Voice is becoming an increasingly part of Mavenir’s AI strategy, with the company looking to turn the mobile network into more than just the infrastructure that connects a call.
Just yesterday (26 August 2026), the software company partnered with Sanas, a real-time Speech AI company, to bring AI-powered voice capabilities directly into its MAVcore Voice AI portfolio.
As a result, the partnership will allow mobile operators to offer real-time translation, speech enhancement and clearer voice communication via their networks, helping people communicate across different languages, accents and environments.
The partnership will also focus on protecting voice communications from emerging threats such as deepfakes and voice fraud.
For Mavenir president and CEO Pardeep Kohli, it is a natural extension of a business that has been working on voice since the company was founded in 2005.
“Voice has always been our bread and butter, even from day one,” Kohli tells Mobile Europe.
“We started when people were going from 3G to 4G. We started working on Voice over LTE and Voice over Wi-Fi because, prior to 4G, everything was circuit-switched.”
Mavenir now has more than 300 customers, with its software used across voice, messaging and mobile core networks.
Bringing translation into the AI network
Sanas has spent years working on speech technology in the contact centre market.
According to Keshava Narayana, Co-Founder and CEO of Sanas, communications networks have “connected people for decades”.
He says: “We believe the next chapter is helping people understand one another more clearly.”
Therefore, Mavenir wants to take that technology into the consumer market.
“When you’re doing a live translation in the consumer world, it’s actually happening in the network, and you’re dealing with millions of people,” Kohli reveals.
Kohli adds Sanas brings years of experience and large datasets covering different accents, environments and noise levels.
“They have trained their models with different data sets for their application. We don’t have to redo that work because the work has already been done.
“What we bring to the table is, how do you scale it to millions of users? How do you deploy an operator network, rather than a contact centre?”
There are also telecom-specific requirements to consider, including emergency services and lawful interception.
“More important is all the other issues like emergency services. That doesn’t happen in a contact centre,” he notes.
More than just translation
Translation is an obvious application for the technology, particularly for families where different generations speak different languages.
Kohli uses his own family as an example. His children do not speak Hindi fluently, while his mother is not completely comfortable speaking in fluent English.
A real-time translation service could make those conversations much easier, but translation is only one part of what Mavenir is exploring.
The company is also looking at AI-powered voice enhancement, including removing background noise from calls, which Kohili explains could be particularly useful in emergency situations.
“If you are in a very noisy environment, let’s say if there’s a big crash and you’re calling emergency services, but if you take the noise out people can hear better what’s going on,” he says.
The company is also looking at whether AI can help deal with the effects of poor network coverage, as well as help operators tackle fraud.
Mavenir already has products aimed at spam and messaging security, but Kohli expects fraudulent voice calls to become an increasingly serious problem as criminals make better use of AI.
A bigger role for the mobile core
For years, operators have invested in each new generation of mobile technology, from 1G all the way through to 5G.
But much of the value created has ended up with companies building applications and services on top of those networks.
“We went from 1G to 5G, and every generation, more value got created, but not by operators themselves,” he explains.
As a result, Kohli believes AI gives operators another opportunity to change that.
“We are trying to help in this transition to AI, so operators can bring more value,” he notes.
While the radio access network provides the connection between the phone and the network, the core is responsible for the services that sit behind that connection.
“RAN is just taking coverage for the radio signal so that your phone gets access. It’s connectivity. It’s access to the radio network. That comes from the core and without this core, we simply can’t do anything,” he says.
Operators have a choice
Kohli believes the telecoms market could increasingly divide between operators that remain satisfied with providing connectivity and those that use AI to develop new services.
He explains: “There’ll be a set of players who will just be happy with just doing connectivity. They will just give you access on your phone, and that’s all they do.”
“But there’ll be another set of players who will try and experiment with AI transformation. They will ask themselves how they can bring more value to my customers?”
Mavenir is already exploring what that could look like, from voice translation and fraud prevention to AI assistants.
Kohli imagines an AI assistant that could sit alongside a conversation and help a user understand what is happening, dubbing it as a customer’s “own private assistant”.
Operators could also have a role in providing the infrastructure needed for AI itself.
They already have data centres, network infrastructure and a physical presence close to customers. That could become valuable as AI applications increasingly demand low latency.
Making AI feel natural
Despite the broader ambitions around AI, Kohli returns to one fundamental challenge: voice must feel natural.
“Voice is actually a very difficult application in the sense that it must flow between two humans. If you don’t have a good experience, you immediately known.
“To do a real-life conversation where we feel natural doing it, and to involve AI in between, so that when I’m speaking and you’re hearing, the language changes or accent changes, things like that, it’s not an easy application to do.
“Our experience working with Sanas is that they have the datasets and the experience of doing it in many different use cases, with different types of populations.
“We’re taking advantage of all of that and bringing it to the millions of users we have through our customers,” he concludes.
IN RELATED NEWS
