Since generative AI entered mainstream conversation, disruption has dominated the headlines. Will automation eliminate accounting jobs? Will finance teams get leaner, smaller and further removed from the strategic table?
What I'm actually seeing inside accounting organizations (including my own) tells a different story. The common assumption is that closing the AI skills gap means teaching accountants new technical skills: prompt engineering, tool fluency and data literacy. That's not the gap I'm seeing. The gap is structural: whether a firm has redesigned its governance, roles and workflows around AI, not whether its people know how to use the tools.
For many of us, AI has stopped being an isolated pilot project or a line item in our software budgets. It's become infrastructure — the kind of investment finance leaders are now expected to justify with real ROI, disciplined oversight and long-term organizational impact. That shift changes what "AI-ready" actually means for a firm or in-house department.
That's a genuinely different management problem than the one most firm leaders were trained to solve. AI initiatives now need explicit governance frameworks and measurable outcomes tied to the firm's broader operating model, not just a faster way to key in invoices. This is the structural work; no amount of individual tool training closes an accountability gap.
The biggest misread on AI in finance is that its purpose is headcount reduction. In the organizations I've seen with the strongest results, it's the opposite: AI is being used to elevate the people already on the team, not replace them. Tipalti's
That's where accounting leadership matters most. If a firm simply automates manual work without redesigning how roles and workflows are structured around that automation, it misses the larger opportunity. The real value doesn't come from staff learning new technical skills; it comes from firms deliberately redirecting time toward the disciplines that build client trust and firm value: analysis, forecasting, advisory conversations, and the judgment calls no model can make. That redirection is a structural decision, not a training outcome.
For most of my career, accounting was a historical reporting function: close the books, explain what already happened. That model is eroding, not because forecasting is new, but because continuous forecasting used to require a full-time analyst rebuilding the model by hand. As that labor gets automated, forecasting stops being a periodic exercise and becomes a standing capability.
For firms serving clients, this is a genuine service-line opportunity. Firms that can offer continuously updated, forward-looking insight rather than periodic historical reporting are positioned to become a different kind of advisor to the businesses they serve. But that shift isn't something a training program hands you. It requires firms to redesign how the function itself is structured and staffed.
None of this means accountants are handing over control. In the rollouts I've been part of, finance wants the ability to review what AI is doing and to see proactive recommendations, not hand full autonomy over to a model.
That distinction should shape how firms build AI-enabled workflows. AI flags the anomaly and surfaces the risk. A human still makes the final call, only faster and better-informed. Getting that right isn't a matter of training staff to "supervise AI better." It's a matter of designing the workflow, escalation path and review checkpoints correctly from the start: structural decisions, not individual competencies.
The takeaway I'd offer other CAOs and firm leaders isn't "AI won't replace you." It's more specific: The real AI skills gap isn't a lack of technical fluency on your team. It's the absence of deliberate structure around how work flows between people and systems.
That means doing three concrete things. First, define which decisions AI can make on its own versus which ones require a human sign-off, and put that in writing, not just in practice. Second, audit where your staff's time actually goes today, and identify the roles where automation should free people up for analysis and client-facing work, not just faster data entry. Third, build a review checkpoint into every AI-enabled workflow, so someone is accountable for catching the exception the model missed. Get that right, and AI doesn't shrink the function. It expands what your team can deliver, while keeping the accountability and control that make accounting a trusted profession in the first place.
