered.

Square rolled out a ChatGPT app and Claude plugin that let customers discover restaurants and place orders directly through AI assistants, with no setup cost or added commission for eligible sellers. Amazon's Alexa+ is coming next as Square expands AI-driven discovery across more channels. What this means for your stack: if you run a Square-based business, check your dashboard for AI discoverability settings, since enrollment appears to be automatic rather than opt-in.

Taxwire raised $25 million led by Headline for an AI system that takes over a company's sales tax registration, filing, and payment across more than 100 countries. The platform builds an auditable trail for each filing, aiming to cut both administrative work and audit risk. What this means for your stack: if sales tax compliance eats real finance team time, tools like this point toward outsourcing the mechanics while keeping an audit trail your team can still review.

Quantifind, which sells AI tools that help banks spot financial crime, raised $200 million led by Summit Partners to expand internationally and grow its Graphyte platform. An independent analysis estimated the platform could cut alert-processing costs at large banks by up to $177.9 million by reducing false positives.

🤖 AI in Finance

Cursor, the AI coding company SpaceX has agreed to buy for $60 billion, is starting a CFO council to build shared benchmarks for measuring AI return on investment. Asana and SentinelOne's finance chiefs are among the first members, with the group's first meeting set for August. What this means for your stack: as AI spend gets harder to track under token-based pricing, watch this group for benchmarks you can borrow instead of building an ROI framework from scratch.

⚖️ Regulation & Reporting

The PCAOB (the board that oversees public company audits) launched a process letting registered accounting firms submit questions and get informal staff views on how to apply its auditing and quality control standards. Firms wanting a more formal review can request a full firm consultation through the board's new intake form.

The FASB (the board that sets US accounting rules) proposed a rule change requiring investment funds to factor in contract-based sale restrictions when valuing restricted equity, after stakeholders said current guidance can overstate net asset value and distort performance reporting. Comments on the proposed ASU (an accounting rule update) are due July 17.

The IRS (the US tax agency) added an online option for individuals to file Form 843, a claim for a refund tied to the Kwong court decision, which lets certain COVID-era periods be excluded when calculating filing deadlines. The online option covers individuals only; businesses must still file on paper, and a Friday deadline applies.

🛠️ The Practitioner

A Journal of Accountancy feature explains how easy access to generative AI is helping fraudsters create convincing fake vendors, invoices, and supporting paperwork, making accounts payable and accounts receivable a growing target. A forensics expert recommends tightening vendor setup reviews and keeping cash-handling duties split between at least two people. What this means for your stack: audit your accounts payable vendor master file for red flags like PO box addresses or shared bank accounts, since AI has lowered the bar for creating a convincing fake vendor.

Quick Links

LinqAlpha raises $22M for AI hedge fund research tools (1 min read) — Ex-Goldman founders' AI platform for hedge fund research now serves 70+ institutions.

🔁 ICYMI

Worth a second look

Secure MCP Tunnel | OpenAI API — OpenAI's technical guide for connecting private/on-prem MCP (a standard that lets AI models call outside tools) servers to its products without exposing them to the public internet, a useful reference as finance teams wire internal systems into AI agents.

Why governance clearinghouses may win the AI agent era — Argues the durable moat in the AI era isn't the system of record but the governance clearinghouse controlling what agents can see, do, and touch, worth revisiting as your team weighs which AI agent platform to standardize on.

Deloitte guide covers how to book revenue for usage-priced AI software — Deloitte's revenue recognition guidance for outcome- and usage-priced AI software remains the clearest reference as more vendors shift away from flat subscriptions, useful alongside this week's story on SaaS data tollgating.

Digits launches an always-on AI system that runs the close continuously — Digits' always-on close system is worth a second look now that continuous-close claims keep surfacing across the market, a good benchmark for what "always-on" actually means in practice.

📚 Definitions

Related to this weeks stories and more

MCP (Model Context Protocol): An open standard that lets AI models and agents connect to outside tools, data sources, and systems in a consistent way, instead of each integration being custom-built. Think of it as a common plug shape that lets an AI assistant safely call into your calendar, your database, or your finance system.

API (Application Programming Interface): A defined set of rules that lets one piece of software request data or actions from another without either side needing to know how the other is built internally. It's the connective tissue behind most "AI in your existing tools" rollouts - when OpenAI or a vendor says something "integrates via API," this is what they mean.

Token: The basic unit AI language models use to measure and price the text they read or generate, roughly three-quarters of a word in English. Usage-based AI pricing is typically billed per token, which is why token consumption is becoming a line item finance teams have to forecast.

Agentic AI: AI systems built to take multi-step actions on a user's behalf - placing an order, running a close, executing a workflow - rather than just answering questions or generating text. This issue's Square/ChatGPT-Claude integration and the ICYMI item on Digits' always-on close are both examples.

GPU (Graphics Processing Unit): A chip originally built to render images and video quickly, now the workhorse for training and running AI models because it can perform many calculations in parallel. GPU capacity, and its cost, is a major line item behind the AI token pricing finance teams are increasingly budgeting for.

CPU (Central Processing Unit): The general-purpose "brain" chip in a computer or server that handles most everyday computing tasks in sequence, one after another. Traditional business software (your ERP, spreadsheets, email) runs mainly on CPUs; AI model training and inference lean much more heavily on GPUs.

Keep Reading