Superloop self-serve AI resolutions top 330,000 cases

Superloop has revealed that its agentic chatbots have resolved more than 330,000 customer queries without human intervention as their combined interaction tallies topped 500,000.

Revealing the milestone at the company’s investor day, Superloop estimated that the AI agents, named Teddy and Mo, had allowed the telco to avoid around 320,000 service calls.

Delivering an update on Superloop’s AI strategy, chief commercial officer Nick Pachos said the agents’ capabilities matched those of its human staff.

The two agents, he said, were able to diagnose issues, initiate workflows and deputise to the telco’s technical support-focused AI agent Refreshify to resolve problems.

“Once a customer has been authenticated, there is effectively nothing Teddy and Mo cannot do that one of our service agents can do and the result has been powerful," Pachos said.

“We've now processed more than half-a million-customer interactions through Teddy and Mo.

"Approximately two-thirds of those interactions are fully resolved without any human involvement."

In addition, Pachos said that Refreshify had generated 400,000 customer interactions, allowing the telco to avoid a further 125,000 calls.

Processify, Superloop’s AI-driven automated provisioning system for new orders, has achieved a same-day activation rate of 75 percent with only 10 percent requiring human intervention.

Interest in AI across the telco sector is intensifying. Mobile and broadband service providers are leaning into the technology to find cost savings and other business dividends as competition increases, placing margins under pressure.

The challenges are particularly acute in the NBN market. Growth of the national network builder’s base has flattened to the point where retailers have entered a Hunger Games phase, only able attract new NBN customers by poaching from each other.

Early last week, TPG Telecom demonstrated its approach, with chief technology officer Giovanni Chiarelli unveiling that the carrier was using AI to monitor and maintain high customer sentiment scores.

Telstra has also demonstrated several use cases for AI and automation across its business, some also aimed at improving customer retention.

Economic case for AI under scrutiny

However, the conversation around the economics of AI has started to shift. Telstra in February said it would examine its AI investments closely to ensure that its costs don’t start to outweigh its benefit.

Commonwealth Bank chief executive Matt Comyn echoed that sentiment in remarks earlier this week at an Australian Financial Review event, saying that he wouldn’t be surprised if companies started to scrutinise their AI costs more closely.

AI technology providers, such as Anthropic, OpenAI and Google, charge for AI based on the volume of text that their models process. The model is to break strings of text down into tokens which carry a unit price to be processed.

For businesses, that means AI costs can be tied to the complexity and scope of their requirements and increase with the amount of information that the models they use ingest and process.

Using AI can be labour-saving, but the convenience of the technology can lead to unforeseen consequences, such as over-adoption – a risk that increases when the technology is used in customer-facing scenarios.

With no queues to navigate and the ability to access it through a smartphone, Superloop said that customer adoption of its self-serve support had been strong “which is what we would expect when support becomes easier and more immediate”.

However, a spokesperson for the telco said that increased engagement did “not necessarily” translate into more AI usage charges.

“In many cases, customers are interacting with automated workflows and diagnostic tools, rather than a large language model. This approach allows us to deliver, fast, simple resolutions, while applying AI only where it adds value,” the spokesperson said.

Furthermore, the telco said that the economic case for using AI would continue to hold as its customer base grew and regardless of changing in its demographic mix.

“The economic case for AI strengthens as we grow, and it does not depend on our customer mix remaining unchanged. That’s because our model was never built on the assumption that every interaction needs a large language model.

“Instead, we have invested in an orchestration layer that determines when AI is needed and when an issue can be resolved through automated workflows and existing systems.

"As a result, cost does not scale one-for-one with customer growth. They scale only with the relatively small proportion of interactions that genuinely benefit from AI,” the spokesperson said.

Pachos told investors and analysts that that the company established the technology foundation for its AI strategy five years ago when it swept out "myriad" legacy systems as part of a digital transformation. The foundation was designed to be AI model agnostic.

“We recognise that those models’ capabilities are improving faster than any individual organisation can keep up with. At the same time, the economics and capabilities of those models continue to change as competition increases and new entrants come to market.

“For that reason, we made a conscious decision not to optimise our strategy around any single model provider or token structure,” he said.