Sales Are Up 20%, Profit Is Down: What Should Your AI Tell You?

Sales Are Up 20%. Profit Is Down. What Should Your AI Tell You to Do?

Sales up 20%. Gross profit down ₹60,000.

Both numbers can be correct. Both can belong to the same month.

Imagine opening that report on a Monday morning. Your sales manager wants to push harder. Your accountant wants to protect margins. Your purchase manager wants to buy more stock before prices rise.

Three sensible people. Three different directions.

Now you ask AI to summarise the report. It produces a beautifully written explanation of all three positions.

You still have to decide what to do at 11 AM.

That gap, between knowing what happened and choosing what happens next, is where AI for business owners needs to get much more useful.

My work spans business consulting, AI workshops and building business software. Across those settings, this is the standard I want to hold AI to: help an owner make a better choice with the time and information available.

Here is a practical way to do that.

The report can be right and the decision can still be wrong

Let us work through an illustrative distributor. These are teaching numbers, not a client result. Sales are net of GST and returns, and both periods use the same accounting basis.

Last month, sales were ₹50 lakh and the gross margin was 18%. Gross profit: ₹9 lakh.

This month, sales are ₹60 lakh and the gross margin is 14%. Gross profit: ₹8.4 lakh.

Sales grew by ₹10 lakh. Gross profit fell by ₹60,000, before operating expenses.

A sales report celebrates growth. A margin report flags deterioration. Neither tells you the cause.

Perhaps the team sold more of a lower-margin product. Perhaps discounts increased. Perhaps purchase costs rose and selling prices stayed the same. Perhaps a return or cost adjustment distorted the comparison.

Each explanation calls for a different response.

An across-the-board price increase might damage a healthy product line while leaving the actual problem untouched.

This is why the next question matters so much: what would we need to know before choosing an action?

AI can help compare explanations, organise supporting records and prepare options. Calculations should come from checked spreadsheet formulas or the accounting system. A convincing paragraph cannot establish which explanation is true.

Use a Decision Card before the next management meeting

In earlier articles, I wrote about the Four Registers and the One Rupee Test: trustworthy inputs and evidence of useful work.

The next step is choosing between useful actions. I suggest a one-page Decision Card with five entries.

1. The decision

Write a choice you can make.

“Margin has fallen” is an observation.

“Should we change the discount on product B for new quotations next week?” is a decision.

The second sentence gives your team something to investigate. It also keeps the meeting from becoming a tour of every number in the business.

2. The evidence and the missing evidence

Specify the company, period and source behind the numbers. Compare like with like: full months with full months, matching product groups, consistent treatment of returns.

Then write what you do not know.

For our distributor, that might be: “We have invoice-level sales and discounts. We have not reconciled landed product costs for this month.”

That missing line changes the decision. You can investigate discounts now. You cannot confidently declare which customer is unprofitable yet.

An AI system that makes this gap visible is helping you think.

3. The choices and their trade-offs

Ask for two practical options and a “wait while we check” option.

For example: restrict discretionary discounts on selected new quotations; investigate whether product mix explains the margin fall; or hold the current policy briefly while checking the cost records.

For each option, ask what it could improve, what it could damage, and what assumption must hold.

Changing a discount may protect margin and reduce conversion. Changing product emphasis may improve the sales mix and leave slow stock sitting longer. Waiting protects you from a premature decision but has a cost if the problem continues.

The trade-off belongs on the same page as the recommendation.

4. The person and the deadline

“Sales team to take action” leaves too much room for everyone to wait.

“Sales manager to bring ten recent discounted quotations and their outcomes by Tuesday at noon” is usable.

If a decision needs your approval, say so. If a manager can make it within an agreed limit, write that limit down.

Otherwise the AI can prepare excellent work that spends three days waiting for someone to say yes.

5. The review

Before acting, choose when you will look again and what you will compare.

For a limited quotation change, review accepted prices, conversion and gross profit on the affected product. Watch for changes in the type of inquiry too. Ten unusually strong inquiries can make a weak policy look good.

A short trial gives you evidence to discuss. It rarely proves that AI caused the entire result.

The Decision Card is complete when you know the choice, its basis, who owns it and how you will revisit it.

Why I would limit the meeting to three decisions

AI makes it easy to produce a list of twenty opportunities.

Your operations manager still has one Tuesday.

My working rule would be to bring at most three decisions into the weekly review. Rank them by the cost of delay, the strength of the evidence, and whether someone can act this week.

Do not add these into a pretend scientific score. Discuss the trade-offs openly.

A large opportunity with missing information may need an investigation. A smaller issue with a deadline tomorrow may need an immediate decision. Both deserve attention, but different kinds of attention.

Try this sentence in your next meeting:

“Report samajh aa gaya. Ab kaunsa decision lena hai, aur kaun lega?”

We understand the report. Which decision needs to be made, and who will make it?

The same method outside finance

Consider a manufacturer whose delivery performance is slipping. AI groups the late orders and points to one production stage. Before buying a machine, the manager checks whether the delays come from capacity, missing material or last-minute specification changes. The first action could be to fix the order-release process. The machine decision can wait for evidence.

Or take a professional services firm whose project backlog is growing. A summary says “hire more people.” A Decision Card asks whether the queue is actually waiting for client inputs, partner approval or staff capacity. Hiring helps only one of those problems.

These are illustrative situations, but the discipline is the same. Move from the visible symptom to the decision that would change it.

This is also the management question behind our work on Growth Intelligence: how can verified business information support an accountable choice? The useful part of that question fits on a sheet of paper. Any owner can start there.

A prompt to use with one report

Use a report you are permitted to share in your chosen AI workspace. Remove unnecessary customer and employee identifiers.

Read the card with the person closest to the work. Ask what the AI missed. That conversation is part of the process.

What to do this week

Bring your latest management report to a 30-minute meeting. Spend the first ten minutes checking one important change. Use the next ten to complete a Decision Card. Use the final ten to agree the action or the missing evidence someone will collect.

Next week, open that card before opening the new report.

Did the action happen? What changed? What did you misunderstand?

This gives your meetings a memory. It also gives AI something more useful to work with than a fresh request to “summarise this month.”

The owner’s advantage is in the decisions that improve after each review.

Which decision has appeared in three of your meetings without anyone owning the next step?

Read next

• AI Employee vs Chatbot: What Happened When 52 Indian Businesses Tried It
• The 90-Minute AI Contract Audit That Found ₹2.4 Crore
• What AI Can and Cannot Do in Your Business: The 4-R Framework