AIR Academy · The open programme · Practice decision · About 7 minutes

The practice decision: the reporting tool

This is a practice decision, not an assessment. You read a short situation, choose what to do at each of three decision points, and write one sentence.

Before you start

A practice decision, not an assessment

  • This is a practice decision, not an assessment. You read a short situation, choose what to do at each of three decision points, and write one sentence.
  • At the end you see a development reading: what you looked at outside the AI, and what you did not. Only you see it.
  • Nothing you choose or type leaves your device. Every company, person, tool and figure is fictional.

The situation

Monday

You manage the office team at Meridian Supplies, a fictional distributor of packaging and consumables with 38 staff. Five of your people prepare the company's recurring reports.

KessidAI is the company's AI assistant. It is connected to the accounting and stock systems. The owner asked it whether to buy a reporting tool.

On Monday the owner forwards you KessidAI's recommendation with one line: "Your view by Thursday. Shall we sign?"

You see the recommendation as KessidAI wrote it.

KessidAIFictional case · training material

Approve the ReportFlow AI subscription.

Your five office staff spend about 200 hours a month preparing recurring reports. ReportFlow AI can automate an estimated 60% of that work.

Hours a month on recurring reports200
Share automatable (vendor estimate)60%
Hours released a month (potential)120
Hourly labour value (internal costing rate)USD 30
Potential annual capacity valueUSD 43,200
Subscription, annual commitment, paid in advanceUSD 12,000
Potential net benefit a yearUSD 31,200

KessidAI's conclusion: the potential capacity value is more than three times the subscription, a potential net benefit of USD 31,200 a year. Recommend approval this week so that the first automated reports run next month.

Decision point 1 of 3 · Structure

Where do you start?

What you looked at

The arithmetic is right: 200 hours, 60%, 120 hours, USD 30 an hour, USD 43,200 a year against USD 12,000.

The sum is right. Nobody has asked what the decision is for. KessidAI answered one question: is 43,200 larger than 12,000?

KessidAI replies

"My recommendation assumes three things. The vendor's 60% applies to all of your recurring reports. Every released hour is worth USD 30 to Meridian. Set-up falls on the vendor. Tell me if any of these is wrong."

A sound first step: the assumptions are now on the table. Nothing outside KessidAI has checked any of them.

The owner replies

"I want fewer hours on reports only if those hours go somewhere useful, and I am not sure every report we produce is still read."

"Two things are fixed. The Monday sales and stock report cannot be late. And I will not pay a year in advance for something we have not seen run on our own reports."

You now know the decision Meridian has: what the released hours would be used for, and which reports are needed at all. That is a different question from the one KessidAI answered.

Decision point 2 of 3 · Test

What do you do with the figures?

What you looked at

Nothing new. The note is as it was.

Both figures have a source line. Neither source has been opened.

KessidAI replies

"It is the vendor's estimate from a demonstration. Tools of this kind commonly report 50% to 70%. Treat it as a planning figure."

You asked the AI to check its own figure. It answered from what it already had, and added a range from nowhere in particular.

Two answers

The finance manager: "200 hours is last quarter's timesheets, five people, every recurring report. I have never counted which reports are opened. Give me two weeks and I will."

The vendor, asked to run Meridian's own weekly sales and stock report: "Happy to. The 60% came from a 30-minute demonstration on one sample report."

The claim the decision rests on is that 200 hours of needed work exists and 60% of it can go to the tool. You now know the 60% rests on one sample, and that nobody has counted which reports are read. You know what would prove the recommendation wrong, and who is counting.

Decision point 3 of 3 · Engage

It is Wednesday evening and your note is ready. What do you do before you send it?

What you looked at

Nothing new. The note is as it was.

The owner has it. The five people who prepare the reports and the three who read the Monday report were not asked. The owner does not know KessidAI drafted the first version. Nobody has said who approves USD 12,000.

The finance manager replies, on Friday

"Noted. I will read it next week."

You told the person who holds the costing rate, and asked no question, so nothing came back before the owner read it.

Three answers

The sales managers: "We cannot run the week without the Monday report. It has to be right by eight on Monday. A wrong figure costs us a day of calls."

One of your staff: "Two of us already use a free AI assistant on our own accounts for the formatting. Nobody has said which company data may go into a tool."

Finance: "Anything above USD 10,000 a year is the owner's signature, not mine."

You hold three things the figures never showed: what the Monday report must do, who signs, and a data question that is the owner's, not the tool's. Your note says KessidAI drafted the first version, so the owner reads it knowing where it came from.

Perform

Thursday

Write the one sentence you would say to the owner on Thursday: what you will commit to, on what evidence, and what would change your mind.

This sentence stays on your device. It is not collected.

In the session, people write a decision in six lines: the action, the reason, the owner, the evidence it rests on, the condition that would change it, and the date to check. Approve with conditions, run a pilot with what it must show, and do not approve now are all decisions when they carry those six lines.

Development reading

What you looked at outside KessidAI

One way to say it to the owner: "I recommend we do not sign this week. First, a two-week count of which reports are read, and one run of our own Monday report through the tool, with the vendor. If the count shows the load stays near 200 hours and the run holds, I will bring back a pilot on two reports, with what it must show. I would change this if the vendor offers a cancellation window and the count comes back quickly."

Compare yours: what will I commit to, on what basis, and what would change my mind?

Both endings are composite fiction, written for this programme.

The case worked through in the room, with three defensible decisions, is on the toolkit page. The case worked through · The five practice exercises · Back to the lessons