Part 3 of a three-part series exploring the evolution of Energy Management development on Cumulocity—from foundational platform capabilities to Digital Twin Manager, AI-assisted development and vertical accelerators.AI is changing how connected products are built — not just what they can do.
Read Part 1 here
Read Part 2 here
For IoT architects and backend developers, the opportunity is particularly practical. AI can remove much of the repetitive work involved in turning an idea into a working connected application: defining asset structures, generating application components, and creating realistic data to validate the result.
But AI alone isn’t the accelerator.
The real speed-up comes when AI has the right platform context, tools, and interfaces to work with. Instead of generating disconnected code that developers still need to integrate manually, AI can work with the capabilities of an IoT platform to help create something much closer to a deployable application.
To demonstrate what that looks like, consider a Smart City Energy Distribution Microgrid built with Cumulocity.
The objective is to monitor a multi-source energy network spanning the municipal grid, renewable generation and diesel fallback, with critical infrastructure including HT incoming feeds, district HVAC systems and UPS panels.
Normally, getting from that description to a functioning IoT application involves a considerable amount of plumbing.
With AI and Cumulocity-native tooling, much of that work can be accelerated across three distinct phases.
Phase 1: Model the connected world
Before building dashboards or writing simulation scripts, the application needs a model of the real-world infrastructure it represents.
That means defining assets, their relationships and the hierarchy that connects them.
For our microgrid, the hierarchy needs to capture multiple energy sources and distribution points, as well as the critical infrastructure consuming and managing that energy.
Rather than constructing that hierarchy manually, AI can help translate the use-case description into an asset blueprint.
The developer describes the required energy network and its relationships. With the appropriate Cumulocity context and tooling, AI can help generate the structure needed to represent it digitally.
The important point is that this happens before simulation. The first output isn’t synthetic telemetry. It’s the foundation of the actual application.
Phase 2: Build the application
Once the asset model exists, the next challenge is turning it into something operators can actually use.
Imagine that the operations team needs an alarm widget showing abnormal conditions across the microgrid. Traditionally, a developer needs to understand the relevant APIs, create the component structure, connect the data and implement the UI before there is anything useful to demonstrate.
AI can shorten that cycle considerably.
Using capabilities such as Cumulocity AI Agent Manager and the HTML Widget Generator with AI, developers can move from intent to working UI components faster. And through Model Context Protocol (MCP) integration, an external AI coding workbench can interact with relevant Cumulocity capabilities while developers continue working in their preferred environment.
Instead of asking a generic coding assistant to produce an isolated snippet, the goal is to give AI useful platform context.
A developer can describe the outcome they need — for example, an alarm widget for critical energy assets — and use AI to accelerate the boilerplate and integration work required to create it.
The developer remains in control of the architecture and implementation. AI takes on more of the repetitive groundwork.
Phase 3: Simulate and validate
Only now do we need the simulator.
Once the asset structure and application components exist, AI can help generate simulation scripts that produce realistic telemetry for the microgrid.
That might include:
- phase voltages
- current
- active and reactive power
- power factor
- energy consumption
- equipment state
- alarms and abnormal operating conditions
The simulated data can then flow through the same application that will ultimately consume operational data.
This gives developers something much more useful than a static UI prototype: a working application they can exercise against changing conditions.
They can see how dashboards behave, trigger alarms, validate asset relationships and identify problems before connecting the application to real infrastructure.
From days of plumbing to under an hour
This is where AI becomes more than a developer convenience.
Across these three phases, work that could traditionally consume days of manual setup can potentially be compressed into under an hour — delivering an order-of-magnitude, or approximately 10x, acceleration for this type of build workflow.
The speed-up doesn’t come from asking AI to build an IoT solution from scratch.
It comes from combining AI with a platform that already provides the building blocks for connected products: asset modelling, device and data management, APIs, visualization, alarms, lifecycle management and integration points.
AI helps developers assemble and extend those capabilities faster.
Cumulocity provides the foundation that makes the acceleration useful.
Faster doesn’t mean uncontrolled
There is another reason this distinction matters.
As AI becomes part of connected-product development and operations, organisations also need to understand how those systems are governed. The EU AI Act introduces a risk-based framework for AI systems used or placed on the EU market, with obligations depending on the nature of the AI system, its risk and the organisation’s role.
Not every automated capability is necessarily an AI system under the Act. A deterministic threshold rule, for example, is different from a system that infers from inputs to generate predictions, recommendations, decisions or content.
Likewise, responsibilities can differ depending on whether an organisation is developing and providing an AI-enabled system or deploying one.
For connected-product teams, this makes visibility important. They need to know where AI is being used, what it is doing, what data and systems it interacts with, and where human accountability sits.
That applies to AI inside the finished product, but it also makes disciplined development practices increasingly important as teams use AI to build products faster.
Build AI into the lifecycle, not around it
The organisations best positioned to benefit from AI won’t be those that simply generate the most code.
They’ll be the ones that combine speed with control.
Cumulocity helps organisations develop, deploy and manage connected products and AI-enabled services on an enterprise-grade AIoT platform. Capabilities around connected asset management, AI-ready data, operational visibility and lifecycle management provide a foundation on which teams can experiment and build without losing sight of how their connected environment operates.
For developers, that means less time spent on repetitive plumbing.
For architects, it means AI can operate within a more clearly defined platform architecture rather than becoming another disconnected layer in the technology stack.
And for organisations navigating requirements such as the EU AI Act, it creates a stronger foundation for understanding where AI is used and how AI-enabled capabilities fit into the wider product lifecycle.
AI is the catalyst. The platform is the accelerator.
The Smart City Energy Distribution Microgrid illustrates a bigger shift in connected-product development.
Start by modelling the connected world.
Use AI and Cumulocity-native tooling to accelerate application development.
Then generate the simulation needed to validate it against realistic conditions.
Phase 1: Model → Phase 2: Build → Phase 3: Validate
The result is not simply faster prototyping. It’s a faster path from an operational idea to a working connected application — with the platform foundations needed to take that application beyond the demo.
That’s the opportunity behind AI-assisted development in Cumulocity: turn days of manual IoT plumbing into a working application in under an hour, while keeping developers in control of what gets built.