Subscribe
Share
Share
Embed
AI is changing how we work, but the real breakthroughs come when organizations rethink their entire foundation.
This is AI-First, where Box Chief Customer Officer Jon Herstein talks with the CIOs and tech leaders building smarter, faster, more adaptive organizations. These aren’t surface-level conversations, and AI-first isn’t just hype. This is where customer success meets IT leadership, and where experience, culture, and value converge.
If you’re leading digital strategy, IT, or transformation efforts, this show will help you take meaningful steps from AI-aware to AI-first.
Stanley Toh (00:00):
Now with ai, if you collaborate more closely with the business unit, understand what are their needs, what are they trying to solve, and then you bring to the table the AI technology that is able to help them solve those problem, streamline their workflows, be it in logistics, be it in finance, even hr, and help them transform.
Jon Herstein (00:34):
This is the AI first podcast hosted by me, John Stein, chief Customer Officer at Box. Join me for real conversations with CIOs and tech leaders about re-imagining work with the power of content and intelligence and putting AI at the core of enterprise transformation. Hello and welcome to AI First Leadership, the show where we explore how AI is reshaping the future of content, enterprise workflows and innovation. My guest today is Stanley Tow, head of Enterprise End User Services and experiences at Broadcom. And I'm your host, John Hurst, chief Customer Officer. At Box each episode we talk with CIOs, IT leaders, and AI pioneers to share real world insights you can act on today. Let's dive into today's conversation. Hello Stanley.
Stanley Toh (01:25):
Hi John. Thanks for having me.
Jon Herstein (01:27):
Of course. It is a pleasure. We've been working together for many, many years and it's very exciting. Yes, it's very exciting to be talking with you today about AI and Broadcom and Box. So maybe we'll start with a little bit of background. So if you could share with folks a bit about Broadcom, a very large and I think well-known company, but some people may not know the full background. So if you wouldn't mind sharing a bit about Broadcom and then your role at the company.
Stanley Toh (01:55):
Okay, sure. Broadcom is a infrastructure technology company. We are both a semiconductor and also a software company. We started as a semiconductor. We do anything that connect anything from your end devices like your laptop and your mobile phone through all the infrastructure, through the fiber to all the switches to the back of to the data center. And now after acquiring ca, Symantec and recently VMware, we are also a software company. So now we are also helping you managing all your data center, your infrastructure, your monitoring your network, basically connecting from your front end devices to your back infrastructure and protecting it at the same time.
Jon Herstein (03:02):
Maybe you can describe kind of your role when we say enterprise end user services and experiences. What exactly does that mean for folks out there?
Stanley Toh (03:10):
I cover anything basically that touches the end user from identity access management to all the collaboration suite, to all the support services, to all the end devices that the user use, be it on their laptop mobility devices to all the ui ux for our portals to anything that is to deal with the IT communication and user training, m and a onboarding. So basically anything that touches the Broadcom workforce
Jon Herstein (03:51):
And the Broadcom workforce is quite large as well. Roughly how many employees?
Stanley Toh (03:55):
So we have roughly about 40,000 employees and about 10,000 outsource partners and contractors.
Jon Herstein (04:03):
Alright, well thank you for explaining that scope. And what I want to jump into now is ai, the topic on everyone's mind these days. I want to start with a bit of a personal take on this, which is for you, what is your why for thinking about spending time on and investing in AI and everything that goes along with it?
Stanley Toh (04:25):
So with ai, if you break it down, there's two kinds of ai. One is to help the business grow. It will help your processes be more efficient, it will help your workforce be more efficient and it will actually change the way your engineers do their coding and troubleshooting. And including in this bucket, you have things like customer support, end user support, your help desk, your cybersecurity, your IT operations, all this, the ai. With ai, it will help them be more efficient because you have so much data and AI can help you digest and make sense of the data and present to you in the manner that is easier for you to digest and search. However, there's a second pool, what we call the feel good ai, and that is the one that you want to try to avoid the sprawl. But those are the one that people like John, if you look at it, you can buy an AI tool so easy these days. You can go to a marketplace, subscribe, turn it on. However, a lot of them are only doing a very specific small things and it doesn't make any impact to the business. If you look at it, AI is not cheap. As you know, AI is not cheap. AI platform is not cheap. So you want to implement something that has impact to your business and your business outcome. It will help you to go to market faster. It's not just for the sake of using ai.
Jon Herstein (06:30):
Yeah, it's a really great point because we're seeing this very, very consistently in our customer base that all of these AI tools and platforms are being sort of thrown at you as a customer, right, as a potential customer of these solutions and sorting out, as you call them, the feelgood AI technologies from the ones that actually add business value is a real challenge. So I'm sort of curious, does Broadcom have a framework for saying this tool or set of tools are enterprise ready versus these others? How does Broadcom think about that? We
Stanley Toh (07:02):
Have a governance from the onset. When AI first came into the picture, we actually form a governance. Even before you do A POC, you have to submit a request to go through legal cybersecurity technology stack review and provide a business justification even before you start POC because we want to try to prevent sprawl because it's so easy right now to buy from the marketplace, as I mentioned. So in the security area, we look at excess control also where the data rec resides and we want to make sure that there's no leakage to, we are r and d design company, we cannot afford our IP to be leaked out and things like that. From the legal perspective, we also look at things like privacy compliance and then as a multinational we have to see, we also have to look at things across country. The policies of each country can be very different. So the legal team will review all that and then for the tech stack, every AI platform out there, they do something, but there's a lot of overlap. We want to make sure that there's no overlap. If we already have a tool, why do we buy another tool that do the same thing? And then lastly, like I say, business justification. What are you trying to solve? What is the ROI and does it provide a business value? If not, why do
Jon Herstein (08:57):
It? Right? And so this governance that you set up, I mean obviously you had governance committees and processes in place before there was ai. So is this a parallel set of processes or have you just integrated the AI evaluation set into the existing processes?
Stanley Toh (09:15):
So we created this as a new governance processes and we integrate now with ai. So anything ai, any platform that has AI in it, even if let's say an existing platform, for example, box ai, it has to go through the same governance
Jon Herstein (09:38):
Before
Stanley Toh (09:40):
We even turn it
Jon Herstein (09:40):
On. So even for vendors you've been working with for years who've gone through all of your governance, if they add AI to the product, then they go through another round of governance effectively.
Stanley Toh (09:51):
Exactly,
Jon Herstein (09:53):
Yeah, that's very common. We're seeing something very similar. Now some of our customers are doing this in parallel with their existing processes and others are just sort of merging the two processes together. So it's interesting just to kind of compare and contrast. Now, when you think about a world where you've got a bunch of these different AI agents and platforms and tools all working together, how are you evaluating them to ensure that they actually can work effectively together? Or you think thinking about them all as distinct solutions?
Stanley Toh (10:23):
Every solution, we look at it individually, but we evaluate through a few things. What are you trying to solve is the first question to turn it, let's say we use box AI and we want to turn it on, what are you trying to solve? So for box ai for instance, we turn it on because in Broadcom we have 2.3 billion files in bot on the box platform, right? 2.3 billion. So there's a lot of content, there's a lot of value in it, and how do you make sense of all this? So that's why we turn on AI so that people can find things better and people can do a comparison, do a search, get inside all of it. So it justified the means why we turn on box ai. We also look at time savings, of course. We also look at efficiency. We also look at improving business processes.
(11:41):
We also look at is the tool that you are implementing or if you turn it on an existing tool set, does it improve time to market your and reduce the time to market for product development? And I know a lot of people may not be thinking of this, but John, when you turn ai, one of the question to ask is, will there be some resources that will be impacted because not everyone is ready? Do you have to re-skill them? Do you have to replace them because the AI tool set need a different set of skills. So we need to also look at resource replacement. And
Jon Herstein (12:42):
I do want to come back to that a bit later when we talk about culture and change and how you navigate through all of that because a very, very valid point. And I know it's an area where a lot of people have concerns in terms of what are the implications of ai. Before we get there though, you've talked about categories of capabilities of ai, things like productivity improvement, time to market, et cetera. Are there any specific use cases that you can share flagship ideas that are already either in production already or that you're working towards where you're going to be able to take advantage of AI and particularly box AI as an example? So one
Stanley Toh (13:23):
Of the key area is customer support. If you look at AI chat bot, they have been there for years. So in Broadcom we implemented a AI chat bot about six years ago. And for the internal support desk, I actually have 85% effective rate. So that's very high. So a key flagship is customer support. We also implemented this for our external customer support. So this is one of the easiest area, if any company that have not implemented any AI strategy, help desk is one of the key area,
Jon Herstein (14:06):
Low hanging fruit, so to speak,
Stanley Toh (14:08):
Low hanging fruit. And there's lot of internal
Jon Herstein (14:11):
And external.
Stanley Toh (14:13):
External, exactly. And there's a lot of what you call that platform out there that you can explore. Right now, not all businesses are built the same. You have certain processes that is very proprietary to you, so you need to choose the right platform. The other area that we implemented is actually for legal. If you look at it, and this is one of the area, we also use box AI for contracts, legal documents. They are long and not everyone can understand the legal language. For instance, if I read a legal document, I have to read the line three, four times before I understand what it's trying to say.
Jon Herstein (15:03):
If you're not a lawyer, right, then you're trying to extract insight from the document.
Stanley Toh (15:09):
Yes. So with ai, it's easier for you to decipher and get a summary of what it's trying to say. And the other thing that we use it very often is your previous contract and your current contract, they can do comparison very quickly and tell you what changes, what are important to you, what are the clauses that are important to you. And AI can pull that out instantly. So it save a lot of time for legal. The other area that we are doing is also to do code assist. So these are more technical for the r and d and they can generate, or I wouldn't say generate, they can code faster and reduce the cycle time. And if you reduce the time of development, you can go to market faster. One area that I'm very excited and currently we're with box professional services, is to reimagine the entire employee journey starting from pre-hire through their career in Broadcom to retire. And we are using all the latest modules and components from box enterprise plus, sorry, E advanced plus, I think that's what you call it, right? And all the AI components in there to see how we can make an employee journey even before they start more seamless and give them a better experience before even they start day one with Brock.
Jon Herstein (17:06):
And this is a vision you've had for some time now, even arguably before the advent of the generative AI capabilities. But how does AI enhance what you believe you can do with the employee experience?
Stanley Toh (17:19):
See, if you look at it, even pre-hire, when you have a job posting, there will be tons of applicants, and traditionally the recruiter have to scan through every single one. Then the hiring manager have to scan through all the applicants. And a lot of the applicants could be overqualified for a position that you're offering or they're underqualified or if there's certain things like visas and things like that you want to work in the us, all those requirements, some of them may not have it, some of them have it. There's a lot of things you have to see through. Now with ai, they can scan through that and then you can actually filter and reduce that to only give you applicants.
Jon Herstein (18:24):
So to be very clear, you're not using AI to make a hiring decision, but you're using AI to help you streamline the filtration process so that you more quickly get to the right pool of qualified candidates.
Stanley Toh (18:36):
Yes.
Jon Herstein (18:37):
Got it. So that's a great example. And that's just one in that employee life cycle that you talked about from pre-hire all the way through retirement. So I'm sure there'll be many, many more. If we take a step back though, Stanley, I mean, how are you all seeing AI really reshaping the mandate of the CIO of the IT ops function? How is it changing things on a bigger picture kind of a basis?
Stanley Toh (19:03):
This is a very interesting question. If you look at it, a traditional CIO or traditional, IT is always viewed as a cost center. We provide platform, we provide, what do you call that, make sure that things work. The
Jon Herstein (19:23):
Utility, the utility kind of
Stanley Toh (19:25):
Utility. Exactly. Keep the lights on and you're responsible for ensuring the infrastructure is stable. However, in the AI driven world right now, the CIO and IT have the advantage of looking at how AI can enhance the business processes. How can we streamline the business processes? How can we make everything more efficient? How can we make the r and d engineers more efficient so that I keep saying go to market faster, right? Because that's all about creating value for the company.
Jon Herstein (20:11):
Yeah. It's not just saving cost and saying, how do I provide the same set of services for less money, but it's actually how do I provide more services, more value, really become a business driver.
Stanley Toh (20:22):
Correct. And on top of that, John, there are also a couple of more things. If you look at it, a traditional it, right? The skillset is focused on infrastructure, networking, traditional software development, just making sure everything is working. Like you say, the utility is working perfectly, keeps the lights on. But on an AI driven, you have to look at different skillset. You have to look at upskilling the talent that has AI expertise. You may need data scientists because you have so much data, which AI tool to use to make sure that your data is relevant. And when you feed into your LLM, you have good data, not hallucinated data. You may need machine learning engineers, even ethics, ai, ethics engineers. So the skillset is very different now. So the CIO and the IT leadership must look at different kind of skillset to support the future IT moving forward in an AI world. The other area I'll think of is John, I would say cybersecurity. Traditionally, you look at DLP, make sure that your IP is not leaked out. We are concerned about brute force hacking, malware, phishing. Now with ai, you have adversarial AI coming. So it is an entire different world, but at the same time, you have better tools to understand all your monitoring data and see how better to protect your infrastructure and your environment. So cybersecurity is also changing.
Jon Herstein (22:39):
So you did a great job of sort of encapsulating some of the role changes and some of the new skills that would be needed are needed. Now. Are there any other big organizational shifts that people should be thinking about where you're taking big chunks of responsibility and moving them from one place to another, given more of an AI first kind of a posture? Anything else people should be thinking about? Organizationally?
Stanley Toh (23:02):
I think one of the area is actually working closer with the business units. Now,
Jon Herstein (23:09):
The
Stanley Toh (23:10):
Traditional, IT has often been looked at as a cost center, and we always try to make sure that we have a seat on the table to be able to shape the direction of the company. Right now with ai, if you collaborate more closely with the business unit, understand what are their needs, what are they trying to solve, and then you bring to the table the AI technology that is able to help them solve those problems, streamline their workflows, be it in logistics, be it in finance, even hr, and help them transform. If you've transform the business unit, the collective business unit will transform your whole enterprise.
Jon Herstein (24:06):
Yeah, you're navigating that while all these things are happening, it's pretty tricky and you're responsible for the end user experience. So I am curious what you've seen and heard from folks, your various stakeholders, employees, internal and contractors in terms of their reactions as you've introduced AI into some of these workflows, mostly positive, negative, fearful, excited. What are you seeing and hearing from your teams?
Stanley Toh (24:34):
From the experience side, right? We have done AI chat bot for a long time and it's easier now. And then with all the AI news every day, everything, everyone knows about it, there's no resistance. Now the problem I have is everyone wants ai. Everyone wants to turn on ai,
Jon Herstein (25:00):
But
Stanley Toh (25:02):
Are we turning on the right tool? Like I say, we want to make sure we do not have AI sprawl. You do not want too many platforms that are doing the same overlap, doing the same thing deployed because they're expensive and every enterprise only have a finite number of dollars that they can spend. And AI is not cheap. And if one tool can do it, why you need to buy multiple tools for multiple bus or multiple teams doing the same thing. We have to look at now, not only the business unit wants ai, but different business units may have different requirement even within one business unit, different teams may have different requirements. So how do you make sense of all that and deploy the right platform? So that's why we have the governance in place.
Jon Herstein (26:11):
Well, and how do you then balance, given the proliferation of all these tools and the availability of all these tools, how do you then balance the experimentation that you need to do to say, well, let's try this out and see if it's valuable, see if it enhances our business processes, but balance that against being managing risk in a sense you want to experiment but not take on too much risk. And so maybe how do you think about that trade off?
Stanley Toh (26:40):
That's why we have that governance. There's this survey that everyone of them have to go through if you want to do even A POC. And in that survey, like I say, we have legal, cybersecurity, technology and business justification. When the users start writing those, it also form a talk in them. Do I really need this? Is there another tool that is already available that I can solve my problem? So it is shifting the culture and the mindset of the workforce to think differently. Not I buy kind of culture anymore is you have to say, do I really need it before I buy and if I implement this and I putting anything at risk. So if you have all those kind of question and you educate your workforce to think that way, then you have a governance that will prevent AI sprawl in the
Jon Herstein (27:58):
Organization. Makes perfect sense. So you're relying a bit on some self-governance people really thinking through what they're trying to do and accomplish, but then you've also got this centralized governance process where they've got to convince a group of people. When you think about getting through all that, not just the governance, but all the mindset change that you referred to, what are some of the top obstacles that you've run into? Are people kind of taking a shadow IT approach of just, well, I'm going to just go do it and then ask for forgiveness later? Data silos, what kinds of challenges are you running into with all this?
Stanley Toh (28:33):
They ask, some people will just turn it on.
Jon Herstein (28:37):
Yeah,
Stanley Toh (28:39):
I dunno if you've heard of this firefly.ai, which is a recording, right? And you invite to join conferences, you'll subscribe notes and things like that. Now in Broadcom for every conference, it could be sensitive. So we do not have AI transcribed turned on.
Jon Herstein (29:03):
Okay. The hard rule, no AI transcription of meetings? No. Okay.
Stanley Toh (29:07):
Because if you look at it, the discussion could be sensitive. It could be a product design meeting, intellectual property. Exactly. It could be a M and a
(29:23):
Discussion, it could be a go-to market strategy, or it could be talking about a sales deal of a potential customer, things like that. It can be very sensitive. Now if you look at it, the other thing is we don't do transcribe is in certain country, if you look at the AI transcriber, it takes action item who attend, who speaks and things like that. You have all this data now available. And in certain countries, especially in EMIR with the workers' council, you cannot monitor an employee performance with any tools. Now, with all this, it's possible for you to do analytics. So there's also that kind of hurdle. So when people turn it on, like I say, cybersecurity now have a big headache. They have to keep scanning. And then when we found something that is not sanctioned, we actually brought it. Got it.
Jon Herstein (30:37):
Now you've talked about there being a lot of excitement and interest and people wanting ai, but there's got to be pockets of resistance out there, folks who are concerned or not interested or worried about the implications for job security, et cetera. So how have you dealt with resistance and overcoming that resistance? And I'm sort of curious about things like executive sponsorship. Do you have leaders talking about the importance of this training, enablement, change management? Just how do you think about overcoming resistance out in the user population?
Stanley Toh (31:15):
Transparency, communication and training. If you want to wait, that's worth repeating.
Jon Herstein (31:21):
Transparency, communication and training.
Stanley Toh (31:24):
Yes. If you look at it, I think every company will go through this, right? There are some people who just embrace AI and some people who will not, there will be a lot of people who just want to jump on the bandwagon without realizing what they're looking for. Each bu each bu the teams within the bu, they want to do something. So there's a lot of those people, but there's also a lot who are not interested. So those people who are not interested because you don't have to use the tool because AI tool is not for everybody. If you're doing your job function well, why do you need something else to help you? So training, educating, make people understand what is ai. I think there's a lot of people know the word ai, but do they really know what AI can do for you? So make people understand.
Jon Herstein (32:37):
Yeah, I've had a few conversations with people where they say, I get what chat GPT or other AI tools can do for me in my personal life, but I haven't quite connected the dots on how it'll help me at work. I know how to do my job, I know what the steps are, I don't really need that. And so I think you have to push through that a bit and start to really show people what's possible. So that transparency and communication is really key.
Stanley Toh (32:59):
Correct. You have to show them the art possible,
Jon Herstein (33:02):
Right? For folks who are out there, maybe not quite as along in their journey as you are, any takeaways for them to answer the question? What do I wish I'd known earlier in this journey? Maybe you can save some people some time and some aggravation. What do you wish you'd known earlier, Stanley?
Stanley Toh (33:22):
So John, I think we have spoken about this in another conversation, and I always say nothing is free. Everything comes to the cost even if it's free, even if the AI platform is free, the cost is what are you risking by letting employee install and use it? If the AI tool is not properly vetted, it could potentially be disaster for your company not only on intellectual property being leaked out, but it could be the company's reputation. So how do you stop the sprawl from the get go is one of the things that we wish we look at it earlier, although we have the governance in place, the governance is the person need to submit to go through, but because of the marketplace and everyone can just subscribe to the tool, how do you control that? So I wish that we found away earlier instead of right now trying to do the cleanup. If you have that at the front at the get go, then you can manage not only the, what do you call that, the deployment better, but also to make your employee more comfortable.
Jon Herstein (35:07):
This may be controversial, but would you recommend, to your point about managing the sprawl earlier, would you say block everything by default and then only turn things on once they've gone through governance? Or would you have something that's a little bit looser than that?
Stanley Toh (35:22):
I would say have better control. You have to block it. You have to block certain AI tool that you know out there. So this is more of a job for the cybersecurity guys. How are you going to vet all the AI tools out there? Is going to be impossible, right? But if you block everything, then it became too restrictive,
Jon Herstein (35:54):
No experimentation if everything's blocked. So that
Stanley Toh (35:58):
Balance, yeah, it is a tough question. So unfortunately I don't have a good answer.
Jon Herstein (36:05):
Well, and I don't think there's a perfect one. And I've spoken to customers who've done both things where they've just by default blocked everything and said you've got to go through governance to turn anything on. And others who've done the exact opposite and said, let people use these things, but put policies in place to say what's okay and not okay. And as you say, there is no perfect answer of depend on company culture and tolerance for risk and all sorts of things. But I do wonder are there any governance controls that you all considered completely? So I'll give you one example that I think we think about a lot, which is if I'm using a commercial model that someone else is hosting, I've got agreements in place that they can't train on our data at all, full stop. Are there things like that that you say no matter what, this has to be true?
Stanley Toh (36:55):
So when we sign on any AI platform, that is one of the things that need to be there. That's one of the clause. Like you say, it is non-negotiable. I think it's non-negotiable for every single enterprise. You cannot use your data to train their model. But the other thing we also look at is the security.
(37:16):
Is it going to be an on-prem or in the cloud? If in the cloud at certain point the data is being processed in a third party data center, how secure is that? Because if you look at it, Broadcom is an r and d company. We are a technology company. We cannot afford data leakage. The other thing that cannot be compromised is hallucination. You want the data to be accurate. So the AI model has to be transparent. So if it is pulling certain things and telling you even our chat bot, if I'm giving you a solution, it help chat bot, I'm giving you a solution, I need to cite where the source is.
Jon Herstein (38:06):
I do think this point about citations is incredibly important. Now you're saying, here's the reference material that I use to give you this answer. And the user can always go and say, I want to go look at that. I'll give you a great example. We see a lot of our customers using box AI for things like policy publication. So here's all of our HR policies. Well, if I have a question about our policies, I can use AI to get the answer, but the employee might want to see the actual words in the policy before they proceed. And so being able to cite the reference source I think is incredibly important. That's a great example and a tip for people. Let's pivot to the future and I'm sort of curious, Stanley, what are you excited about? What trends are you following most closely? As we sit here today, agents are kind of the big topic, but who knows what it'll be three months from now? So I'm sort of curious, what are you and the team kind of tracking in terms of emerging trends? Right now
Stanley Toh (39:05):
In the customer support area, we are testing out agents. If you look at it right, we are better chatbot. You can enhance the experience of the end user. You can help them be, instead of waiting for an agent, talk to an agent, time wasted. The AI chatbot can help them solve their issue better. If you look at John, a lot of people are pivoting to self-service. Now, five, 10 years ago, people like white glove service, we always hear about white glove service. But now it is trending towards self-service. Even when you go to a fast food chain, you have a self-service kiosk.
(39:52):
Even you get to Starbucks for your morning coffee, even before you get there, you use the app to order. So people are used to self-service. So with the agent ai, it will actually enhance that. It will execute multi-step processes and help them solve more complex problem. In the traditional AI chatbot is more of a, you ask a question, I give you a few links I give you, these are the steps, but right now it's more conversation and say, Hey, can I execute this for you? And the agent will call another agent to solve it for you. So now you have self-healing on top of self-service. So I think that is one area that's very exciting. The other exciting area is more on knowledge, knowledge management and internal search. Like I say, 2.3 billion files, right? 18 PBI of content in box. Now how do I use all that knowledge, all that content and develop better knowledge base for my users? That's one of the area. The other area is obviously IT operations. If you look at it, an automated agent, AI can actually self recover when you have an outage. Instead of traditionally it's just notifying a support person to go look at this. And then they have to acknowledge they come in, they restart the server or they increase their CPU or they increase the memory. But with AI agent, you can do all that autonomously.
(42:05):
In Broadcom, we are actually deploying and it autonomous IT operation client lifecycle management end to end. This one is for your laptop management, removing the human middleware where from the start of ordering through the lifecycle of the laptop until retirement and recovery, everything is autonomous. So these are some of the areas and some of the future areas that I think is pretty exciting, especially in the end user support space.
Jon Herstein (42:47):
That laptop example is fascinating. And I assume you're not doing that with a single agent, but multiple agents that handle different parts of that process?
Stanley Toh (42:55):
Correct. So you have, see if you look at it, right, the process to deploy a laptop, I think most companies always the same. Someone order, do I have it in stock? If yes, I ship it out. If no, what do you call that? I order one. And then you do the shipping, you have the tracking, you have the receiving, you all these are manual steps, but right now you have agents running around and doing all this for you and you remove the human middleware. So I think it is pretty exciting.
Jon Herstein (43:37):
Maybe it's early days for this, but just given that specific example, any best practices that you all have learned about how you get these agents to work well together? I mean, it's easy to think about a conversational chat bot that just answers questions, but now you've got these different agents, as you said, running around. What have you learned about how to make sure they work well together? So
Stanley Toh (43:57):
If you look at AI agent, they are currently in the early stages of experimentation. Although it is widespread, it's still in the early stages. And in my personal view, we need to have the human in the loop currently to make sure that agent is doing the right thing. You need to incorporate human intervention so that you have checkpoints for human oversight. Sometimes you may need to override something. And especially, particularly in critical decision making, you don't want to rely all on your AI agent. You need to make sure that what you call it, that is checks and balances and you can override all you can execute. Ensuring transparency. Transparency. I keep saying that word. When you design an agent, you have to clearly explain what the agent is doing so that you can actually trust the agent and
Jon Herstein (45:08):
Documentation you're thinking about not just what the agents are doing, but when and where you insert human reviewers or approvers into the process checkpoints, that's all part of the redesign.
Stanley Toh (45:20):
And then at the end of it, you also have you facilitate a user feedback, John. So if you look at it, like I say, everyone is trying experimenting with Asian right now, right? So you're not a hundred percent sure if it's doing the right thing, right at the end of it. If you have a feedback loop, you can keep improving on it and go back to what my boss always say, verify and trust.
Jon Herstein (45:56):
I like it. It's a good mantra. And maybe to that point a bit further, but if you're the CIO and you're being asked to greenlight some new and AI initiative, what are two or three questions that you should be asking before you say yes?
Stanley Toh (46:12):
I would say the first question is what are you trying to solve? The second one is what's the cost? Not only implementing the platform, but are there any risks? So all those have to be factored in. And if you look at cost, it's not only buying the platform, do you have the right resources to support the platform, right? All these are post post-implementation costs, the ongoing support, how difficult is it to maintain it? And then lastly, is it a passing trend or is this sustainable? That's the last question I think that is important is right now I say everyone wants to jump on the bandwagon. Sure, but are you solving the right problem? And this is also going to be used long term instead of just a passing trend. You do it for six months and then boom, no one uses it anymore. So I think those are kind of some of the key question you have to ask.
Jon Herstein (47:27):
You're the CIO, you've said yes, you asked those questions, you've greenlit the pilot. What's the single piece of advice you would give that CIO on how to make sure that's successful?
Stanley Toh (47:39):
I'll always go back to the question while you're trying to solve, right?
Jon Herstein (47:43):
Okay.
Stanley Toh (47:44):
Find a use case that is manageable, gather the support, and then define what is the business outcome.
Jon Herstein (47:53):
I think always starting with that question of what's the business outcome to your point, what's the problem you're trying to solve? And sometimes really candidly, is AI the right solution for that problem? Maybe it's not, right? Maybe so. Maybe there's other ways to solve it. And right now I think there'll be a bias to, well, let's do it with ai, and maybe that's actually not what you need. So it's a great point. So Stena, this has been great. I like to wrap with a couple concepts that I think about a lot in the world of customer success. And that's basically three things. The idea of making sure we're delivering value to our stakeholders, that we're thinking about the culture and the change to people's roles and what they do every day. And then the experience, which I know you think about a lot, the experience for the users in adopting these new ways of working. So maybe you can say a few words about how you think about the critical path to value realization, important criteria for change management and the culture, and then what's your criteria for success from an experience perspective. So I know there's a lot wrapped in there, but value, culture, and experience. How do you think about those three?
Stanley Toh (49:01):
From the value perspective, I would say that, like I say earlier, enterprise operate with a finite budget. So we tend to implement a solution that often we need to implement a solution that is sustained, can be sustainable, utilize and deliver the right business value. On the culture side, I would say that we cannot stop this. The AI revolution is coming. If you look at it, we cannot stop technology advancement and breakthrough every decade. There has always been a technology breakthrough. Start with internet and worldwide web. Remember back in the days we have nothing. And then you have internet, you have worldwide web. See how it changes our life now? Now you have social media and tweets and stuff like that now, then you have personal computers and lan that revolutionized how people work. Then you have mobile phone and smartphones,
Jon Herstein (50:10):
Right? And cloud
Stanley Toh (50:12):
And virtualization and cloud computing. Now you have AI and ml. So to me it is don't stop, embrace it, embrace the change. But with that said, we do not want to sacrifice the integrity of the company for the sake of change. I think that's very important. Your company culture, your company integrity, your company reputation. Without that, there's no company. And on the experience side, I would look at at least three things. One is security. Obviously need to make sure that whichever AI platform you implement must have a robust cybersecurity measure, data privacy protocol and things like that in place industry regulation, especially cross country, if you are a multinational, but you have to go on the fine line of, what do you call that, being too restrictive until it's not productive or too loose. So it is a very fine line to walk on. Security is one of that. The other one I would say confirm, no hallucination. So this is critical, right? So transparency is critical to me. And I would say the last will be adoption and training. You must have a solid adoption and training plan because it has to be when, like I say, there's a finite dollar. The first one we spoke about, you have a finite dollar of budget. So when you implement it, it must be sustainable and people keep using it. It's not a fad, it's not a phase where after three months you have a costly AI platform that no one uses. So from an experience, I'll look at those few.
Jon Herstein (52:32):
And you think the key to sustainable use is enablement, training, communication, it sounds like. So not just throw the tools out there and hope people get value from them, but actually teach them.
Stanley Toh (52:45):
Exactly.
Jon Herstein (52:46):
Great words of advice. Stanley, thank you. Thank you so much. This has been a great conversation. I really enjoyed it. And I just want to say we truly appreciate the partnership with you and with Broadcom over the years. Appreciate you taking the time to share your insights with all the folks who are listening. And for all of you who are listening, I just want to say thank you for listening to AI First Leadership. Again, thank you to Stanley for contributing to today's conversation. And for the rest of you, if you found value in this episode, please subscribe, share it with your professional network, and we've got additional resources out there, companion materials on box.com. So please visit there and join us for our next episode as we continue artificial intelligence, content innovation, and the evolving landscape of enterprise technology. Thanks again. Take care. Thank you. Thanks for tuning into the AI first podcast, where we go beyond the buzz and into the real conversations shaping the future of work. If today's discussion helped you rethink how your organization can lead with ai, be sure to subscribe and share this episode with fellow tech leaders. Until next time, keep challenging assumptions, stay curious and lead boldly into the AI first era.