In 2025, software engineering underwent profound change. A new generation of AI models (Cursor, Claude Code, Codex), paired with a maturing agentic infrastructure, crossed an invisible capability threshold. By the end of the year, developers moved from “AI helps me write code I carefully review” to “I can orchestrate teams of agents that automate most of the development process.” What made it happen wasn’t smarter models, but the infrastructure around them finally catching up.
The rest of the enterprise world is next. But most industries aren’t close to being ready, especially one of the world’s most important: retail. In order to understand current retail data challenges, it is helpful to take a step back and examine where enterprise software has been. The same forces that reshaped many other industries are converging on retail, but with even higher stakes.
PHASE ONE: THE SYSTEM OF RECORD
In the 1990s, enterprise resource planning systems solved a real problem for retailers. SAP and Oracle each promised one system, standardized data, and consistent processes. It worked, but implementations took years, armies of consultants, and systems so rigid they couldn’t keep pace with the market. By the time data were clean, the world had shifted.
PHASE TWO: THE SAAS EXPLOSION
The cloud era democratized procurement. We could spin up a SaaS tool in days. But the side effect was severe fragmentation. The average enterprise today runs 275 to 342 SaaS applications, more than doubling usage over the past five years, with no shared data layer and no consistent logic. Cloud data warehouses added powerful infrastructure for IT, but nothing for the retail business users who needed an answer by Monday morning. Storing everything in Snowflake doesn’t tell a category manager what to do at their biggest retail account next quarter.
PHASE THREE: PERSONAL ENTERPRISE SOFTWARE
AI automates tasks, but also enables nontechnical people to build software. A demand planner, trade marketer, or category manager can describe what they need and get a working application. The individual becomes the builder.
I call this personal enterprise software: applications built for, and increasingly by, the people doing the work, instead of the IT department procuring on their behalf.
What’s needed but often missing, is specialized infrastructure including the tools, data, and context that unlock what AI is already capable of. We cannot run retail on slide decks and text files scattered across shared folders. Generic AI on generic data produces generic answers. That’s an extremely expensive search engine.
WHAT NEEDS TO BE IN PLACE
Before we can build the castle, we need the support system. Here’s what that includes.
A vertical data foundation. The AI must know the industry deeply before a user types a prompt. In consumer packaged goods and retail, that means years of normalized, harmonized retailer data from point-of-sale, inventory, velocity, distribution, and space, in a consistent semantic layer. It takes years to build and a generic cloud provider can’t replicate it. The foundation is the moat.
Transparent logic chains. Speed without accuracy introduces risk. Every recommendation must be traceable to its data inputs and business rules, not a black-box output. Without this, teams spend as much time validating outputs as they would conducting the analysis themselves. Trust compounds when reasoning is visible.
Context that compounds. AI without memory is limited to one-off analysis. Vertical platforms learn your language, your KPIs, and your decision patterns, and carry that forward.
When these three are in place, the fourth follows.
The ability to build. Nontechnical users can create purpose-built applications in hours for pricing, promotions, supply chain, category management, and retailer meeting prep. Developers build net-new applications on the same backbone, and the platform becomes an ecosystem.
THE BATTLE PLAYING OUT RIGHT NOW
There is a fight inside every major enterprise and it’s about control. On one side is IT and procurement, protecting hundreds of millions in commitments to last-generation platforms. On the other hand are business users who can’t wait for modernization.
Finally, the business user has another option. For example, Crisp client Jeff Garde at Kraft Heinz Away From Home developed a full distribution strategy with a top retail partner in under 10 minutes using vertical AI. Teams that would have waited 18 months for IT approval are running their own applications today. Enterprises can now enable their business teams to become application builders.
This evolution is a workforce shift and a technology shift. Instead of teams grouping around functions, they can rally around outcomes. An analyst who used to pull data, handed it to someone who built a model, who handed it to someone who built a deck, can now do all three before lunch. It is about leveraging teams that are sharper, focused, and closer to the decisions that matter. The most innovative companies are mandating AI proficiency the way they once mandated email.
WHAT COMES NEXT
The on-premise era gave rise to SAP and Oracle. The SaaS era gave rise to Salesforce and Workday. The success of the personal enterprise software era will be defined by the depth of industry specialization and data, and the quality of the infrastructure on which individuals can build.
The shift is already underway. The question for those in retail: Is your organization enabling your business to respond in real time, or are you still waiting for IT to approve?
Are Traasdahl is CEO and founder of Crisp.
