Battery-as-a-Service is often defined by the scale of its physical infrastructure: how quickly a network of stations can expand, how reliably batteries can be swapped, and how efficiently assets can be utilized. In reality, the physical scale is only half the equation. Beyond a certain point, sustainable growth depends on an organisation's ability to capture, interpret, and operationalise knowledge at scale. This intelligence layer is becoming the foundation of next-generation BaaS platforms.
The first generation of Battery-as-a-Service was defined by infrastructure. The next generation will be defined by intelligence. As battery service networks expand across cities and serve millions of battery exchanges, the challenge is no longer simply adding more stations or batteries. It is building systems that learn from every operational interaction and become smarter with every decision. Thousands of daily operational decisions, ranging from onboarding verification to station deployment and exception handling, begin to define system performance.
Increasingly, Artificial Intelligence (AI) is moving beyond isolated use cases and becoming part of day-to-day operational workflows. Rather than being limited to dashboards or retrospective analytics, AI can help surface relevant information, standardise decisions, and reduce dependence on fragmented institutional knowledge.
One of the most important expressions of this approach is what we refer to as the Second Brain—a structured knowledge base combined with a hosted MCP server that powers AI analytics agents across the organisation. Its purpose is to make institutional knowledge accessible in natural language, enabling teams across functions to query complex operational questions and receive answers grounded in validated organisational context. Crucially, this system is not static documentation. It is designed as a learning loop: Incorporating corrections from domain experts, reinforcing validated knowledge, and improving continuously without requiring model retraining. The implication is significant. Knowledge no longer degrades as it scales, it compounds.
A similar transformation is underway in field operations. Historically, many verification processes in onboarding, such as asset validation or quality checks—depended on manual inspection and subjective judgment. As scale increased, so did variability.
{{/usCountry}}A similar transformation is underway in field operations. Historically, many verification processes in onboarding, such as asset validation or quality checks—depended on manual inspection and subjective judgment. As scale increased, so did variability.
{{/usCountry}}Recent advances in computer vision and Vision Language Models (VLMs) are enabling parts of these workflows to be automated. These systems now support workflows such as baseplate verification, RC extraction, motor controller validation, and visual similarity checks for onboarding assets. The impact is twofold: Operational turnaround times decrease, while consistency and standardisation improve significantly across high-volume processes. What was once dependent on human interpretation is now encoded into scalable, repeatable intelligence systems.
Network planning is another area where data-driven approaches are becoming increasingly relevant.
Historically, decisions on where to deploy new battery swapping stations often depended on local knowledge, field surveys, and operational experience. Today, these decisions can be supported by combining multiple data sources, including computer vision-based estimates of EV density, optical character recognition (OCR), mobility data, traffic patterns, and demographic information.
AI models can synthesise these inputs into structured decision-support tools, helping prioritise locations based on objective criteria while complementing human judgment rather than replacing it.
The goal is not to remove human judgment, but to augment it with systems that are more comprehensive and continuously learning. Taken together, these efforts reflect a broader shift: From digitising operations to embedding intelligence within them.
The evolution of Battery-as-a-Service reflects a broader shift occurring across many industries. Digital transformation is moving beyond the digitization of existing processes toward systems that continuously learn from operational data.
In this model, every service interaction, inspection, maintenance event, or operational decision contributes to a growing body of organizational knowledge. Over time, this creates feedback loops that improve planning, operational efficiency, and decision quality.
As Battery-as-a-Service ecosystems continue to mature, physical infrastructure will remain indispensable. Equally important, however, will be the knowledge and intelligence systems that help organizations operate that infrastructure more effectively. The next phase of growth is likely to be shaped not only by the expansion of assets on the ground, but also by the ability to transform operational experience into continuously improving decision-making.
(The views expressed are personal)
This article is authored by Tejasav Khattar, senior vice president, engineering, data science and product, Battery Smart.