There's a case study circulating on LinkedIn at the moment that every accountant should read.
Nathan Carroll, a founder and CEO, recently published a piece on the Xero and Anthropic partnership and what it signals for the finance industry. Craig Carroll, a founder and investor, responded with his own account from a private session he attended with a group of founders and their CFOs.
The session involved a detailed review of a business that had already cut its operational finance costs by 40% through the application of AI. No redundancies. No single new software platform. The finance function had been completely rebuilt from scratch.
The method is the part that matters
The 40% gets attention. But how they got there is what every practice should be paying attention to.
The team mapped every workflow end to end, identified decision points, and reallocated tasks across AI agents, automation tools, and humans. Repeatable work went to AI agents. Manual handoffs were automated. Humans moved into judgement calls, exceptions, and client communication.
Craig is clear that this wasn't theoretical. It wasn't a roadmap or a future plan. It was already done.
Most practices right now are asking which AI tools to adopt and which platforms to evaluate. The businesses seeing results like this have moved past that. They're asking whether their current system is worth keeping at all. That's a different question, and it leads somewhere different.
The jobs going first are ones every practice will recognise
Craig makes a pointed observation about where the impact is landing first. Offshore finance teams.
In the business he reviewed, a meaningful portion of remote finance roles had already been displaced. Not gradually, but decisively. The structured, repeatable work those roles were built around had been automated out of the system entirely.
For accounting and bookkeeping practices, that list of affected tasks will be immediately familiar. Data entry, reconciliations, document handling, standard reporting. For many firms, that's a significant share of how time is spent and how staff are deployed.
His view is that the market is still underestimating this. The Xero and Anthropic partnership is the most visible wave, but it isn't the only one. The shift is broader than any single platform announcement.
The part the industry is carefully avoiding
Most commentary on AI in finance steps around something that deserves to be said directly.
Roles are disappearing. In some businesses, displacement is already happening faster than new work is being created. Craig is direct about this: the finance industry is entering a period of incredible efficiency and innovation, but also one where parts of the finance workforce may be structurally displaced faster than new roles are created.
For practices with staff whose work sits in those categories, the useful question is which parts of the current model are genuinely protected and which are more exposed than they look. Waiting for that to become obvious is leaving it too late.
The tool conversation is a distraction
Nathan's original piece makes a point that Craig builds on. The discussion across the profession is dominated by platform announcements, software integrations, and AI feature releases. The Xero and Anthropic partnership is a prominent example.
Both Carrolls argue this is the wrong layer to be focused on. Finance isn't becoming a better-tooled version of what it was. It's becoming, as Craig puts it, a system of orchestrated agents rather than a function delivered by people or point solutions.
The firms that have understood that are already operating differently. The gap between them and the firms still evaluating tools is opening, and it will get harder to close.
Where the work that matters actually sits
The firms navigating this well are deploying people more deliberately, with a clearer sense of where human involvement produces something a system cannot.
Judgement, client relationships, complex decisions, communication that requires context and trust. For accountants and bookkeepers, that's where the most defensible value has always lived. As automation absorbs the repeatable work, that value becomes more visible and more important to protect.
Practices that are honest about which parts of their model are built on expertise and which are built on process are in a much stronger position to make deliberate decisions about where to go next.
The question worth sitting with
Craig ends his response with something that applies directly to every practice in this profession.
What happens to the humans in the system?
That question deserves more airtime than it's currently getting. The practices willing to ask it openly, and to act on the answer, are the ones that will still be shaping this profession in five years.
Read Nathan Carroll's original article and Craig Carroll's response on LinkedIn here.
