AI Deep Tech

0:06Agentic AI has to be designed around responsible AI. Can you audit? Can you trace? Can you control? The risk we run is we create systems that recursively learn and change without the signal and the switch to turn them off. Ahmed Mazari is the former President of Microsoft Asia and now the founder of Singapore based AI startup Kaijentic. In this episode, we explore why the era of simply searching for answers is over. Technology is moving beyond generative AI to tools that actually do the work, acting as a team mate that perceives problems and executes solutions in real time.

0:56Find out how Agentic AI is changing the way we build, manage and grow the modern enterprise. Ahmed, now you have moved from steering a massive ship at Microsoft to then launching A speedboat, I think is the best way to describe it, at Kaijentic. And you call Kaijentic that intelligence layer. I would love to know a little bit more about that. What exactly do you mean by the intelligence layer? The intelligence layer is a connection between human actions and human intent, and it's an ability to enable the human or humans to make better decisions. I mean, if we step back and think about what enterprises are built around, enterprises are actually built around functions because they're built around hierarchy. They're not built around knowledge. And if you compare it to the learning of an individual, a child that's born learns from observability and that observation happens in good times and bad times, at school, at home, with friends, in social connections, with music, that those are in many ways not connected systematically in an enterprise.

2:10And our intent is to ensure that we're able to bring the signals that happen at an in an enterprise to the action that needs to be taken today in an effort to change the metabolism, right? How do we change the metabolic rate of an enterprise? You change the metabolic rate of an enterprise by bringing more information, more insight. But finally the human makes the decision, and that's the human in the loop. So tell me a little bit more about this concept of metabolism then.

2:42Metabolism no different from humans. I mean, if you think about when the printing press came in, what is the printing press? The printing press's knowledge. What is AI? AI is knowledge what was electricity, another general purpose technology. It changed the pace at which we worked. So we clearly see AI as a general purpose technology which has the ability to significantly change the pace at which enterprises work. And that's what I call metabolism. I have to ask, why would you leave such a big organization and then say, OK, I'm I'm going to go do my own thing? Well, it started first with a a yoga retreat. I actually said 7 or 10 days of doing nothing, no caffeine, you know, just clean up your mind.

3:32And it really made me think the first, you know, landline telephone at our home was at the age of 13. And today we live in a world of, you know, 500 Meg on your phone as as Internet speed AI. Like where in the world would you have seen that in about from the landline for like 30-40 years? So much change. So I asked myself, I said, look, I never had the courage, the capital, the capability, the expertise of the experience to participate in anything. Earlier I was still in the phase of, you know, I have to find and create a foundation of learning and, and confidence. I can't miss this like I can't miss this. I can't try.

4:13It incredible. So then when we think about go back to the layman and their perception of AIA, lot of people see it as predictive guess is the next word or number maybe, but it's there to help supplement you talk about that human intent. Can you explain that difference then maybe succinctly for me between that? The AI Intelligence Slayer in many ways is to help the enterprise understand and connect dots about stuff that for example, if you say let's not pay the invoice, the AI understands that as an action that needs to be taken, right?

4:56If you look at the agentic layer, and if you have taught the agentic layer, because agentic is about agency, what is agency? Agency is accountability, auditability, traceability and reliability, right? So if you can teach your agent so when you say please do not pay the invoice, the next prompt you, they will ask you is when would you like to pay it? When you say, well, when the cash flow is ready for payment, then the agent knows that they have to, it has to go and check for the cash flow.

5:34According to recent reports, generative AI could add trillions of dollars in value to the global economy annually. Asia is currently leading the charge in deployment, with markets like India and Singapore adopting AI tools at a faster rate than many Western counterparts. However, the challenge remains moving from pilot programs to actual enterprise wide value. Can you maybe give us a very concrete example of a task that maybe a standard LLM like ChatGPT just can't do, can't do and and cannot that an agent actually can? If you were a company that was shipping a very critical product that was passing by the Suez Canal, right, In today's world, that is static information lying in a supply chain system if you've not used an ejectic layer. But if you program your ejectic layer to say, look for all critical shipments, I want to overlay geopolitical risk. So now your system is going through geopolitics and saying, by the way, your shipment is in the Strait of Gibraltar. It will hit the Suez Canal in X hours. But there is this risk that we could get impacted by. If it shuts down, our shipment

6:56is stuck. Do you want to make a change now? You as human in the loop then can make a decision. The bot or the the agent is not making the decision. The agent will recommend based on insights. But does that also leave U.S. Open to mistakes? I know this sounds very negative, but is there a chance where it will make a mistake and whereby that mistake may be irreparable? Where potentially millions of dollars gets transferred to the wrong account? Or maybe if we take too much of A hands off approach we leave it to the AI and that is at our detriment. I think human agency and human intent are very critical in the age of AI.

7:39Our advice is that AI has to be designed around responsible AI. Can you audit? Can you trace? Can you control? I think the risk we run is if we create systems that recursively learn and change without the signal and the switch to turn them off. So any enterprise must have those levers. Any humans that work with on the consumer side must understand the risks, and companies have to ensure that those kill switches are available. I am not in the camp of autonomy of agents. I'm in the camp of enabling humans to make the decisions and ensuring that humans understand the impact of those decisions. You've launched this venture with SNBC Group, 400 year old Japanese bank that has allocated I think it's a $340 million war chest to generative AI.

8:42So they're effectively your customer zero in that sense. You are building out this whole scope and system for them. How are you ensuring that whilst you're doing this that you're actually building a global product as opposed to just a single tailored solution for one organization? We have one of the biggest banks in the world as Customer 0, so we have a sandbox to try and experiment, right? When do you ever get that? When do you ever get that? We're so fortunate 2 You know when you have a start up, you're thinking about capital, right? We got that sorted, right?

9:18So we have a customer that's backing us. Third, we're an independent company. We have an independent board. We can experiment on the most complex things. And if we can solve for tacit knowledge, human knowledge, connections in a 400 year old bank, I reckon the other is not going to be that very challenging. Why Japan though? I was interviewing a very young engineer and he said to me something fascinating. He said Japan has been 2000 since 1980. Japan had products, solution and innovation that was far ahead of where they were and even today in the physical world, Japan still does a very good job.

10:05But in the digital world, it has remained under invested and hence the bank which is at the epicenter of organizing and enabling industries is the best place for experimentation and enablement. So then in the next few months then with your focus on the intersection of business success but also responsible AI, what do you think is the one major shift in how we work that people just aren't seeing just yet? I mean, let me kind of parse the question slightly differently, right? Almost always when something new and clever happens, we overestimate the impact in the short term and underestimate the impact in the long term, because the pace of change is quite dramatic.

10:50What I see is that there will come a time not so far in the future where we will start having a conversation of we have 200 employees and 5000 agents in our team. And there's a very interesting, there's actually a pharma company in the US who has named their Chief HR Officer as also the Chief AI officer because their thinking is that AI is capacity capability, right? Just like humans. The most, most critical aspect is you said it in your question, responsible AI, is it guardrail, right?

11:34Is there auditability? Can you look at the action log? Can you define what intent? Because the difference between look, if you think about it, the difference between age, AI and human is intent, right? And conscientiousness. We do certain things because of conscientiousness, right? That agents and AI cannot. Remarkable. Thank you very much. Thank you, Kelly. A fascinating look into the mind of a founder who has seen the industry from the very top. A big thank you to our special guest, Ahmed Mazari, founder of Kaijentic.

12:10You've been watching I See Your Trade brought to you by I see Markets trade up to I See Markets. We'll see you next time.