Email in the AI era: why retention gets more valuable as ads get more crowded

AI has made advertising much cheaper to produce. It has not made it cheaper to run, and the two facts are connected.

Ad space is sold by auction. When producing an ad took a team and a budget, that limited how many businesses could compete. Now a business of any size can make good-looking ads in an afternoon, and more of them are bidding for the same people's attention.

The prices show it. Meta reported that its average price per ad rose 9% across 2025 and was up 14% on the year in the third quarter of 2026.¹² That is a global average across all advertisers, not the cost of winning a customer for any one brand, but the direction is clear. New ad channels fill quickly too: within 200 days of launching ads in ChatGPT, OpenAI said tens of thousands of advertisers were using them.³

None of this is an argument against advertising. It is how most new customers are found. It is an argument for getting more from each customer once the advertising has done its job.

The economics of keeping customers were worked out long before AI. Research by Frederick Reichheld of Bain & Company, as summarised by Harvard Business Review, found that increasing customer retention by 5% raised profits by between 25% and 95%, and that winning a new customer costs between five and twenty-five times as much as keeping one.⁴

Those figures are decades old and drawn from many industries, so they should not be read as a forecast for any one business. The reasoning behind them has only got stronger. If the cost of acquiring a customer rises every year, the second and third purchase are where the first one starts to pay for itself.

An ad rents attention from a platform. An email goes to someone who chose to hear from you.

Email is how most businesses reach those customers. There is no auction to win and no algorithm choosing which of your customers see the message. The list is yours, and it moves with you if you change platform. In Litmus's 2025 survey of marketers, 35% said email returned between $10 and $36 for every dollar spent and another 30% said between $36 and $50.⁵ Those are marketers' own estimates, not audited results, but few channels are described that way by the people who run them.

There are two kinds of marketing email. A campaign is sent once to many people: a launch, a sale, a newsletter. A flow is sent automatically to one person when they do something: sign up, start a purchase and leave, become a customer.

Brands send far more campaigns, and earn far more per email from flows. Klaviyo, one of the most widely used email platforms, reports that flows make up 5.3% of sends and produce nearly 41% of email revenue, with revenue per recipient almost eighteen times that of campaigns.⁶ Omnisend, a competing platform, found the same pattern in its own data: automated emails were 2% of sends and 30% of email revenue.⁷ Both companies sell email software, but they are measuring different customers and reaching the same conclusion.

The reason is timing. A flow arrives when the person has just shown interest, with a message about the thing they were looking at.

The figures are averages across Klaviyo's customers in its 2026 benchmarks.⁸ Omnisend adds a useful detail: abandoned cart and welcome messages together produced 76% of all orders from automated emails.⁷ If a brand has only two flows working well, those are the two.

These benchmarks come from online retail, because that is where the public data is. The pattern holds in any business. A software company's email as a trial ends, a clinic's follow-up after an appointment and a firm's reply to a new enquiry are flows too: sent to one person, at the moment they showed interest.

• —Welcome. A short series, not one email. Who you are, why the product is different, the proof, and a reason to place a first order. This is the highest-attention moment you will get with a new subscriber.
• —Abandoned cart. Or its equivalent: an unfinished booking, quote or sign-up. Start with a plain reminder showing what was left. Follow with an answer to the likely objection: delivery, returns, whether it works. Keep any discount for the last message, so you are not paying people who would have bought anyway.
• —Post-purchase. How to get the most from what they bought, what to expect and when. This is the flow that turns a first sale into a second, and the one most businesses leave as a bare receipt.
• —Win-back. For customers who have not bought in a while: what is new, and a reason to return. It earns the least per email and still pays, because the alternative is buying that customer again through ads.

A few rules apply to all of them. Each email should have one job. What the email shows should be what the person looked at or bought. An offer, when there is one, should be impossible to miss. And the claims should be ones you can support, which matters as much in an inbox as in an ad.

Three changes in the last few years affect how email should be run and measured.

1. 01Sending rules are stricter. Since February 2024, Gmail has required bulk senders to authenticate their email properly, offer one-click unsubscribe, and keep spam complaints below 0.3%.⁹ Brands that send too often to people who do not want it now risk not being delivered at all.
2. 02Open rates no longer mean much. Apple's Mail Privacy Protection loads emails automatically, whether or not anyone reads them. Litmus estimates it now affects roughly 55 to 60% of all email opens.¹⁰ Clicks and orders are the numbers to trust.
3. 03AI is starting to read the inbox first. Gmail introduced AI summaries and an inbox view that highlights what it judges to need action in early 2026.¹¹ An email may be summarised before it is seen, so the subject line and the opening words need to say plainly what the email offers.

Flows are valuable and neglected for the same reason: they are a lot of work to build well. A full set for one brand can run to thirty or forty designed emails, each needing copy, images, products and timing, and each needing a refresh when the offer or the season changes. Many brands build two or three, and stop.

This is work AI is well suited to, with the same condition as for ads: the emails have to be built from what is true about the brand. Made that way, a brand can have every flow in place, in its own style, and updated as often as the business changes.

At Lightvector, email is planned and produced by the same system as the ads, from the same record of the brand, so the customer who was won by an ad about one promise is welcomed by an email that keeps it.