AI deflation: what gets cheaper for consumer brands, and who keeps the savings

Meet Priya. She runs a made-up travel mug brand.

Last year, her weekly report took a full day. Now a first draft is ready before her coffee cools.

Ad concepts used to take a week of freelance time. Now ten arrive in an afternoon.

Her bookkeeper adds it up. About $90,000 a year, saved.

Priya is thrilled. Then the Monday numbers land.

Ad cost per order is up. Again. Her rivals got the same tools, and they are bidding harder.

So where did the money go?

AI deflation is the falling cost of knowledge work, and for most consumer brands the savings get competed away unless they become better decisions or something rivals cannot copy.

What we mean by AI deflation

Deflation means prices falling over time. Here we mean one kind of price: the cost of an hour of thinking work.

Knowledge work is work done with information instead of things. Pulling a sales report. Writing a listing. Checking whether ad spend matches the invoice. Drafting an email flow. Most of it used to need a person for every hour.

AI now does a large share of the first draft. A person still checks it. But the hours per task fall, and so does the cost per task.

There is early evidence for this. One large study followed customer support agents who got an AI helper. They became nearly 14% more productive (NBER working paper, Generative AI at Work). Economists track this as output per hour. The U.S. Bureau of Labor Statistics says rising productivity tends to lower labor cost per unit of output (BLS Productivity and Costs, Q2 2026).

That is the whole idea. The cost of a unit of analysis, reporting or drafting goes down. The cost of a unit of product does not.

A word on what this post is. It is our view, not settled economics. We have no survey of brands to point to. What we have is a way to think about it, a made-up brand and a checklist.

Here is where it gets interesting. The cheap part and the expensive part are not the same part.

What gets cheaper, and what does not

The split is simple. Work done with words and numbers gets cheaper. Work done with atoms and auctions does not.

Physical costs are still rising. U.S. consumer prices rose 3.7% in the 12 months to September 2026, and energy rose 20.4% (BLS Consumer Price Index, September 2026). AI did not make fuel cheaper.

The ad line deserves a closer look. An ad auction is a live sale of ad space, where the highest useful bid wins the slot. Google, for one, runs an auction every time someone searches. Your bid, your ad quality and your rivals’ ads all decide where you land (Google Ads Help). Your cost per click is set by the market, not by you.

When making ads gets cheaper for everyone, more brands can afford to test more ads. More tests mean more bidders and more bids. So AI can lower what you spend making an ad and raise what you spend running it. We think that is the most likely place for savings to leak.

Why the savings tend to get competed away

Here is where it gets expensive.

Here is the economics in one line. A cost cut that every rival gets is not an edge. It is the new floor.

Say you are Priya. You save $90,000 a year on reporting and creative work. That figure, like every number in this post, is illustrative. Your three closest rivals save about the same. What happens next?

• One rival cuts price. Shoppers compare mugs side by side, so you follow. The saving goes to the customer.
• One rival raises bids. It wins more top ad slots, so you raise yours. The saving goes to the ad platform.
• One rival keeps it. For a while. Until the first two moves force its hand.

This is not a law. It is a pattern we would expect in any market where shoppers can compare easily and many sellers can get the same tool. Most consumer categories look like that.

A moat is anything that keeps rivals from copying what makes you money. A loyal customer base. A product people search for by name. Better records of your own business than anyone else has. Savings turn into profit when they feed a moat, and leak when they do not.

An illustrative P&L, before and after

Back to Priya’s made-up travel mug brand, with $3 million a year in sales. All numbers below are illustrative and rounded, in thousands of dollars.

“Knowledge work” means people and tools for reporting, analysis, creative drafts and ops checks. That includes part-time help and agency hours spent on that work.

Read it left to right.

Year one. Knowledge work drops from $300,000 to $210,000. That is $120,000 saved on people, minus $30,000 for AI tools. Profit jumps from 5% to 8%. This is the number everyone quotes.

Competed away. Rivals catch up. Bids rise, and the brand spends $90,000 more on ads to hold the same sales. Profit is back to 5%. The brand is no worse off. It is just not better off either.

Turned into decisions. Same savings, used differently. The brand spends its freed hours finding the ad sets that lose money and cutting them sooner. It stops pushing a mug that looked like a best seller but lost money after fees. Ad spend falls by $60,000 with no drop in sales. Profit reaches 10%.

We set the ad change in the third column equal to the savings to make the point clear. Real markets are messier. Some savings stick for longer, some leak faster. The direction is what matters.

Notice what did not move in any column: product cost and fulfillment. AI did nothing to the cost of the mug or the cost of shipping it. If you want those lower, you still have to negotiate.

So how do you end up in the last column, not the third?

Three ways to keep the savings

1. Make better decisions, faster

The real prize is not the cheaper report. It is the call the report makes possible.

A weekly report that took a day often arrived too late to act on. A report ready Monday morning can change Monday’s budget. The value lives in the decision, so measure it there: dollars of ad spend cut from losing campaigns, stock-outs avoided, prices tested.

For a way to find the products that only look profitable, see contribution margin for Amazon and Shopify.

2. Own something rivals cannot buy

Every brand can buy the same AI. Not every brand has clean, written records of its own business. That means what each product costs to land, which ads send shoppers where, and what was tried last spring and why.

That record is hard to copy because it is yours. It also makes every AI answer about your business better. Our guide to AI-ready data covers what that record needs.

3. Spend the savings where costs do not deflate

If the savings will leak somewhere, choose where. Put them into the product, the packaging or the customer experience. Those are the things shoppers compare. A better product holds its price. A cheaper report does not.

Where AI deflation goes wrong

Faster wrong answers. A cheap analysis that compares cash paid with orders placed is not a saving. It is a cost that arrives later. The ten ways AI misreads Amazon and Shopify numbers are a good list of traps.

Reports nobody reads. When reports cost nothing, teams make more of them. Ten reports nobody acts on are worth less than one that changes a budget.

Cutting the judgment. The person who knew a fee jump was a one-time charge is worth more now, not less. Cut the copy-and-paste hours. Keep the people who say “that number looks wrong.”

Counting the saving twice. Year-one savings often show up in a plan as permanent profit. Our view is to plan as if they will be competed away, and treat anything you keep as upside.

Where Synthesis fits, and where it does not

Synthesis is ontology software and an AI harness in one plan. It writes the business down once, so every number means one thing, and keeps the AI you already use on those numbers. It checks figures against their source. For example, Amazon sales tie to settlements and ad spend ties to each account. It sends a Monday report and can make approved changes in your ad accounts and store.

So it sits on the cheaper side of the table: analysis, reporting and ops checks. It is $999 a month for the whole team.

It will not lower what your mugs cost to make or ship. It does not set ad prices. And it does not decide for you. The point of cheaper analysis is better calls, and the calls are still yours.

A checklist you can run this week

1. List your knowledge work in dollars. Reporting, analysis, creative drafts, ops checks. Include agency and freelance hours.
2. Mark which lines AI could cut by half. Be honest. A first draft is not a finished job.
3. Write down where the savings will go. Ads, price, product, profit. If you do not choose, the auction chooses.
4. Track ad cost per order every month. If it climbs while your own costs fall, the savings are leaking.
5. Pick one decision to speed up. Cutting losing ad sets, reordering stock or testing a price.
6. Write down your own records. Landed cost by product, by date. Where each ad sends shoppers. What you tried and why.
7. Plan as if year-one savings will fade. Keep what sticks as a bonus.

The short version

AI makes thinking work cheaper. It does not make products, freight or ad slots cheaper. When everyone gets the same discount, the market tends to take it back.

So treat AI savings like any other windfall. Decide where it goes before someone else decides for you. The brands that keep it will be the ones that use it to see sooner and decide better.