Issue #069 June 10, 2026

AI Costs, Career Pivots.

One thing happening in AI. One career insight. One paper or tool. One hot take.
That's the whole issue.


You're Benchmarking AI Costs Against Software — You Should Be Benchmarking Against Headcount

Boris Cherny, the engineer behind Claude Code, told Fortune that most companies evaluate AI investment by comparing it to other software purchases. That's the wrong frame. The right comparison is what it costs to hire, train, and retain a person to do the same task. For accounting firms, this reframes the entire ROI conversation: a $500/month AI that handles first-pass AP matching isn't competing with a $50/month SaaS tool — it's competing with 20 hours of staff time per week.

Stop asking 'is this AI tool worth the subscription?' Start asking 'what would I pay a competent person to do this, and is the AI cheaper?'

The Accountants Landing Senior Roles Right Now Have One Thing in Common

A software engineer recently went viral for landing a £100K+ UK role without relocating — by reverse-engineering job descriptions and building a targeted skill signal. Accountants can run the same play: the firms paying premiums right now aren't looking for people who 'know AI,' they're looking for people who have automated something real and can describe the before and after in dollar terms. One documented workflow improvement — AP cycle time cut from 12 days to 3, month-end close shortened by two days — is worth more on a resume than a dozen AI certifications.

Pick one process you own, automate part of it this quarter, measure the result, and write it down. That story is your leverage.

Claude Code Is Being Used to Build Internal Finance Automation Without an Engineering Team

The business problem: you know exactly what your close process needs, but you can't afford a developer to build it and your IT queue is six months long. Claude Code — Anthropic's agentic coding assistant — lets someone with process knowledge and minimal scripting background build, test, and iterate on automation pipelines directly. Firms are already using it to write Python scripts that pull from ERP exports, flag anomalies, and format output for review — without a single engineer in the loop.

If you have a controller who's comfortable in Excel and curious about automation, point them at Claude Code before you budget for a developer.

The Billable Hour Isn't Dying — But the Junior Staff Model That Supports It Is

Everyone's focused on whether AI will replace accountants. The more immediate disruption is structural: the traditional pyramid — partners billing high, juniors billing low to cover overhead — breaks when AI does the junior work faster and cheaper. A mid-sized firm in Chicago already cut its first-year associate class by 40% this hiring cycle and absorbed the volume with agentic AP and reconciliation tools. The partners didn't lose revenue. They lost the training pipeline that used to produce their next senior staff.

AI isn't flattening your fees — it's flattening your bench, and you won't feel the talent shortage until it's two years too late to fix.