Autonomous operations begin when AI moves beyond recommendations and starts acting within utility processes. Smartflex eliminates data and process fragmentation by unifying customer operations on a single platform. Alexandria, its native AI engine, builds this foundation to connect analysis with action, within the authorization levels defined by each utility. A real-world billing complaint use case illustrates the potential: a projected reduction of nearly 50% in handling time per case.
From recommendations to action
Utilities have spent years automating repetitive processes. The next step is autonomous operations, where AI agents detect an event, investigate its context, execute authorized actions, and escalate only the decisions that require human judgment. This model applies across a wide range of customer operations, from complaint handling and collections management to unusual meter reading validation and field work order follow-up.
Adding AI agents to fragmented operations is not enough. An agent can make informed decisions only when it has access to complete information and can act in the systems where operations run. When commercial, metering, and field data live in separate applications, the agent analyzes only part of the case, leaving an analyst to carry out the next steps. Autonomy therefore depends as much on the underlying platform as it does on AI.
Smartflex and Alexandria: the foundation for autonomous operations
Smartflex is Open Intelligence’s unified customer operations platform for electricity, water, gas, and telecommunications providers. It helps utilities eliminate data and process fragmentation by unifying Customer Information Systems (CIS), Digital Customer Experience (DCX), Meter Data Management (MDM), and Mobile Workforce Management (MWM) within a single database.
The result is a single source of truth and end-to-end processes that run without intermediate integrations. This design eliminates at least 14 integration points between these four applications, which typically operate as disconnected systems.
Alexandria, Smartflex’s native AI engine, operates on this unified foundation. It enables utilities to configure agents for specific business objectives across customer service and field operations, including supporting complaint resolution, accelerating collections, and handling exceptions such as unusual meter readings and unsuccessful work orders.
Each agent monitors operational events within Smartflex, evaluates possible actions, and executes those authorized by the utility. Alexandria works with the language model the utility chooses, including private models, masks personal data when it is not needed, and logs every action.
Because agents work within the same platform used for billing, customer service, and field dispatch, analysis and action happen in the same environment. This is what makes the transition from recommendations to autonomous operations possible.
A real-world example: handling consumption-related billing complaints
Consumption-related billing complaints are one process where this model is being put into practice. In 2025, seven out of ten such complaints at a multi-service utility were resolved without changing the bill. Even so, an analyst had to review every case and substantiate the response, a process involving six manual steps and 22 to 24 minutes per complaint. Most of the effort went into investigating, not deciding.
In the AI use case under development, Alexandria takes on that investigation. It prepares the analysis and proposed response for an analyst to validate before replying to the customer. The projected results are:
- Handling time per case: from 22–24 minutes to 11–12 minutes, a reduction of nearly 50%.
- Total complaint-handling costs: 47% lower.
- Investment payback: approximately 5.7 months of operation.
The same pattern extensive investigation supporting a decision governed by established criteria appears across other operational processes. This makes complaint handling a useful starting point for understanding how autonomous operations work more broadly.
The four pillars of autonomous operations in Smartflex
Four capabilities enable Alexandria to move from analysis to action within Smartflex. Each applies across multiple processes. The complaint-handling use case, alongside other Smartflex results, illustrates how these capabilities work in practice.
1. Real-Time Triggers: operations respond as events occur
In traditional operations, an event, a complaint, a payment, or an out-of-range meter reading usually waits in a queue until someone reviews it or systems synchronize.
In Smartflex, each event immediately activates the corresponding agent or workflow, with customer and service information available from the outset. In the complaint-handling use case, registering a complaint triggers the agent to gather the relevant service and customer information, including the customer’s own description of the issue, and filter out ineligible cases before beginning the analysis.
2. Agentic Workflows: the investigation adapts to each case
A traditional workflow follows a fixed path. In an agentic workflow, the agent determines which information to consult based on the objective and context of the case. Each finding informs the next step.
For a high-consumption complaint, the agent verifies that the complaint was submitted within the required timeframe, checks that meter readings are consecutive, compares the disputed consumption with previous months, and calculates whether the deviation exceeds regulatory thresholds.
Instead of cross-referencing data across multiple screens, the analyst receives the case with the analysis and proposed response ready for validation.
Unified information already helps utilities accelerate customer service with Smartflex. One customer, for example, reduced its average customer request resolution time by 87%.
3. Closed-Loop Execution: diagnosis leads to action
Closed-loop execution connects the diagnosis with the actions needed to resolve a case, executing them directly within Smartflex without intermediate integrations or delays while another system synchronizes.
This real-time coordination can help reduce unnecessary field visits, unproductive travel costs, and complaints related to avoidable service disconnections.
In the complaint-handling use case, Alexandria provides the proposed response together with its supporting evidence: meter readings, consumption data, calculated thresholds, and a suggested conclusion. The analyst then validates the response.
By building this foundation, the utility can progressively authorize additional actions, such as closing a complaint or adjusting a bill, as it defines the conditions under which those actions may be executed.
4. Human-in-the-Loop Governance: the utility stays in control
Alexandria’s autonomy operates within the confidence thresholds and authorization levels defined by each utility. These controls determine which actions can be executed autonomously and which require approval, enabling a gradual, process-by-process approach:
- Assisted: the agent investigates and prepares the response; a person validates it. This is the current scope of the complaint-handling use case.
- Supervised: for decisions that exceed a defined threshold, such as credit adjustments above a specified amount or service disconnections, the tangent completes the analysis and proposes an action for a supervisor to approve with a single click.
- Autonomous: for high-confidence tasks with low financial risk, such as adjusting meter readings within normal ranges, the agent acts directly, provided the utility has explicitly authorized it.
Every decision is logged and auditable a key requirement for regulators and auditors in any market.
This traceability also reduces compliance effort by creating structured documentation of decisions involving readings outside normal parameters. One multi-service utility reduced the time needed to respond to oversight bodies requesting information on estimated consumption by approximately 77%.
Moving toward autonomous operations with Smartflex and Alexandria
Autonomous operations require more than adding AI. They depend on agents working with unified data and processes. Smartflex eliminates the fragmentation that prevents AI from acting, while Alexandria turns that foundation into agents that investigate, execute authorized actions, and escalate decisions requiring human judgment.
Complaint handling is a starting point. The same model extends to collections, unusual meter readings, and field operations. Autonomy is built one process at a time, reducing repetitive work, increasing capacity to handle exceptions, and allowing teams to focus their judgment on the decisions that truly require it.
Identify the processes where your teams spend more time investigating than deciding. With that assessment, Open Intelligence can estimate the savings Alexandria could deliver and determine which processes are ready to advance toward autonomous operations within Smartflex.
Schedule an assessment session with our team.
