Giving up on interfaces to rely entirely on agents or APIs is a mistake. As foundational technology becomes cheap and abundant, craft, design, and user experience become the core differentiators that determine whether an enterprise application stands out.
Rather than a single omnipotent agent handling all workflows, software will sustain a blend of horizontal agents, specialized vertical agents, and domain-rich user interfaces. Context switching and focused craft make dedicated interfaces necessary alongside agent-to-agent interactions.
A single-line blank textbox is a poor user interface because users fail to grasp the model's full capabilities and default to basic search behavior. Foundational agent design requires dynamic, contextual controls and explicit tracking of human versus synthetic contribution provenance.
Engineering teams should not prematurely optimize for LLM costs during feature creation. Builders should focus first on shipping magical experiences that achieve adoption; cost reduction and model routing should be engineered later once usage hits enterprise scale.
An effective corporate context layer cannot treat organizational history as static facts. It must actively maintain an updated social graph of current employees and ongoing projects while extracting distilled memories and learnings from past documentation.
Most companies claiming to train proprietary in-house models simply apply domain-specific data and fine-tuning onto existing open-weight foundations. Without proprietary raw data and continuous price-performance advantages, generalized foundational models eventually surpass these bespoke models.
Proprietary static business information is losing its asymmetrical leverage as AI levels access to knowledge. Business success increasingly hinges on imagination asymmetry—the creative ability to recombine technologies and apply intelligence to organizational problems in novel ways.