Skip to content

Your controller found a faster way to close the month.#

SEO phrases: AI ROI, change management, month-end close, human approval, controller-faster-close-month.

An office at dusk. One person at a desk works under a glowing "Human Approval" frame while a thread of light runs from their screen to a team at a conference table. Caption: One person's way becomes the team's. Then keep a human approving what goes out.

Your controller found a faster way to close the month.
Sales has a prompt that cleans up follow-up notes.
Operations has a spreadsheet someone finally automated.

Everybody feels faster.
The company's numbers don't move.

That's the AI ROI problem I keep seeing. Not whether the model works. Whether one person's better way of working ever becomes the company's way of working.

Here's what I keep running into, and the data backs it. Almost everyone feels the gain. Almost no one can point to it on the bottom line. In McKinsey's 2026 survey, 80% of respondents say AI improved their own productivity. Only 37% report positive EBIT contribution from AI, essentially flat from 2025.

The failures point the same way. By the end of 2025, at least half of generative AI projects were abandoned after proof of concept, and "unclear business value" was one reason (Gartner, January 2026). Its warning is blunt: "Without change management, even technically excellent GenAI tools see minimal adoption."

And almost nobody is changing how the work runs. Only 30% of organizations are redesigning key processes around AI. Another 37% report using it only at a surface level, with little or no change to underlying business processes (Deloitte, January 2026).

These are enterprise studies, but I see the same split in a 30-person shop: the tool works for one person, and the business still can't point to the number.

A saved hour isn't earnings until it changes the work. It has to remove a handoff, shorten a close, cancel a task, reduce rework, or give the owner a number to manage.

My read: getting past the trap starts with three boring things.

  1. Start with one team and one baseline. Hours per month-end close. Days from request to approval. Follow-ups sent before a lead cools.
  2. Turn one person's better way into the team's way. When the controller works out the month-end check, the whole finance team should be able to run it.
  3. Keep a person approving what goes out. Then automate the repetitive part after the judgment is clear.

Inside JOV AI it works the same way. Agents prepare the work, a person approves what goes out, and each workflow gets an owner and a number.

We haven't booked our own company-wide AI number yet. I'd rather say that plainly than make one up. What we have is the operating loop: source check, draft, evaluation, human approval, published record. The baselines come next, one workflow at a time.

Feeling faster happens on its own.
Booking it doesn't.

Which team of yours is using AI without a number yet? Tell me the team and the one metric you wish moved, here or at dminix@jov.ai, and I'll tell you where I'd put the baseline first.

Sources#