[0:10] Matty Sirois: Good morning, good afternoon, good evening, wherever you're joining us. Thank you for joining our webinar today. I'll just let a few people come in through the waiting room and then we'll get started. I can see a few people jumping in.
[0:26] If you're joining us from anywhere in the world because we're doing a few different locations, you can always drop in the comments or in the chat where you're joining us. We got a few from Singapore, over in the UK as well in London, down here in Sydney, which is where we're currently based, which is quite good. Fully global.
[0:57] Thanks everyone for joining us today for "Understanding AI Agents: Your Key to MVNO Differentiation in 2025." Before we kick things off, I'll just go through a few housekeeping slides.
[1:12] Just so you guys know, this webinar is being recorded, so we will send out the recording after today's session, along with a few different relevant resources for you to go through. We have a new ebook you guys have got access to as well as a few different articles.
[1:37] If you do have any kind of tech questions during the webinar - if you can't see anything, if you can't hear anything - our admin can help out through the chat window. Best case, usually log in, log out, do a little reset usually works, and worst case, you will get the recording, so no stress there.
[1:54] If you do have any questions throughout the webinar, you can use the Q&A there at the bottom as well. We will do a Q&A section at the very end, so please drop in your questions. If we don't get to them, we will include that in our FAQ doc following today's presentation.
[2:10] To introduce myself: My name is Matty Sirois, I am the marketing director here at Pendula. I'm essentially just your MC. The real hero to this is Patrick Tang, who I'll now hand over to for his introduction.
[2:11] Patrick Tang: Thanks very much, Matty. Thanks again everyone for joining. As you can read there for yourself, Patrick Tang, the head of solutions, consulting and professional services here at Pendula, also based in Sydney with Matty. And we'll just get straight into it.
Agenda
[2:30] So today we got a relatively simple agenda:
- First, we'll spend a few minutes just going over AI agents 101, specifically why they matter for MVNOs
- Then we're gonna go through a few examples of use cases for agents - something a bit more concrete that we've discussed with a number of our customers and we're in the process of implementing
- Finally, we'll run through a quick demo experience of what talking to one of these AI agents is gonna look like for your customers, speaking on their handset
[3:08] I'll briefly hand back over to Matty after that, before we finish up with the Q&A. As Matty said, please feel free to put your questions into the chat, and we will answer them at the end of the session.
What are AI Agents?
[3:27] AI agents are a way of helping you automate a lot of those business processes, customer interactions, whatever they may be, that take up a lot of the time of the various teams within your organization, within your MVNO. That could be anything from a customer service agent to internal questions and everything in between.
[3:57] One of the key things that an AI agent is able to do is actually take the next best action. So it's not just like a chatbot where you're just endlessly talking to it and at best it's just gonna maybe send you a link to some information (although of course it can do that). It's actually able to decide what your intention is, what you're actually asking for, and do its best to try and actually complete your query or help you with whatever you're asking it to do.
[4:30] As part of that, an AI agent is integrated into all of the various systems within your MVNO as well. Whether that's the OSS and BSS, it could be other information such as your CRMs, or it could be other external data sources as well. You may have a data warehouse, you may have a support ticketing system, a separate support ticketing system.
[5:06] It's continuously hooked into all of those systems, which then allows it to do that final point on the right-hand side here, which is proper personalisation at scale.
[5:35] So what I mean by that is not just being able to send a message that's got "Hi, Patrick" or "Hi, Matty" as the salutation. It actually understands you as a customer when you're talking to it. So it's gonna understand what plan you're on, it's gonna understand where you're from potentially, ideally what handset you have, whether you have previous cases open or existing cases open with us, what your sentiment's likely to be based on the resolution of those cases.
[5:35] All of that is fed into the AI agent, and when it makes those decisions for next action and how to respond, it's going to take that into account.
Why are Agents important for MVNOs?
[5:49] For a lot of the MVNO customers we talk to, they operate in a very crowded market, and it's very, very hard to differentiate themselves. A lot of MVNOs have obviously defaulted to price, which is just a race to the bottom.
[6:07] But having an AI agent really allows you to differentiate yourself to your customers. The agent is going to understand the target market that you're going after. And as I mentioned, it's able to provide a much more personalized experience. So when somebody calls in or messages in for assistance, or whatever it might be, it's already got a lot of the context there that you don't have to start by playing 20 questions before you can even tell the customer service agent what you're after.
[6:41] Obviously a big part of this is that these agents run in the background all of the time whenever it makes sense for your customers, and what that allows you to do is scalably engage your customers without also scaling up your costs as well.
[6:57] It's able to reach out at the right point in time on the right channels to your customers for things such as upsell opportunities. Maybe there's a new iPhone coming out, or maybe the customer's used up all of their data for the month, or maybe based on our information, we suspect the customer might be going overseas, so maybe we'll offer them a roaming pack or something like that.
[7:23] And similarly, it really allows you to save on your operations cost as well.
[7:30] Again, based on our interactions with our customers, a lot of the support calls that come in follow the classic 80-20 rule. 80% of the queries are more or less of the same nature, and a lot of them are relatively simple to fix. It could be things such as resetting passwords, it could be changing addresses, updating information - all of that stuff that really doesn't take a human and their intelligence to do.
[7:59] An AI agent is able to fully take that from the customer query all the way through to the end of updating the OSS or BSS, for example.
[8:11] And obviously being agents, they're 24/7. So you don't have to worry if a customer is calling in at 2 AM in the morning, for whatever reason - maybe after a bit of a big night on Friday - the AI agent is still ready to go.
Implementation strategy: Crawl, walk, run
[8:33] Speaking with a lot of our customers, depending on their maturity both as an organisation and their maturity with AI in general, they may have different comfort levels on how they want to implement AI. A lot of the customers aren't yet ready to unleash the AI to go and talk to their customers without some kind of human or operational oversight.
[9:02] So for a lot of our customers, we propose this classic crawl, walk, run strategy.
[9:03] For crawling, we suggest that you provide the AI agents as an additional capability for your internal operations team, so it could be for your support team, for example.
[9:14] The agent is there, it's connected into all of the previous support cases, it knows your FAQ, it knows your systems, it knows the common problems and resolutions.
[9:25] Instead of an agent speaking with a customer having to put them on hold, going to talk to an expert, or maybe looking something up in a wiki or some kind of data repository, the agent can quickly put it into the AI agent. The AI agent tells the agent exactly what information they need to convey back to the end customer, and you can move on with your cases in a much quicker manner, and hopefully achieve that first call resolution.
[9:55] In the walk kind of scenario, it's somewhere in between the two that I've talked about.
[10:01] For the walk, you still have the AI agent making the decisions in the backend, but you've essentially approved all of the responses or copy that the agent is going to provide back to the customer. And what this allows you to do is keep the agent on some very fixed guardrails. The agent can only respond with specific paragraphs or words that you have already approved.
[10:30] Outside of that, it's not going to be making things up, and you don't have to worry about things such as hallucinations.
[10:38] Finally, for those customers who maybe have gone through the crawl and walk, or maybe just the ones that are a little bit more mature and a bit more comfortable around AI agents and AI capability in general, they can let the AI hopefully take everything from where to go.
[10:49] So the AI agent is talking to the customer, it is making up the copy, in this case it decides what it needs to do - whether that is respond to the customer or actually take action, such as changing people's plans, charging their credit cards (with approval from the customer), and anything like that.
Real-world use cases
[11:21] So let's go through some real world examples of how these agents can potentially help as well.
[11:32] The first one very much does talk to that kind of crawling approach where you've got an agent talking to an operations agent directly internally with your staff. As I said, it's got all of the context it needs - most importantly, those historical tickets that are in your organisation. So it knows what the common problems are, what people call in about, and what, more importantly, how they've been assessed and resolved previously.
[12:08] It also obviously understands everything about this customer. So maybe we know this customer has been with us for 12 months, and when they signed up they signed up for a new deal for the latest iPhone.
[12:21] It's able to pull up all of the relevant information associated with the person as well. So when the person's asking the agent for whatever it needs, the AI is supplementing all of the information so the agent can provide the best possible experience for the end user.
[12:44] The second one is around 24/7 support. Not all MVNOs have the resources to run a 24x7 call center - that can be quite an expensive and investment-heavy kind of thing to do. Most of the ones we talk to have a business hours call center. But at the same time, obviously, people don't stop using their services outside of business hours.
[13:11] What this allows you to do is have the agent talk to your customers. So you could advertise a QR code, for example, for them to text in to talk to an agent, or maybe they could call in and they'll get the standard out-of-hours response, but they can opt for a call or a message directly from an AI agent to try and sort them out if it's after hours.
[13:35] You can see on there an example of the interaction where we're telling them that the agents will be back tomorrow at 9, but let's see if the AI can help you. If it's a simple query, the AI can maybe help you out if it's a billing issue.
[13:49] Or if the AI is determined it's not able to help, at the very least it can schedule a callback. And just like you would with a real person, it's not just "we'll call you tomorrow." It's going to potentially ask you when's a convenient time for you, and that can be as granular as you want. That can be something as simple as "give me a call in the morning or afternoon," or it could be specific around "give me a call next Friday at 3:30 PM."
[14:19] Here is an example that we're gonna run through in a minute with the demo, where we've got an agent specifically trained for eSIM adoption and support. Again, speaking with a lot of our MVNO customers, while eSIMs are a fantastic technology, there's a lot of people that still haven't had an experience with them. Especially the first time they try and use an eSIM, they get lost and don't really know what to do.
[14:45] So we've created an agent that basically allows an AI to help your customers onboard with an eSIM. It can walk them through the steps needed to activate the eSIM on their device, and it can even help them troubleshoot if they run into problems.
Live demo: eSIM support agent
[15:06] Let's see the demo in action. What I've done here is I've mirrored my phone onto my computer. And in this case, I am gonna use WhatsApp to just talk with an AI agent. The theory is maybe that I've landed in a new country, I've just bought an eSIM when I've landed. Obviously, I don't have any service, so I'm not gonna be able to communicate with them via SMS, for example. But hopefully they're gonna have WiFi while they're at the airport or something, and I can initiate a conversation with them as you're about to see here.
[15:47] So they've received the email telling them that they've purchased their eSIM. But this person, as I said, doesn't know what to do. In the email, there's instructions to trigger this message to come from the AI agent.
[16:04] And as you can see here, our mythical Down Under Mobile MVNO has reached out to them, and the customer's asked the AI how can I activate the eSIM essentially.
[16:16] You'll see that they can respond in natural language - there's no prompting or anything like that.
[16:22] The first thing that we need to do here is identify what device the customer is actually using. So I'm gonna go here - I have a slightly older iPhone, but hopefully I'm up to date with my iOS.
[16:34] The agent has been provided with all of the information on all of the latest handsets and operating systems, so it's gonna give you specific instructions based on your responses here.
[16:45] I'm gonna quickly scan this. I don't need you to read all of it, but I can see on point 3, it's actually asking me to scan a QR code. In my case, however, I dropped my camera while I left the airplane - I'm sorry, I dropped my phone when I left the plane and I've broken the camera, so I'm not able to follow these instructions. So let's see what the AI is gonna tell me to do.
[17:09] So it's gone back and actually looked at alternate instructions. It's told me actually I can manually go ahead and enter, instead of using the QR code, all the information needed.
[17:20] So I'm gonna go ahead and add an eSIM. As I said, this is just a demo running on my phone, so you'll see here it's gonna say I can't use the camera (i.e., broken), but it does allow me to enter in any of the required information. We unfortunately don't have our own MVNO Pendula, so when I hit next, this isn't going to activate, but pretend it does.
[17:44] I can then go back to WhatsApp and let the AI know that I've completed the steps.
[17:56] And hopefully, once that's done, I'll be getting reception. I'll let the AI know that the eSIM is now activated, and I can continue with my holiday or whatever it is and hopefully make use of the eSIM.
[18:17] And with that I'll briefly hand back over to Matty.
Discovery workshops
[18:22] Matty Sirois: Thank you so much, Pat. That was great. It's always good seeing an agent in action.
[18:27] Just before we jump into the Q&A section, I will just mention we are running AI agents discovery workshops. Essentially what these are are 60-minute sessions with your team and our expert services team to run through some use cases you guys may have. If you're not sure where to start with AI agents, it's a great kind of free session to jump in. We can do some exploratory stuff, figure out where your pain points are, where in your customer journey can be improved with agents.
[18:51] We've had a lot of success with those. A lot of people found it very beneficial just to go in there and have a whole bunch of questions, or they know where their pain points are and they want to see where agents can get launched.
[19:00] So reach out afterwards - we'll send out those resources and the link to our discovery workshop submissions where you can put your entries in there.
Q&A session
[19:09] And now we'll jump over to some Q&A. I did see a few questions come in. I'll start off. There's actually 2 questions that were pretty similar, so I'm gonna merge them for you, Pat. There's 1 around how do AI agents handle sensitive billing disputes that kind of like privacy issue that we always hear around agents.
[19:28] Patrick Tang: In that regard, basically all of the AI agents that we provide customers are based on their own individual AI accounts. And what that means is that any data that's fed into any of the AI agents remain private and unique to you.
[19:47] So none of it's going to be used to train any kind of AI models. It's never going to be surfaced in other queries. If you go out there and use ChatGPT and ask it a question, it's not going to repeat what some of your customers have said or anything like that. You can rest assured that everything is safe and secure, and of course, leveraging all of the wonderful certifications that both Pendula and our sub-processes have, such as SOC 2 and ISO 27001.
[20:18] Matty Sirois: Beautiful stuff. There's another one around what happens if an AI agent can't answer my questions or any of my queries like reaches a dead end.
[20:27] Patrick Tang: Unfortunately, of course, all technology has its limitations. We can't guarantee that the AI is gonna be able to solve all the problems, but of course you want to be able to make sure your customers' queries are resolved.
[20:40] So generally speaking, with the AI agents, they have instructions in regards to determining whether it can or can't help a person. And even if it thinks it can, we obviously don't want it sitting there endlessly asking the customer questions. People will get frustrated and then sick of it after generally, you know, 3, 4, or 5 interactions.
[21:01] So there's always the capability to hand off back to a physical agent. That could be someone actually directly asking "I want to talk to someone" or the AI making that determination.
[21:14] And on the back of that, what you can do is have the AI agent put the person directly into the queue for the call center for somebody to talk with them next, or it could be something such as raising a ticket within your ticketing system or sending an email, or all of the above, or anything in between as well, just to make sure the agent has identified that it can't help, but then a real person is gonna reach out to the customer or connect with them straight away.
[21:46] Matty Sirois: Very nice. I'll do one more because I've seen it come a few times, and we've gotten it even before this webinar - pricing for Pendula agents, pricing model, pricing structure, anything around that.
[21:56] Patrick Tang: So I run the solutions consulting team and professional services. I try not to get involved in pricing, but I believe all of the AI agent pricing is on our website at pendula.com.
[22:08] Matty Sirois: I can confirm that. So if you head to our website, you can see the dropdown there for AI agents, and there's a pricing tab in there.
Closing
[22:35] Beautiful stuff. For any other questions, please do submit them. I think that about covers the big ones that I've seen. We will be sending an FAQ doc if there's anything else that we've missed after this webinar.
[22:35] So yeah, otherwise, thank you guys so much. We appreciate you taking the time wherever you were joining us. Have an amazing day, morning, evening, and we'll speak to you soon. Thanks so much.
.webp)
.webp)
