AI data deployment is being used to cut costs when it should be driving growth

May 14, 2026

— min read

Henry Owen

Product Marketing Manger

Most businesses deploying AI right now are using it to cut costs rather than generate growth. The logic is appealing: same output, fewer people, better margins. But it misses the more valuable opportunity. Two people working well with AI will outperform what four people could do without it. One person given double the workload and a Claude Enterprise subscription mostly just burns out, and probably starts looking for another job. The businesses that actually win with AI will be the ones that deploy AI data deployment against growth targets, own their own data and decisions, and treat an AI data platform as a capability multiplier rather than a headcount substitute.

The margin trap

When a business asks how AI can help, the answer it most often hears back is: you can do the same work with fewer people. The unit economics look good on a slide. Same output, lower headcount, cleaner margin.

The pattern shows up with uncomfortable clarity in a letter published in The Guardian recently. A freelance memoir writer described how her employer restructured her work around AI: she interviews clients as before, an LLM now writes the first draft, and she is paid half her previous fee to edit the result. The editing takes as long as the writing used to. Same hours, half the money. Her employer calls it efficiency. She calls it a cynical way to get work done at half the cost.

The Guardian article that prompted her letter documented something similar playing out across the translation industry.

Paul Coggins, CEO of Kleene.ai, has been watching this across businesses well beyond publishing.

"Too many businesses aren't using AI to drive growth. They're using it to improve margins at the expense of jobs."

AI gets deployed to handle the parts that were never the real source of value. The assumption then follows that the human contribution has been proportionally reduced. Often it has not been. Often it has just been repackaged at a lower rate.

What deploying an AI for growth actually looks like

A marketing team using an AI data platform for growth does not look like one person cranking out twice as much email. It looks like a team that can test things that would previously have taken too long to be worth bothering with.

The more valuable applications tend to be further upstream than most businesses look when they first start evaluating AI tools. Which customers are quietly drifting toward lapse before they actually leave. Whether the marketing spend attributed to a channel by that channel's own reporting matches what the sales data actually shows. What inventory demand is going to look like before the purchase order goes in. None of that is about doing the same things cheaper. It is about having information that changes what you decide to do, at a point when there is still time to act on it.

The companies doing this well are not running leaner teams. They have more people doing more interesting work.

How it applies to your data stack

The same choice shows up in how companies spend on data tools. Most of that spending right now goes toward reducing friction — fewer manual exports, faster reports, one fewer person touching a spreadsheet. Margin thinking applied to data infrastructure.

The growth version of that question is harder to answer but more worth asking. Which decisions does your business make badly because the information arriving is too slow, too incomplete, or too disconnected from what is actually happening? Inventory calls based on last quarter's numbers. Marketing budget decisions made from platform-reported attribution that everyone suspects is wrong. Customer retention conversations that happen after someone has already left.

An AI data platform running on well-structured, connected data can change all of that. But it cannot change it if the underlying data is fragmented across systems that do not talk to each other, or if nobody has done the work of making it trustworthy enough to act on.

The data stack audit guide is a practical starting point for figuring out where yours actually stands.

The data sovereignty problem nobody talks about at procurement

Paul's LinkedIn post flags something that tends to get missed entirely when cost is the primary lens.

"This shift is now being fueled by the infrastructure giants on a buying spree, scooping up companies that integrate AI: their AI (Claude, Gemini or ChatGPT). Lock-in won't show up on the invoice as 'lock-in', it'll show up as bundled pricing you can't escape and the slow death of data sovereignty."

When speed and price are the main evaluation criteria, nobody asks where the data lives or what happens to it downstream. That is fine until it is not. These tools are not being built out of generosity and the business model is not subscription revenue. It is your data becoming more central to their infrastructure over time until moving costs more than staying.

"The infrastructure giants want your data inside their models. Kleene flips that. We build the models, trained on your data — and LLMs simply interrogate them. The intelligence stays in your data layer. Which LLM you use becomes a choice you make, not a dependency you're stuck with."

What feels like a reasonable deal in 2025 tends to look different once you try to leave.

What the businesses actually winning with AI data growth strategy is

They own their data and their models reflect their specific situation rather than what the average company in their sector looks like. When a better tool comes along they can actually switch to it rather than calculating the switching cost and staying put.

"The businesses that win the AI era won't be the ones with the biggest vendor. They'll be the ones with the cleanest data, the clearest decisions, the freedom to plug in whichever model, cloud or tool serves them best — and trusted partners that understand their business."

Getting this right does not just improve margins slightly. It builds a compounding advantage over the businesses that decided AI was primarily a cost management tool and optimized accordingly.