The Keyboard Problem
Every retail stack ships with a human permanently bolted on. The next platform shift isn’t a smarter operator. It’s no operator.
Ilyas Kurklu, CEO & Co-founder, Replenit
On January 9, 2007, Steve Jobs put four phones on a slide and called them “the usual suspects.” A Motorola, a BlackBerry, a Palm, a Nokia. Then he pointed at the thing they all had in common: a plastic keyboard, fixed in place, occupying the bottom forty percent of the device,“whether you need it or not.”
The room laughed. The industry didn’t get the joke for another two years. Because Jobs wasn’t attacking those four companies. He was attacking an assumption so deep that every vendor had stopped seeing it: the interface must be physical, so the interface must be fixed. Every roadmap in the industry was a variation on better keys. Smaller keys. Backlit keys. Predictive keys. Nobody was asking whether the keyboard should exist.
I’ve written before about the three taxes retailers pay to run their customer relationships, the segment tax, the content tax, the martech tax. Customer pays these tax, and tbh vendors care so little!
Rebuilding audiences that decay the moment you save them. Producing forty creative variants to approximate relevance. Paying seat licenses for tools that generate work about the work. Three new products, three invoices. None of them offers optimization in a AI period!
Here is what I didn’t say plainly enough: they are not three taxes. They are one tax, itemized. And the landlord collecting all three is the same assumption Jobs pointed at in 2007, a fixed interface, sitting in the middle of the machine, whether you need it or not.
In retail’s case, the keyboard is the human operator.
The lineup. Four operator consoles, each with a human bolted to the bottom forty percent, and each now shipping “AI assist” for the keys. The fifth device has no keyboard at all.
Better keys
Look at any “AI-powered” release in the customer stack this year. AI subject lines. AI send-time optimization. AI audience suggestions. Copilots that help you build the segment faster, draft the campaign faster, read the dashboard faster.
Every one of these is a better key. Smaller, backlit, predictive, and bolted to the same board. The human still defines the audience. The human still intervenes the journey. The human still schedules the send, reads the report, and starts over next Monday. AI is being sold as a way to type faster on a keyboard that shouldn’t exist.
The taxes aren’t the disease. They’re the symptoms of one assumption: that a human must sit in the middle of every decision.
That’s why the segment tax never goes away no matter how good your CDP gets, segments exist because people can’t reason about customers one at a time, so we compress millions of humans into thirty boxes a team can hold in its head. The content tax exists because thirty boxes still isn’t relevance, so we brute-force it with variants. The martech tax exists because every tool in the stack is an operator console, and consoles need operators, training, agencies, and admins.
Remove the operator, and all three invoices stop arriving at once.
One landlord, three invoices. Each tax compounds with scale because the deciding capacity, the team, stays fixed. Software’s cost per decision doesn’t.
The math the keyboard can’t do
Here is the arithmetic nobody in the stack wants printed. A strong CRM team can genuinely reason about, hold in working memory, argue about in a meeting, maintain with discipline, maybe a few dozen segments. That number does not move. Hire more people and coordination eats the gains; it’s why ten-person CRM teams still ship the same “VIP / active / lapsed / at-risk” grid as three-person teams.
Get Ilyas Kurklu’s stories in your inbox
Join Medium for free to get updates from this writer.
Meanwhile, the number of decisions your customers actually require is not a few dozen. It is the number of customers you have, times every lifecycle moment: is this the week her serum runs out, is this the family drifting to a competitor, is this the wrong month to discount, is this the right adjacent product, for this person, today. A million customers is a million standing questions, refreshed daily.
Why segments exist. Averaging was never a choice. It was the only output a fixed-capacity operator could produce. The gap is the market, platforms, and the lack of vision.
Full screen
The iPhone didn’t win by shipping a better keyboard. It won by making the interface software, so the whole surface could become whatever the moment required. A keyboard when you type. A map when you navigate. Gone when you don’t need it.
Apply the same move to the customer stack and you get something that is categorically not another marketing tool. You get a employee. An AI CRM Manager that builds a memory of every customer, every product, and the brand; applies expert retail skills, replenishment, cross-sell, churn, winback, engagement, promotion discipline, and produces one Golden Decision per customer, per moment. Not a prediction handed to a human to act on. The decision itself, dispatched as a ready-to-send instruction to the channels you already own.
Notice what does not get replaced: the usual suspects. Braze, Salesforce, Klaviyo, Bloomreach, Insider, those are the glass and the radio. They stay. What goes is the operator layer between your data and your customer. This is why we’ve stopped calling this category martech. Martech is consoles for operators. This is commercial intelligence, the deciding layer of the business, and retail is only its first proving ground.
Same data, same channels, no keyboard. The operator layer between data and customer is replaced by a worker that decides, dispatches, and learns.
This isn’t a concept video
In 2007 the room could dismiss the demo, until people held one. So here’s what holding one looks like. At Mumzworld, MENA’s leading mother-and-baby retailer, our AI CRM Manager was layered on top of their existing Braze setup. Seven days to first value. Three channels activated. On the replenishment workflow alone: over 100X ROI, because it stopped asking “which segment should get the restock campaign” and started answering “is this the week this family runs out of formula,” a million times over, every day. L’Occitane, eBeBek, and Beko run on the same engine, each on top of the stack they already had: https://replen.it/case-study/ai-crm-manager-mumzworld-braze
Nothing was ripped out. Nothing was replaced except the assumption.
The incumbents will respond the way incumbents always do: better keys. More copilots, more assist panels, more AI sprinkled onto operator consoles, because their business model is the console, and the console needs you sitting at it. That’s not cynicism; it’s structure. BlackBerry’s finest engineers spent 2007–2009 perfecting keyboards. They weren’t wrong about keyboards. They were wrong about whether keyboards should exist.
Every tax you pay in your customer stack is rent on a human bottleneck. Stop renting the bottleneck.
We’re not shipping a better keyboard. We removed the human from the equation, and gave the humans better jobs: setting strategy, guarding the brand, deciding what the machine should optimize for. The operator role ends. The conductor role begins.
Ilyas Kurklu, CEO & Co-founder, Replenit · Warsaw
Maestro, the world’s first AI CRM Manager: https://replen.it/ai-crm-manager
