Writing

A Tiny Eval Harness for Finance Claim Drift

The Drift Test lab is seven synthetic claim/fact cases—not an LLM benchmark—built to practice spotting value, period, unit, and definition mismatches before they enter a decision pack.

Walking the Variance Workbench: What to Inspect in 30 Minutes

A hiring-screen walkthrough of the plan-to-actual variance prototype: zero-plan handling, source references, review states, and deliberate limits.

What a LatAm SAP Cutover Taught Me About Finance Systems

Experience-labeled lessons from leading regional finance teams through an SAP implementation across Latin America: ownership, cutover readiness, and controls that match how teams close.

Why the Statement Workbench Is Static on Purpose

A Cloudflare Pages–compatible finance review prototype stays browser-only: deterministic TypeScript, synthetic fixtures, and no model calls—so the control boundary stays inspectable.

Small AI Errors Can Drift Into Important Decisions

A plausible AI error can be copied forward through reports and models. Finance controls need ways to detect small, consistent drift before it compounds.

Embedded AI, Point Solution, or Neither?

A practical framework for deciding how to govern AI features in existing software, dedicated tools, and internal builds.

A Durable Skill for New FP&A Analysts

Why checking whether an answer makes sense is a more durable finance skill than fluency with any one prompting technique.

Data Lineage Is a Practical AI Bottleneck

AI-generated finance answers inherit ambiguity in source data, metric definitions, and ownership. Data lineage makes those dependencies reviewable.

How AI Changes the FP&A Learning Path

The scariest part of agentic AI in finance isn't that it might replace junior analysts. It's that it might quietly remove the exact repetitive work that used to be how people became good analysts in the first place.

The Continuous Close: Start with Reconciliation

The vision of a fully automated, real-time close is oversold. But a genuinely useful, narrower version of it is quietly becoming real for teams that get the fundamentals right. The difference is in what you don't automate.

How to Measure AI ROI in Finance Pilots

Set a baseline, track workflow adoption, and separate time saved from quality improvements when evaluating an AI pilot.

Adapting Model Risk Policy for Language Models

Most finance orgs still govern AI tools using a model risk framework built for regression models and rate curves. That framework doesn't ask the right questions about an LLM. Here's what to add, not throw out.

AI Agents in FP&A: Where Judgment Still Matters

Agents can connect finance workflow steps, but teams still need RACI, approval gates, and reversible outputs before automation earns trust.

What CFOs Are Prioritizing in 2026: AI, Digital Finance, and AI Agents

Key findings from Deloitte’s Q4 2025 CFO Signals survey—why technology transformation, automation, and AI agents top the agenda for North American finance leaders.