Routine enquiries at charging stations

The majority of support requests follow predictable patterns: membership questions, account issues, payment queries, and how-to guidance for using a charging point. Before automation, agents responded to each of these manually.

Proto’s AI agent now handles these enquiries instantly across WhatsApp – in English and Malay – drawing from ChargEV’s integrated knowledge base. Common questions about pricing, promotions, and account management are answered without agent involvement, around the clock.

Finding and checking charging stations on the road

Drivers often reach out when they’re already on the move, looking for the nearest available charger. The AI agent is connected directly to ChargEV’s charging station network, allowing it to check charger availability and operational status in real time.

Customers can ask whether a specific station is functional, find the nearest available charging point, and receive up-to-date information on pricing and active promotions – without waiting for a human agent. For drivers planning a longer journey, this kind of instant, accurate information is core to the product experience.

Image recognition for driver issues

One of the most time-intensive parts of ChargEV’s support workflow was handling screenshots. Drivers experiencing errors in the ChargEV app or at a physical charging station would send images to the support team, who then had to interpret each one individually before responding.

Proto integrated OpenAI Vision into the AI agent to automate this step. When a customer uploads a screenshot or photo – of an app error, a charger display, or a user interface issue – the agent identifies the problem from the image, provides an appropriate response, and routes the conversation accordingly. Where the issue can be resolved automatically, it is. Where it can’t, the ticket is escalated with the image context already captured.

Image recognition now handles 71% of visual queries automatically, removing most of the manual review that had been generating a backlog for the support team.

Escalation to live agents

For cases the AI agent cannot resolve, the conversation is escalated to a live support agent, with the full conversation history preserved – including any images the customer has shared. Agents have immediate context and do not need to ask the driver to repeat themselves.

This unified handover means AI handles volume and agents handle exceptions, without any gap between the two.

Results across the network

Within two months of going live, the AI agent was managing 64% of all customer enquiries without human intervention – with a 44% reduction in agent involvement. As the agent was trained on new query types and ChargEV’s knowledge base expanded, automation reached 70%, with the system handling over 16,000 monthly interactions across channels.

The solution allowed ChargEV to maintain fast, accessible driver support across its 400+ station network in Malaysia and Singapore – without proportionally increasing the size of its support team.