AI & Automation

Thirty percent of accounting professionals name manual data cleanup as the top barrier to advisory work, ahead of staffing. The average firm runs ten applications and loses five hours a week to re-entry.

Professional services leads on AI adoption, but only about a quarter deploy firm-wide. The difference is workflow redesign, not tools. What the leading quartile did differently.

More than 40% of agentic AI projects are forecast to be canceled by 2027, and escalating cost leads the causes. Why open scope, not capability, is what ends most agentic builds.

Agentic AI pilots in lending stall at one workflow because no one can produce a record of what the agent did. Why observability, not model quality, is the constraint on scaling.

Firms protect client data with sophisticated rigor and classify their own process knowledge not at all. Adoption is compounding at 20.2% of firms, 36.8% in professional services. The method gets encoded either way.

The largest US banks put AI agents in client vetting and transaction accounting, not the storefront. Three tests decide where a lender's first agent actually earns its keep.

Three-quarters of small professional services firms use AI. A third saw revenue move. The gap is not the tools, it is the unrebuilt operation the tools run inside.

Keeping a human in the loop by reviewing every file rebuilds the bottleneck agents remove. The 2026 operating question for lenders is where judgment should sit, not whether a person is involved.