One NZ Reduced Service Deployment from 10 Days to 10 Minutes: How AI Agents Revolutionized Order Provisioning

Telecom operator One New Zealand faced a challenge: the process of deploying new services for customers typically took around 10 days. It was slow, costly, and lacked transparency. After implementing AI agents, the same process now takes less than 10 minutes, and the company gained unprecedented visibility and control over every stage of order processing.

In a large telecom company, service deployment is a complex dance between multiple systems, and every step requiring manual input or verification is a bottleneck. The lack of end-to-end documentation and "tribal knowledge" makes the process opaque, slows it down, and multiplies the risks of errors. And when a process takes 10 days, it directly impacts customer satisfaction and operational costs. This way of working is no longer necessary; there's a solution available today.

The hidden problem of 10-day deployment

At One New Zealand, tens of thousands of new service orders are processed annually. Until recently, this process was entirely manual and handled by a remote team. The problem wasn't just the slowness, but also the complete lack of transparency: there was no end-to-end documentation, and much of the process knowledge existed as "tribal knowledge." If a problem arose, it was almost impossible to figure out which stage it occurred at and who was responsible.

This meant that every order, passing through complex ecosystems of digital, Salesforce, and Oracle platforms, could get stuck at any stage, requiring manual intervention and stretching deployment time up to 10 days. Lost orders, dissatisfied customers, and huge operational costs were the reality.

Why conventional automation wasn't enough

The company already had experience with automation, but traditional approaches focused on "low-hanging fruit" – simple, isolated tasks. In this case, the problem was systemic: multiple platforms and scenarios were involved, and simply automating individual steps was not enough. A tool was needed that could orchestrate the entire end-to-end process, from start to finish, taking into account all possible variations and exceptions.

This is why One New Zealand turned to AI agents: they could not just execute scripts, but act as intelligent coordinators capable of adapting to changing conditions and taking on all the routine management of a complex process.

Designing the AI agent for order orchestration

The company decided to start with the most difficult task to prove the viability of AI agents in the most challenging conditions. The agent was designed as a central orchestrator, intended to manage the entire service deployment process. Its main tasks included:

• End-to-end coordination. The agent had to manage the flow of data and actions between various systems: digital platforms, Salesforce, and Oracle.
• Exception handling. Given the vast number of variations in real orders, the agent needed to be able to recognize non-standard situations and either process them according to defined rules or escalate them to a human.
• Transparency. The main goal was not only acceleration but also ensuring full visibility of every step of the process, so that order status could be tracked at any time.

A strict rule was established: any situation falling outside strict regulations must be immediately handed over to a human for decision-making. The AI agent served as a tool, not an autonomous decision-making entity.

Implementation: from testing to live orders

The implementation proved challenging. The team went through approximately 120 configuration changes during user acceptance testing, which took three weeks instead of the one week originally planned. The main reason: the lack of test environments that adequately replicated real-world process variations. It was necessary to test on live orders to understand how the agent handled unpredictability.

This approach, though longer, provided invaluable insights. Thanks to detailed logging of agent actions, the company gained a complete picture of all process variations, identified non-obvious problems, and was able to optimize both the process itself and the agent's operation. Ultimately, the system was perfected, and the team gained a tool that could work under any conditions.

Results

Reducing deployment time from 10 days to 10 minutes is a monumental breakthrough. But, as Cy Wright, General Manager of One New Zealand, notes, the most significant outcome was not speed, but transparency. The ability to see every step of the process allowed not only for optimization but also for identifying training needs for employees and predicting the impact of changes in source systems.

How to replicate this in your business

The One New Zealand case demonstrates that AI agents are capable of solving the most complex and convoluted process problems. If you have end-to-end processes that suffer from slowness, lack of transparency, and reliance on "tribal knowledge," an AI agent could be the solution. Here's where to start:

• Don't shy away from complex tasks. If an AI agent can handle the most difficult process, it can handle the rest. Choose problem areas where the potential benefit is maximal.
• Ensure full visibility. The main advantage of AI agents is not just automation, but also detailed logging of every step. Use this for a deep understanding and optimization of your processes.
• Integrate the agent into a unified orchestration system. An AI agent should not just be an isolated tool, but part of a larger platform capable of coordinating actions between various systems in your company.
• Keep humans in the loop for complex decisions. The AI agent should be a tool that prepares information and handles routine tasks, but final decisions, especially those affecting customers, should remain with a human.

If this case sounds like what's happening in your company, our manager can help: he'll analyze your business and niche for free and point out where an AI agent would bring a real result in your case. Message the manager