workflow first
Why your AI pilot did not change the work
The tool was never the constraint. The handoffs, the exceptions, and the missing evidence were.
You bought the tools. People were trained. Six months later the work looks the same.
That is the most common outcome, and it is not a tooling failure. A workflow is a chain of decisions, handoffs, exceptions, and evidence. A tool improves one step of it. If the step was never the constraint, improving it changes nothing downstream.
Where the time actually goes
Ask where a request waits, and the answer is almost never "in the step a model could do." It waits at a handoff nobody owns, at an exception with no defined path, and at a decision that cannot be made because two systems disagree about the same number.
What to do instead
Start from the outcome and work backwards through the chain. For each step, decide whether it should be human, deterministic software, AI-assisted, or an external service, and what evidence is required before anything acts.
Most steps come back deterministic. That is a good result, not a disappointing one.
Start with one workflow.
Thirty minutes, no deck. Bring a process that costs more than it should and we will tell you honestly whether there is something here worth paying for.