Waymo revealed this week that 70 remote assistance agents oversee its entire fleet of 3,000+ autonomous vehicles. That's one human for every 43 cars on the road. The vehicles drive themselves. When they hit something ambiguous — a confusing intersection, an unusual obstacle — they phone home. A human advises. The car decides. Most of the time, the vehicle resolves the situation before the agent even responds.
That ratio — 1:43 — is what an AI-native operating model actually looks like at scale. Not theoretical. Not projected. Running, today, on public roads.
Now look at your practice. One bookkeeper handling 8 to 10 client accounts. One senior reviewing the work of maybe 3 or 4 staff. Those ratios haven't changed meaningfully in a decade. The tools got faster, the software got better, but the number of humans per unit of output barely moved.
If you're running a CAS practice, the Waymo number isn't a tech story. It's a staffing story. And it's telling you something specific about what your firm looks like in 18 months.
The ratio is the strategy
Waymo's model works because driving — the repetitive, pattern-based, high-volume execution — is handled by software. Humans handle exceptions. The 70 agents aren't monitoring every vehicle in real time. They're consulting. They step in when the system flags something it can't resolve, then step back out.
Replace "driving" with "transaction coding." Or bank reconciliations. Or monthly close prep. Or routine client reporting. The structural parallel is exact. Your practice has an execution layer — repeatable, pattern-based work that follows documented rules — and a judgment layer where experience, context, and client knowledge actually matter.
Right now, most firms staff both layers with the same people. That's about to look as inefficient as putting a human driver in every Waymo.
And the restructuring proof already exists outside our profession. Klarna went from 7,000 employees to 3,000 over three years while growing revenue. AI now handles work equivalent to 853 full-time customer service staff. Revenue per employee approaching $1 million, up from $575,000. The work didn't disappear. The ratio changed.
Fewer people, different jobs
I wrote recently about The Talent Trap — what happens to junior development when AI handles the foundational work that used to train new accountants. This is the structural sequel. Not how you develop people, but how many you need and what they actually do.
The answer Waymo demonstrates — and Klarna confirms — is fewer people doing fundamentally different work. The human role shifts from executing to supervising. From producing deliverables to reviewing AI output, handling exceptions, and making the judgment calls that require client context.
Your bookkeepers don't disappear in this model. They become agent monitors — overseeing AI workflows across 20, 30, 40 client accounts instead of manually processing 8 to 10. Your seniors don't do less. They do different — reviewing, approving, catching the things AI flags and the things it misses.
The firm that figures this out first doesn't just move faster. It operates at a structurally different ratio — and either captures that margin or reprices to win clients from firms that haven't restructured.
This is already happening — even in my practice
I'll be direct about where I am on this. When we first started deploying AI workflows, it actually created more work. Experimentation has a cost — building prompts, testing outputs, fixing what didn't work, training the team on new processes. That's real, and anyone who tells you AI is plug-and-play on day one is selling something.
But after it stabilized, the shift became undeniable. We're now routinely looking for inefficiencies that can be eliminated — problems that AI can help us solve — and redesigning the process at the same time. It's nowhere near finished, but it's underway. A team member recently left, and we're not replacing them. The efficiencies we've built are expected to absorb the work. That's not a one-time cost saving. That's a permanently different operating ratio.
I'm not at 1:43. Nobody in CAS is — yet. But the direction is clear, and the distance between where most firms are and where this is heading is larger than most practice owners realize. Tomorrow, we're publishing a piece on how the cost of digital agents just collapsed — and what that means for the math underneath every staffing decision you're making right now.
The question you can't defer
Waymo didn't get to 1:43 by adding remote agents to a human-driven fleet. They rebuilt the entire operating model around what autonomy made possible. Firms that bolt AI onto their current staffing model — same roles, same ratios, same workflows, just slightly faster — will capture a fraction of what's available.
The real question isn't whether you can use AI to make your team 10% more efficient. It's what your firm looks like when you design it around a fundamentally different human-to-client ratio. Different headcount. Different roles. Different economics.
If you don't have someone inside your firm thinking through that redesign right now, you're already behind. Our 15-Day AI Practice Transformation program is built for exactly this — a CAS owner and their champion mapping the practice around what AI makes possible. Three phases. Fifteen working days. A firm designed for what's coming, not optimized for what's leaving. Visit theaiaccountant.ai to apply.
Your practice ratio is the scoreboard. Start moving the number.