The Skills Gap Hiding Behind the Technology
Accounting firms are deploying AI at pace. Agentic AI reached a tipping point in January 2026, with some firms automating over 80% of tax prep and cutting audit document analysis time by half. The technology works. The question no one's asking is whether your people can actually develop the skills that matter once AI handles the mechanical work.
Strategic judgment, relationship building, connecting client comments to opportunities. These can't be automated. They also disappear first when people are burned out. Research shows decision quality drops without real recovery. Pattern recognition suffers. Nuance gets missed. Complex situations get squeezed into risky simplifications.
82% of employees are currently at risk of burnout. That's the stat firms should worry about more than which AI platform to license.
The Advisory Training Problem
Most firms have incorporated "become more advisory" into strategic planning. The execution fails because advisory skills require presence and cognitive bandwidth. When people operate in survival mode, they don't have access to higher-level thinking. They can't read room dynamics. They miss the client comment that signals a strategic opportunity.
Firms investing in AI training unlock 40 extra hours per employee annually. Only 25-37% actually do it. Meanwhile, 46% of finance leaders cite generative AI skills as their top gap for 2025-2026. The global AI accounting market hit $10.87bn in 2026, growing at 44.6% annually. Firms with AI strategies are 3.1x more likely to see ROI, but 63% deployed AI without measuring value.
What Actually Needs Training
AI supports rather than replaces jobs. New roles like AI Accounting Analyst are emerging. But digital natives lead uptake while older staff lag. Legacy systems and data issues delay 63% of implementations. The skills that matter now include prompting, basic Python or SQL, and critical thinking. Also: the capacity to think beyond immediate tasks.
The technology will continue advancing. The question is whether you're building an environment where people can actually develop the capabilities AI can't replicate, or whether you're training burned-out staff on new platforms and wondering why the advisory pivot isn't working.