Retail Technology Has a Speed Problem, Not a Scale Problem

The retail industry spent the past decade building big. Scale was every CIO's mantra — massive systems designed to support millions of users and petabytes of data. And for a while, that strategy worked. E-commerce platforms handle holiday spikes. Cloud infrastructure powers businesses of every size. The engineering challenge of scale has largely been solved. However, solving scale created a different problem.

Retailers that once competed on their ability to build robust infrastructure now find themselves unable to respond quickly enough when the market shifts. The systems they built to grow are now the systems preventing them from adapting. And in the artificial intelligence era, adaptability and responsiveness, not size, is the moat.

The Real Legacy Burden

The conventional wisdom is that legacy systems are a technology problem: old code, outdated platforms, risky customizations. That's true, but it misses the deeper issue.

The real burden of legacy isn't the codebase. It's the hundreds of manual workarounds that grew up around systems that couldn't adapt. Every time a retail system couldn't handle an exception (a new fulfillment rule, a regional pricing variation, a promotional edge case), someone built a workaround. A spreadsheet. An email chain. A manual review step. Over time, those workarounds became invisible infrastructure. Each one added decision latency, and each one made the next change harder.

When a retailer finally tries to introduce AI into pricing, demand planning, or fulfillment optimization, the AI can't reach the actual work because the actual work doesn't live in the system of record. It lives in tribal knowledge, shared drives, and the judgment of the three people who know how the process actually runs.

That's not a modernization problem. It's a decision-speed problem – and it compounds every quarter it goes unaddressed.

Enterprise tech leaders have succeeded in building big. Velocity was traded for size, and that trade-off is starting to show. In the AI era, scale no longer acts as a protective moat. It’s become a constraint, locking retailers into systems that are difficult to change when it matters most.

The Shrinking Moat

AI has fundamentally changed the economics of building software. A fulfillment reliability issue that a retailer estimated would take six months to re-engineer can now be resolved in weeks using AI-accelerated development. A pricing optimization workflow that once required a multi-quarter integration project across ERP, point of sale, and demand planning can be prototyped as a thin, instrumented slice, tested against real transaction data and refined based on what actually moves margin.

This isn't theoretical. Engineering teams using AI-assisted development are delivering at multiples of the pace of traditional approaches, not by cutting corners, but by compressing the iteration cycle so dramatically that the constraint shifts from build speed to decision speed.

The implication for retailers is stark. The historic first-mover advantage, once protected by the high cost of building competitive systems, has eroded. Retailers burdened by large operational legacy codebases may find themselves outpaced not by better-funded competitors, but by smaller, faster ones that can build modern capabilities organically in this new way of working.

The competitive metric is loop speed — the time it takes to turn data into a decision, a decision into a deployed change, and a deployed change into measurable signal. The retailers winning today are compressing that loop from quarters to days.

Goodbye 12-Month Road Maps 

If your technology planning cycle is longer than a single promotional season, you're building for a market that no longer exists by the time you ship.

The traditional approach (extensive upfront planning, detailed specifications, long-term road maps) made sense when the cost of change was high. Organizations invested months in requirements gathering and architecture because getting it wrong was costly at best and catastrophic at worst.

AI has inverted that equation, and what replaces the long-horizon road map is a fundamentally different commitment model. Instead of front-loading planning to reduce risk, leading retailers are inverting the sequence: identify the highest-risk assumption in a proposed change, build the smallest meaningful test against real operational data, and scale investment only when real-world signal justifies it.

AI makes this viable because iteration costs have collapsed. However, the discipline isn't in building faster. It's in knowing what to test first. The retailers still tied to traditional development cycles will find themselves locked in planning meetings for a feature while competitors build, test, learn, and refine multiple iterations of that same capability.

Taking Back Control

The path forward is not rip-and-replace. Retailers don't need to abandon the systems they've invested millions in building. They need to build on top of them, using AI to extend and enhance existing operational systems with capabilities tuned to the specific operations that drive margin: pricing, fulfillment, demand planning, and inventory allocation. Well-understood functions stay on established platforms. Competitive differentiation gets built, tested, and refined at the speed AI now allows, but speed-to-deploy is necessary, not sufficient.

The hardest step in any AI-accelerated capability isn't building it. It's changing how people work. A pricing optimization that merchandisers don't trust will be overridden manually. A demand signal that planners can't interrogate will be ignored. An inventory algorithm that store managers don't understand will be worked around, and we're back to the spreadsheets and tribal knowledge that created the legacy burden in the first place.

Retailers need systems that learn fast, where data flows into decisions, not dashboards. Every day of internal adoption friction is a day your customer sees the old experience.

The moat isn't infrastructure anymore. It's loop speed, and the organizational willingness to trust what it produces.

Derek Perry is the chief technology officer at Sparq, a provider of digital engineering services and enterprise AI solutions.