Integrating Software with Goals and Guardrails

Implementing and integrating enterprise software is hard. Successfully implementing software often necessitates such disparate tasks as complex schema migrations, ongoing ETL pipelines, one-off glue code, and customized visualizations or workflows.

One lens we can use to view these implementation complications is that most products to date are built to solve a generalized version of a problem. Products try to solve the problem uniquely for each business (no one likes to spend time on implementations), but a natural tension exists between a generalized product and the unique needs of each business. The boundary where a generalized product meets a unique business is the realm of implementations. Typically this customized product-business boundary is where services companies like TCS, Accenture, and Infosys get involved. They in part solve the last mile by building bespoke solutions.

However, AI may be ushering in a change in how product-business boundary solutions are built. AI may enable us to build products at the product-business boundary which can account for the uniqueness of each business. The change may be driven in particular by the (amazing) abilities of AI to generate high-quality code. We may be moving to a world where implementations are defined by “goals and guardrails” (see our previous post on goals and guardrails) and where AI-generated code plays a pivotal role in migrating and maintaining enterprise software.

The result could be a world where businesses can quickly optimize their operations and more deeply customize their workflows. Paraphrasing Jeff Bezos, businesses can “focus on what makes the beer taste better” and not the endless months of product implementation and integration.