Learn how Palo Alto Networks is Transforming Platform Engineering with AI Agents. Register here
Register now
Optimize compute, storage and data
Choose copilot or autopilot execution
Continuously improve with reinforcement learning
At autocon we held a panel to share in-depth insights on the business case for AI and autonomous systems, tackling some big questions about autonomy and AI:
The panel was comprised of industry leaders who are playing a key role in shaping today’s tech infrastructure:
Read on for the perspectives our panelists shared.
Q: Matt (Northwest Venture Partners): PayPal is a monster organization. It's probably constantly under cyber attacks. How does PayPal handle the volume, and how is it using autonomy today?
Sri (PayPal): Our business case for autonomy started with a need for very high reliability. We found ourselves in a situation where we thought we had great reliability and uptime. But our customers would come in and tell us, “You're nowhere close to where you're telling us you are.”
We were measuring uptime based on whether the servers were up. It is called “ATB” or the availability to do business. We had a bit of a wake up call when we started to measure reliability from a customer’s perspective.
The first time we measured it, we thought our uptime was 99.95%, but when we saw that from the outside in, it was less than three 9s.
That’s when our journey started. It took us through a level of automation, and eventually led us into autonomy to a point where we have not only been able to achieve four 9s, but now we achieve about 200 days of five 9s in a year.
Q: Matt (Northwest Venture Partners): How do you handle the risk of using autonomy? How willing are you to think about the co-pilot, and what is your willingness level to go fully auto?
Sri (PayPal): We use standard approaches to manage risk. We do a lot of testing ahead of launching it across our whole production fleet. We roll it out in one availability zone and on a small fraction of our servers at a time to build confidence. Then, we use that confidence to roll it out everywhere. It's called a “release strategy". You can also call it a canary strategy. Lastly, there is a place for the co-pilot, and there is a place for the autopilot. In fact, we are using autopilot for autonomy in infrastructure.
Matt (Northwest Venture Partners): Was your staff feeling threatened when it came in? What was the initial reaction?
Sri (PayPal): Not at all actually. They were more than happy to give up the kind of work which had a high toil and low satisfaction. In fact, it allowed the staff to free themselves up and move to more meaningful work, so they could actually deliver better to the customers.
Q: Matt (Northwest Venture Partners): We know RingCentral is extremely efficient. Was autonomy and AI pushed down from the board in RingCentral, or was it something you pushed up to the board? What were the dynamics when you brought autonomy into your organization?
Kira (RingCentral) (RIngCentral): It was both ways. From the board, you get, “what are you doing about AI?” But from within, there's a need for autonomy that continues to grow. Especially for us, as we are running the service at 5/9ths. So, the amount of overhead without autonomy is immense and the risk is also high. Autonomy is important in terms of cost, but even more for reliability. The optimization that autonomy provides in the product itself, or in the services behind the product, is also super important. So that is really where more of the pull and push within the organization takes place.
Matt (Northwest Venture Partners): How big of a topic is autonomy at the board level? Are they asking more about it or are you telling them more about it?
Kira (RingCentral): We're telling them about it. There's an expectation that it's inserted into every element of the process because it is expected to help with everything we do.
Matt (Northwest Venture Partners): You also acquired a company around AI autonomy recently. Can you tell us a little bit about that, and what were the dynamics around that?
Kira (RingCentral): A couple of years back, we bought a company that provided us with a very early natural language understanding and models that could transcribe every single conversation that took place on the RingCentral network. For clarity, we are talking about billions of minutes, both voice and video. We put that into a product that we made available to the market. It can be applied to revenue intelligence, quality intelligence, contact center - and every single conversation that takes place.
Q: Matt (Northwest Venture Partners): Palo Alto Networks is one of the top cybersecurity companies on Earth. The industry is very unique because it is constantly under attack. What is Palo Alto Networks thinking about autonomy?
Suresh (Palo Alto Networks): Palo Alto has gone through a transformation into the cloud, but autonomy is not uniformly distributed in everything we do. There certainly is AI in our product in terms of finding out new vulnerabilities, creating new abilities for our internal testing and so on. In terms of risk, we have built these co-pilots and rolled them out internally to our support workforce so they can use them to answer customers’ questions. For now, it is in an augmented mode, and I don't think we are ready to go fully auto yet, given what we've seen with the latest technologies for autopilot.
Q: Matt (Northwest Venture Partners): How do you view autonomy, and measure that balance of being aggressive and not aggressive? Also, you started to roll out autonomous IT functions on serverless. Are you thinking about expanding it?
Suresh (Palo Alto Networks): We are definitely thinking of expanding (autonomous IT functions on serverless). The partnership with Sedai has been very interesting. Sedai, in its current form, is applicable to parts of what we do and there also are parts where it doesn't apply yet. In fact, I was happy to see on the roadmap that Sedai is extending to GCP Dataflow, GCP VMs, and so on. So it'll become more relevant and I expect that we'll grow.
I will also explain with an example why autonomy is not yet possible in parts of our work. We've been involved in about five acquisitions. The backends of these companies have to integrate with us and present a single console - a single pane of glass to our customers, so that all the complexity of these backends coming together gets hidden.
Stitching these backends together and running them as one uniform service means that not everything we get from vendors can apply as is. In some cases, we build a layer of intelligence and development on top of what we get from vendors. In others, we go deeper. In partnership with Sedai, we can build on what we already have.
Q: Matt (Northwest Venture Partners): Every company is now complaining about the costs of cloud infrastructure. At the same time, we're being asked to roll out AI. How much cost pressure is there and how are you handling the costs of this new AI infrastructure?
Suresh (Palo Alto Networks): Cost is an important factor. As we first try to apply automation internally, we can easily measure the improvements in efficiency and a clear ROI.
In the case of co-pilots that we have applied, our support teams are getting more efficient. A lot of our contract workforce handles first-level cases, which are the easiest ones to solve, so we can easily see a path to shrinking that staff. I believe that in all of these cases, especially internal use cases, we can get the desired ROI.
As far as external use cases are concerned, I think it's a standard product market fit problem. If we use AI to do stuff that people don't really care about, then they won't pay for it, and then it's not going to pay for itself. So we'll have to build products that add real value so we can charge a premium for it.
Kira (RingCentral): I second everything that Suresh (Palo Alto Networks) said. It's all very similar in our case. The cost is incurred but then it translates into internal savings in labor, etc. AI has opened new opportunities for us that weren't there before. The adoption in the market is moving much faster than we've ever seen.
Sri (PayPal): In terms of an investment framework, we've taken a two by two - “internal and external” on one axis and “assist and act” on the other. The value of investment is no different than any other technology. Either it has to have a productivity multiple or a customer metric multiple that pays for it. The economic framework is the same as those used in the past. AI doesn't present any new method of evaluation for that. It is expensive but this is where we are also diversifying by using vendor-only solutions from the big players like Google, Microsoft, and OpenAI.
Q: Matt (Northwest Venture Partners): How stressed are your DevOps and SRE teams?
Sri (PayPal): Our DevOps and SRE teams are not any more stressed than they've been.
Kira (RingCentral): Our teams are definitely less stressed than they were a decade ago. Without autonomy I don't know how we'd do the work we do today as we are running at super high availability, Every company is now asked to be more efficient. That just means more autonomy. It's almost synonymous for me.
Suresh (Palo Alto Networks): We have seen a lot of growth, so I'd say our team is stressed because we are not fully autonomous. The good news for us is that the Sedai approach has made clear that certain things are possible.
Q: Matt (Northwest Venture Partners): Some companies have customer-facing AI and they are not sure about how much uplift they can get in revenue. Is there a foregone conclusion that you can increase your ASPs or is the jury still out?
Suresh (Palo Alto Networks): “Companies will definitely see an uplift (in revenue) with value creation.”
Kira (RingCentral): "It's still early but we expect an uplift. We are all willing to accept loss to some extent assuming that ASP will go up as we create more value."
Sri (PayPal): "The value addition from AI will allow us to pass it on to the customer in some cases because the value will outweigh the price. In other cases, the cost that is incurred can be offset with something else within the enterprise and not be passed on to the customer. In payments, it's a race to zero. You can keep increasing the prices and you will lose to the competition very quickly.”
The panel discussion revealed several important insights into how leading companies are approaching AI and autonomy. These industry leaders shared their experiences, challenges, and strategies for leveraging autonomous systems to drive business value. Here are the key themes that emerged:
October 3, 2024
November 25, 2024
At autocon we held a panel to share in-depth insights on the business case for AI and autonomous systems, tackling some big questions about autonomy and AI:
The panel was comprised of industry leaders who are playing a key role in shaping today’s tech infrastructure:
Read on for the perspectives our panelists shared.
Q: Matt (Northwest Venture Partners): PayPal is a monster organization. It's probably constantly under cyber attacks. How does PayPal handle the volume, and how is it using autonomy today?
Sri (PayPal): Our business case for autonomy started with a need for very high reliability. We found ourselves in a situation where we thought we had great reliability and uptime. But our customers would come in and tell us, “You're nowhere close to where you're telling us you are.”
We were measuring uptime based on whether the servers were up. It is called “ATB” or the availability to do business. We had a bit of a wake up call when we started to measure reliability from a customer’s perspective.
The first time we measured it, we thought our uptime was 99.95%, but when we saw that from the outside in, it was less than three 9s.
That’s when our journey started. It took us through a level of automation, and eventually led us into autonomy to a point where we have not only been able to achieve four 9s, but now we achieve about 200 days of five 9s in a year.
Q: Matt (Northwest Venture Partners): How do you handle the risk of using autonomy? How willing are you to think about the co-pilot, and what is your willingness level to go fully auto?
Sri (PayPal): We use standard approaches to manage risk. We do a lot of testing ahead of launching it across our whole production fleet. We roll it out in one availability zone and on a small fraction of our servers at a time to build confidence. Then, we use that confidence to roll it out everywhere. It's called a “release strategy". You can also call it a canary strategy. Lastly, there is a place for the co-pilot, and there is a place for the autopilot. In fact, we are using autopilot for autonomy in infrastructure.
Matt (Northwest Venture Partners): Was your staff feeling threatened when it came in? What was the initial reaction?
Sri (PayPal): Not at all actually. They were more than happy to give up the kind of work which had a high toil and low satisfaction. In fact, it allowed the staff to free themselves up and move to more meaningful work, so they could actually deliver better to the customers.
Q: Matt (Northwest Venture Partners): We know RingCentral is extremely efficient. Was autonomy and AI pushed down from the board in RingCentral, or was it something you pushed up to the board? What were the dynamics when you brought autonomy into your organization?
Kira (RingCentral) (RIngCentral): It was both ways. From the board, you get, “what are you doing about AI?” But from within, there's a need for autonomy that continues to grow. Especially for us, as we are running the service at 5/9ths. So, the amount of overhead without autonomy is immense and the risk is also high. Autonomy is important in terms of cost, but even more for reliability. The optimization that autonomy provides in the product itself, or in the services behind the product, is also super important. So that is really where more of the pull and push within the organization takes place.
Matt (Northwest Venture Partners): How big of a topic is autonomy at the board level? Are they asking more about it or are you telling them more about it?
Kira (RingCentral): We're telling them about it. There's an expectation that it's inserted into every element of the process because it is expected to help with everything we do.
Matt (Northwest Venture Partners): You also acquired a company around AI autonomy recently. Can you tell us a little bit about that, and what were the dynamics around that?
Kira (RingCentral): A couple of years back, we bought a company that provided us with a very early natural language understanding and models that could transcribe every single conversation that took place on the RingCentral network. For clarity, we are talking about billions of minutes, both voice and video. We put that into a product that we made available to the market. It can be applied to revenue intelligence, quality intelligence, contact center - and every single conversation that takes place.
Q: Matt (Northwest Venture Partners): Palo Alto Networks is one of the top cybersecurity companies on Earth. The industry is very unique because it is constantly under attack. What is Palo Alto Networks thinking about autonomy?
Suresh (Palo Alto Networks): Palo Alto has gone through a transformation into the cloud, but autonomy is not uniformly distributed in everything we do. There certainly is AI in our product in terms of finding out new vulnerabilities, creating new abilities for our internal testing and so on. In terms of risk, we have built these co-pilots and rolled them out internally to our support workforce so they can use them to answer customers’ questions. For now, it is in an augmented mode, and I don't think we are ready to go fully auto yet, given what we've seen with the latest technologies for autopilot.
Q: Matt (Northwest Venture Partners): How do you view autonomy, and measure that balance of being aggressive and not aggressive? Also, you started to roll out autonomous IT functions on serverless. Are you thinking about expanding it?
Suresh (Palo Alto Networks): We are definitely thinking of expanding (autonomous IT functions on serverless). The partnership with Sedai has been very interesting. Sedai, in its current form, is applicable to parts of what we do and there also are parts where it doesn't apply yet. In fact, I was happy to see on the roadmap that Sedai is extending to GCP Dataflow, GCP VMs, and so on. So it'll become more relevant and I expect that we'll grow.
I will also explain with an example why autonomy is not yet possible in parts of our work. We've been involved in about five acquisitions. The backends of these companies have to integrate with us and present a single console - a single pane of glass to our customers, so that all the complexity of these backends coming together gets hidden.
Stitching these backends together and running them as one uniform service means that not everything we get from vendors can apply as is. In some cases, we build a layer of intelligence and development on top of what we get from vendors. In others, we go deeper. In partnership with Sedai, we can build on what we already have.
Q: Matt (Northwest Venture Partners): Every company is now complaining about the costs of cloud infrastructure. At the same time, we're being asked to roll out AI. How much cost pressure is there and how are you handling the costs of this new AI infrastructure?
Suresh (Palo Alto Networks): Cost is an important factor. As we first try to apply automation internally, we can easily measure the improvements in efficiency and a clear ROI.
In the case of co-pilots that we have applied, our support teams are getting more efficient. A lot of our contract workforce handles first-level cases, which are the easiest ones to solve, so we can easily see a path to shrinking that staff. I believe that in all of these cases, especially internal use cases, we can get the desired ROI.
As far as external use cases are concerned, I think it's a standard product market fit problem. If we use AI to do stuff that people don't really care about, then they won't pay for it, and then it's not going to pay for itself. So we'll have to build products that add real value so we can charge a premium for it.
Kira (RingCentral): I second everything that Suresh (Palo Alto Networks) said. It's all very similar in our case. The cost is incurred but then it translates into internal savings in labor, etc. AI has opened new opportunities for us that weren't there before. The adoption in the market is moving much faster than we've ever seen.
Sri (PayPal): In terms of an investment framework, we've taken a two by two - “internal and external” on one axis and “assist and act” on the other. The value of investment is no different than any other technology. Either it has to have a productivity multiple or a customer metric multiple that pays for it. The economic framework is the same as those used in the past. AI doesn't present any new method of evaluation for that. It is expensive but this is where we are also diversifying by using vendor-only solutions from the big players like Google, Microsoft, and OpenAI.
Q: Matt (Northwest Venture Partners): How stressed are your DevOps and SRE teams?
Sri (PayPal): Our DevOps and SRE teams are not any more stressed than they've been.
Kira (RingCentral): Our teams are definitely less stressed than they were a decade ago. Without autonomy I don't know how we'd do the work we do today as we are running at super high availability, Every company is now asked to be more efficient. That just means more autonomy. It's almost synonymous for me.
Suresh (Palo Alto Networks): We have seen a lot of growth, so I'd say our team is stressed because we are not fully autonomous. The good news for us is that the Sedai approach has made clear that certain things are possible.
Q: Matt (Northwest Venture Partners): Some companies have customer-facing AI and they are not sure about how much uplift they can get in revenue. Is there a foregone conclusion that you can increase your ASPs or is the jury still out?
Suresh (Palo Alto Networks): “Companies will definitely see an uplift (in revenue) with value creation.”
Kira (RingCentral): "It's still early but we expect an uplift. We are all willing to accept loss to some extent assuming that ASP will go up as we create more value."
Sri (PayPal): "The value addition from AI will allow us to pass it on to the customer in some cases because the value will outweigh the price. In other cases, the cost that is incurred can be offset with something else within the enterprise and not be passed on to the customer. In payments, it's a race to zero. You can keep increasing the prices and you will lose to the competition very quickly.”
The panel discussion revealed several important insights into how leading companies are approaching AI and autonomy. These industry leaders shared their experiences, challenges, and strategies for leveraging autonomous systems to drive business value. Here are the key themes that emerged:
.svg)
.svg)
