Resilient Supply Chain — How Leaders Keep Business Moving
Resilient Supply Chain is for supply chain, operations, procurement and technology leaders who need practical ways to manage disruption, reduce risk and keep business moving.
First launched in 2019 as Digital Supply Chain and later known as Sustainable Supply Chain, the podcast continues today as Resilient Supply Chain, with the same extensive archive of more than 525 episodes covering supply chain technology, resilience, logistics, procurement and sustainability.
Each Monday, technology and sustainability analyst, keynote speaker and former SAP Global VP Tom Raftery speaks with the executives, operators, founders and innovators redesigning how goods, information and decisions move through global supply chains.
These are candid conversations about what works in practice—not polished PR narratives or vague predictions. Guests explain how they are responding to supplier failures, geopolitical shocks, volatile costs, capacity constraints, changing regulations and rising pressure to improve both performance and sustainability.
The podcast examines:
Faster planning and decision-making under uncertainty
Supplier risk, sourcing strategy and business continuity
AI, automation, visibility and operational intelligence
Warehousing, logistics and fulfilment performance
Scope 3 emissions, circularity and responsible sourcing
The systems, incentives and organisational changes needed to turn data into action
The central question is simple: what helps a supply chain withstand disruption, adapt quickly and continue serving the business?
New episodes are published every Monday at 7am CET. Resilient Supply Chain+ subscribers also receive bonus analysis, highlights and briefings on emerging industry trends.
Follow Resilient Supply Chain for practical lessons from the people solving real operational problems—and building supply chains that perform when conditions do not go to plan.
What should AI agents handle overnight, and what should be waiting for a person in the morning? Elon Salfati describes a workflow in which agents carry out tasks overnight and bring the decisions that need his input into one inbox. In his example, a thousand agent runs leave him with four decisions to make.
In this episode I speak with Elon Salfati, CEO of Salfati Group, about how to divide work between AI agents and people. We look at how an overnight workflow can keep projects moving and what happens to accountability when software starts doing the work.
Questions we cover
What should an overnight AI workflow hand back to the team? · 9:46 Elon describes opening an inbox in the morning and seeing the decisions that need his input. A short list lets him make the calls that will unblock several projects.
How does accountability differ between buying software and hiring a service provider? · 19:48 With conventional software, the customer uses the tool and owns the outcome. Elon contrasts that with a service provider taking responsibility for delivering the work, and asks where AI services fit between those arrangements.
Who should be accountable when an AI agent takes action? · 26:56 Elon argues that a person should remain accountable. His approach is to give agents the routine work and bring the decisions back to the people responsible for the result.
If disruption hit tomorrow, would you know where your supply chain was most exposed? In 15 minutes my free scorecard helps you assess 27 resilience statements, calculate your score, and turn the result into three priorities and a 30 day action plan. You can download the scorecard free at tomraftery.com/scorecard.
Podcast supporters I'd like to sincerely thank this podcast's generous Subscribers:
Alicia Farag
Kieran Ognev
Gary Lynch
And remember you too can become a Resilient Supply Chain+ subscriber - it is really easy and hugely important as it will enable me to continue to create more excellent episodes like this one and give you access to bonus episodes of topical, timely supply chain resilience analysis.
🎤 Looking for a keynote speaker on supply chain resilience, AI, and sustainability? I help leaders understand the forces reshaping global supply chains, from geopolitical disruption and climate risk to AI and real-time decision-making.
Download my Speaker Pack: https://tinyurl.com/spkrpck
Finally If you have any comments/suggestions or questions for the podcast - feel free to just send me a direct message on LinkedIn, or send me a message using this link.
The way that it's set up means that it allows us to go to sleep and come back in the morning and basically have an inbox where we say, a thousand agents executed over the night. Here are the top four decisions that you need to make. Just let me know what you want to do. And I'm unblocking so many projects at once in like 30 minutes in the morning.
Tom Raftery:
The real question isn't how much work AI agents can execute. It's whether the decisions they leave to humans are the right ones made with the right information. Good morning, good afternoon, or good evening, where everywhere in the world. Welcome to episode 147 of Resilient Supply Chain stories and strategies that keep business moving. I'm your host, Tom Raftery. My guest today is Elon Salfati, CEO of Salfati Group, where he works with organisations who are adopting AI. We explore why automation projects fail, what happens when critical knowledge lives in people's heads, and how to give AI greater operational responsibility without surrendering human judgement. Because faster execution doesn't automatically mean better operations. Let's dive in. Elon, welcome to the podcast. Would you like to introduce yourself?
Elon Salfati:
Yeah. Thank you for having me Tom. Sure. I'll do like a quick introduction. So I'm currently the CEO of Salfati Group, which is a company that is specialised on AI solutions and delivery. We're mostly working with non-technical organisations. And we constantly see surprising experiences in the market. We have this huge almost like a staircase experience that we see in the market these days where we have, very non-technical people that don't touch AI at all. And then we have like this big step into people who consume AI and then this big step into now non-technical builders, come into play. And we have so many people out there building new things. And then we have like the final step where people do a lot of like more AI intensive work. And mostly what we do and what we spend our time on is to take people through these steps and making sure that they get to the, final stages as quickly and efficiently as possible, but, learn from the journey. Besides that, I'm also doing my PhD in AI security at Imperial College and advising the UK Parliament on, AI adoption.
Tom Raftery:
Okay, fantastic. And where do AI projects in operations most often fail? Is it like the model? The data? The process? Or the operating model around it? You know, what, What goes wrong?
Elon Salfati:
We honestly, we see it across the board. So the way that we look at it is we look at it as cycles, right? Or rings. At the very core of it, if you are in an organisation that doesn't consume or store any data and everything is like, very segregated and it's not prepared, and you would fall on the data part. And the more you go out the rings, the, like, you have like different types of problems. But the thing that we see the most, honestly, is around the change management aspect of it. There is this wave right now where we have like this amazing technology, coming into like operations, but many, many organisations out there perfected their operations to the point that it's almost their ip. And, companies like ourself or, even without us just AI coming into place and saying, you know what, we're not going to use AI just to make your operations slightly better. Instead we need to help you rethink what your operations look like now that AI is in place, right? Because it changes a lot of the methodologies. And we see a lot of frictions around that.
Tom Raftery:
People don't love change.
Elon Salfati:
Well, here's a surprise for you. They really don't.
Tom Raftery:
I am shocked. I'm absolutely shocked. And talk to me about Agentic AI then, because that's a, a large part of what you do. Where does hype meet operational reality there?
Elon Salfati:
Yeah. we have seen, through these steps that I've mentioned before, we have seen a lot of people use AI in different ways. And they all start from, yeah, I'm going to deploy Claude Enterprise within the organisation. I am an AI native organisation now, not really, not yet. And then they kind of go one step up and they say, we're going to deploy some agents across the organisation. We're going to give an agent to whatever, every employee and I will make all of my employees basically become managers. But what we see often is that most organisations on like the human capital side of it, are not really ready for these type of operations. Because especially like the bigger you are, you bring people for a very specific niche expertise. But AI is exactly the opposite, right? It's the thing that knows a lot of everything. So bringing someone who has like an expertise of a niche to manage something that is this broad, rarely works well. So what we have seen and where we spend most of our time on is flipping the methodology. So instead of a human with an army of agents, we strongly believe in an army of agents with a human.
Tom Raftery:
Talk to me about how that works in practise.
Elon Salfati:
Yeah. So for us, what we do is when we come to most companies, we start to walk them through their operation backbone, right? The flows that they take, the steps that they take in the organisation. The main premise is that in most operations that we have these days, the moment the company kind of 5:00 PM closes the door and people are going home, then the company basically stops working, right? Not fully, but basically. The premise of AI, or the premise of an army of agents with a human is that the company never shuts the door. The agents run constantly 24 7. And what we want to bring into the table is the ability to take the core reason why we hired people and magnify that. We sometimes tend to forget it, but in a lot of cases, we would hire someone for their creativity and their ability to make strategic decisions. We rarely hire people. just to put data into a CRM, right? We believe in humans more than putting data into systems. So the premise of an army of agents with the human says that we're going to look across all of your operations, all of the steps, and we are going to colour code them. We're going to start with green, which would be like human moments, conversations like this. I assume you don't want to interview an AI version of myself. And then there are human decisions, right? Although I don't need to be like in engagement with the human, I, I want to be the one to direct the strategy of my company. Even if AI can do it, I don't want it to do it. Like it needs to be my vision. And then the third step, which is where AI come into place, is hands on keyboard, right? Zero added value of someone doing the work, putting data into the CRM. This is where AI come into place. So when we go and we deploy all these agents across the organisation now data ingestion and data extraction and processes and engagement, all of that gets automated. But what doesn't get automated intentionally, is the ability to stop the process and say, Hey Tom, I need your decision on this approach. Like, tell me where do you want to go? AI will make it easy for you, but it's, your call.
Tom Raftery:
So when you talk about the factory or the company or whatever closing at five o'clock, AI agents can continue to work, but they likely hit a decision point at some point where they have to reach out to a human and say, Hey, Tom, which way do you want to go here? So at that point, they're not 24 7 either, right?
Elon Salfati:
Well, I think this is where the magic happens. Like the bigger the organisation is, the more of those things you have to do. So if you think per process, then yes, right? At some point you need a decision and you need to stop. But the beauty of it all is that you don't run one process. You run a thousand, a hundred thousand processes in parallel. So even us, for us as a company, we designed the company around this concept of personal agents and functional agents, right? So every person in my company has like their own personal agents and they configure it however they want to set it up however they want. And then we have functional agents, right? So customer success or product managers and so on. Which we defined as a company like this is how we do things. And the way that it's set up means that it allows us to go to sleep and come back in the morning and basically have an inbox where we say, a thousand agents executed over the night. Here are the top four decisions that you need to make. Just let me know what you want to do. And I'm unblocking so many projects at once in like 30 minutes in the morning.
Tom Raftery:
And obviously then comes the question, what should always stay with humans no matter how capable the agents become?
Elon Salfati:
Oh, that's a great question. I've spent the past five years like in the trenches of AI, and I'm not one of the people that believe that it's going to take over. I think it's definitely going to change. And, I think that what we, tend to forget, especially because I think that what's magical about, we call it AI, but it's the wrong name, right? The magical thing about LLMs large language models is that it speaks our language, right? So it, it feels almost authentic. But in practise it's just predicting, you know, what sentences have been used the most across history. It means that take strategy for example, let's take something simple like an operational strategy around whatever ap. If we go and let AI build a strategy for us. And similar to us as a small company that does that, there will be 10,000 more companies that ask the LLM for the same question exactly. So, yeah, I don't want AI to do my strategy because I need a unique strategy in order to win. And I also don't believe that creativity, whatever, you know. I tend to say to people that if an apple will fall on an LLM server, they would not discover gravity. Um, And yeah, it's not gonna happen. Like the, the technology is not there.
Tom Raftery:
Sure, sure, sure, sure, sure. And, where do companies then accidentally automate creativity or strategy, and what are the implications of that?
Elon Salfati:
Well, it's easy. And it's, so easy to fall into this trap. The obvious thing that we all see day to day, you open LinkedIn and you see a thousand posts, they all look the same. They all sound the same. They all have like this weird way of saying things. It's easy to create a thousand LinkedIn posts with LLM. It's much, much, much harder to create the, post that actually capture attention and say something meaningful. And I think people fall though. It's easy to create, you know, a project milestone on how we want to get these big factory, running. It's much harder to do it right.
Tom Raftery:
Yeah. Yeah, yeah. Let's talk then about something supply chain leaders understand painfully well, the things that only work because Mary or Ahmed or Jao knows the work around. So tribal knowledge. Why is tribal knowledge such a dangerous blind spot for AI rollouts?
Elon Salfati:
Yes. So the way that most AI systems out there work is that, or the thing that makes them powerful, let's say, is the ability to access a lot of data that was before that, hard to get to. So PDFs and documents and derive insights out of it. When something is locked in someone's mind, there is no tool in the world that get it out of there. And especially now with all these changes that AI bring, I think there are a lot of people who also feel from the change. And it's a way for them to protect themselves. Like there are the things that only I know and AI will never do, and I will never tell the AI what to do. Um, And, and I think this is how this is an interesting management challenge. Again, like the way that we solved it is to say that every person in the company has their own personal agents and it's a replica of them. And whenever I'm doing performance assessment across my company, the better your agent is, the better you are. If your agent generates for the company six, seven times than what I pay on a salary, and it's because that person made it accessible, sure. Why not? Everyone wins.
Tom Raftery:
Yeah. Yeah, yeah, yeah. And then what kind of operational knowledge rarely makes it into systems, but makes all the difference when something breaks?
Elon Salfati:
Human relationships. And I think this is where most operations are happening. If you go and you talk to like this guy in the second company that you need something from, but you already know that his sister just got married. And if you come with a gift and you get the thing that you need. This is way beyond AI to do like this quick, quirky relationships. Those rarely come into the system and we can help, like AI can help manifest it in the way that we do operations. We're trying to bring it up. It's never fully there. But the more it is, then the more interesting your value that you get out of AI becomes. Hmm.
Tom Raftery:
Yeah. Yeah. And there any way to capture that expert knowledge about relationships or whatever else without forcing people into endless documentation exercises?
Elon Salfati:
Oh, fantastic question. The way that we go about it, and I feel like a lot of people kind of miss on this opportunity now, is connect your whatever notetaker that you have in the company to the AI knowledge base. In three months, you would see some magical responses from AI after that. The amount of knowledge and insight that go into a conversation. Even think about us talking now, right? I'm pouring into this conversation everything I believe in. The moment, AI puts this, it's virtual hands on it. It gets magical.
Tom Raftery:
There's an important trade off here. Capturing knowledge from everyday conversations could be enormously valuable, but those conversations also contain commercially sensitive information, personal details, and things people never intended to share widely. So knowledge capture needs governance from day one. So Elon, when you walk into a company, how do you decide which process is the right first target?
Elon Salfati:
So the way that we approach it is, people don't really think about how they are going to measure success. Usually AI come in, like the stress to implement AI come from two places. And they tend to come out of fear, right? So it's either the boat coming into place and saying, yeah, we have to have an AI initiative, like go do AI, whatever that means, right? And then the second thing is because you saw the market and you saw the competition and someone doing something in the, in your kind of vertical with AI, and now they get an attention or customers coming in and asking, where is your chat bot? But people don't really think about how they're going to measure that success, right? And what we're trying to do is we're trying to go across the company and say, and, and going to put it into a sentence basically, right? If this one thing is going to change by this matric, by this amount in that metric then we had a successful initiative. And if you can put it into the sentence, like, the effort that goes into encapsulating it into one sentence is I think where it becomes really easy to understand where you want to do those AI initiatives. And some parts of the business, it becomes really, really hard to put. And also the measurement of it becomes really, really hard. And in some parts of the business and especially operations it becomes much easier. And, and this is where you want to go.
Tom Raftery:
Hmm. And what then does a good first AI operations project look like?
Elon Salfati:
So in many cases what we see is that a process that gets augmented by AI would then become at least two to three times faster from end-to-end. And the way that we do that is across the board, try to figure out what are the processes that has the most manual steps in it that bring zero value. Going back to like putting data into a CRM or an ERP system. The more steps you have across the board that you can automate away, the more impactful it becomes, and the more it frees your team to do the work that they can, really contribute to the company.
Tom Raftery:
Interestingly, your model fixes price around an outcome. What has to be true operationally before anyone can credibly promise that?
Elon Salfati:
It's an interesting perspective to, to look at it. So the way that we look at it, and, if we'll do another conversation in six months, I'll let you know if it goes well or not. but it is just a strong belief that I have that like the industry right now is basically building a new business model. And I think that what we, we were used to is that we have the SaaS solutions of the world, right? You go, you buy Salesforce, you get a tool that helps you do the work, but you are still accountable for everything that comes out of it. If you are horrible at putting data into Salesforce, then Salesforce bring you zero value. If you're great at it, then tonnes of value. And then on the other side of it there are, the professional service providers. and they take full accountability. It can get like really, really expensive. And then when they leave, you get stuck with a tool that you cannot manage. What I think we're ending up with is somewhere in the middle. It means that we are getting into an industry where I can come and say, are going to get the full experience of a SaaS solution. So you kind of forget the product, you forget about operating it, you forget about security patches. All of it kind of get done by, our agents, right? We will be accountable for the outcome, but it needs to be an outcome that we can measure. So when we go and we look at the process and we look at the, our spend per iteration of that process, if we can go and measure that, we cut down the process by 50%. I'm going to take a a big risk and I'm going to take a bite of what we gave you back, right? But you still get back and you still get it back on the second year. And I think it creates like this new environment where operation people get much more empowered to do things their way or according to their vision without being limited by the SaaS solutions and without taking on the risk of, getting or becoming a legacy organisation in two years.
Tom Raftery:
And how do verification and proof change the risk equation?
Elon Salfati:
So with AI, of course, comes a lot of new risk, but it also means that comes a lot of new processes, right? So where we cut down on manual entry of data, we do get more verification points into place. And I think this is how we, mitigate a lot of the risk in place because AI can, of course, like we all see it from, working with cloud and GGPT can say something with a lot, a lot of confidence, but in practise it's nonsense. We see less and less of it to be honest, but still it happens. I think the more verification step would be put along the way, the more it becomes guaranteed that we mitigate the risk. On the flip side, it also means that a lot of people now need to optimise the way that they verify stuff. Because now going in the way that, Claude and ChatGPT tend to give you a, a result is to produce a book, a hundred pages that hey, there is everything that you need to know, and no one is going to read that. It is too much, right? So I think that optimising how you're doing verification and how you're guaranteeing along the way that the agents are going through the directions that you wanted it to go is becoming slightly more interesting these days.
Tom Raftery:
And then where have you changed your mind most over the last couple of years?
Elon Salfati:
Around the way that people, work with AI. So we see a lot, a lot of people these days that are non-tech, technical in nature, they're more from the operations side, they're more from the business side. And now they, open the cloud application. They can build something in, two seconds and they can build like a applets in two seconds. Which is great. But you know, when you look on it from a technical perspective, from an AI perspective, or even from a security perspective now, like you have like this huge surface that has so much risk attached to it and so much, challenges on how you do proper development and how it scales and so on. I was not the biggest fan of it and, especially me being an engineer but it grew up on me. Like, I think the moment you give people the ability to do it safely, which is hard if you just open cloud or ChatGPT and build an application, but if you do it in like a proper corporate way and you give every person who is trying to build something, you inject into the regions directly, a lot of skills and a lot of capabilities, and a clear direction on how to go through like the corporate network and the corporate workflows and the ability to ask the right questions. I can now make every non-technical builder much, much, much stronger because in many, many cases we have like this huge gap between the technical people who build a solution to the domain experts, right? Put me in a factory. I I might build you a great application, but without a direction of a domain expert, I have no idea what I'm doing. And now we can put it into their hands and even if it's a communication medium, it's still fantastic.
Tom Raftery:
Sure. Yeah, yeah, yeah, yeah. But then on the flip side, what's harder than clients expect once agents start touching real workflows?
Elon Salfati:
So when agents start to touch real workflows a lot of the problems or the challenges become around how to understand it. I think we all think about AI these days as a chat interface or a way to communicate and converse, but in practise, most people are not used to this mechanism. Maybe in five to 10 years. And I think this is also the, like, these chat conversations are easiest to build, but in practise a lot of people don't really think about that way. And anything related to the, the, I'll call it human AI, relationship and interfaces are becoming much, much harder because there is a lot of people, especially in operation that come from, 40 years of experience in doing something a certain way. And people usually look at it in like a bad perspective. Well, you If you've done it for years, you're just not comfortable with change. In many cases though, it comes from a lot of experience on why they do it that way.
Tom Raftery:
Yeah.
Elon Salfati:
and just balancing between, learning from the wisdom of experience to saying you just need to think about it differently is, is a tricky thing. And I think it touches that component a lot.
Tom Raftery:
Going back to the tribal knowledge we, we discussed earlier,
Elon Salfati:
Yeah.
Tom Raftery:
Let's bring in the less glamorous bit governance because, you know, it's the broccoli of AI transformation, but unfortunately it's, it's kinda useful. So who should be accountable when an AI agent takes action inside an operation?
Elon Salfati:
Hundred percent a human. I, there is no, no doubt. And this is why we're building a lot of the AI agentic workflows, the way that we build them to empower people to make the decisions. But I don't want AI to make the decision. I want AI to do the, grunt work, right? The one that brings zero value but needs to get done. And I actually spoke to a couple of founders, I think five or six years ago. I was so fascinated by the way that they approached their product, I took it with me for a long time. So they were building, an AI version to look on, screening images x-ray images with AI. And what they have discovered is that if they want to have this solution being used in hospitals, they need to go to like massive FDA approval. And the go-to market is insanely hard. What they have done, which was really, really smart, was to look at it as a QA system for the doctor. Same technology, same everything, but instead of it saying, this is the result or this is the result, the doctor goes and whatever, do their own like traditional process and they go the qa, if they found something else, they pop it up. If they find any kind of misalignments, they raise it up and you can, you can look at it and you can accept it, and you can look at it and ignore it. And in many cases it saves lives, right? And taking that mechanism where we look at AI as a way to execute operations are really well defined. And also helping us make decisions is I think where most of the magic happens, Because, even the smartest people on earth, we cannot, at every decision point in time think about all of the ways that it can impact that decision. But with AI, we can do that, right? I can take every decision on, we will ever encounter in, in a process, and think about it in a thousand different ways, and I can raise it to the human to say, Hey, have you thought about like this angle as well? Sometimes the human say, they get like this eureka moment. Oh right, this is the thing. But it still sits with the humans, like the direction, the strategy all has to sit with humans.
Tom Raftery:
So then how should companies prove what their AI did, why it did it, and who approved it?
Elon Salfati:
So the way that we go about governance is we are going about it in the way that we have always done it, right? So we have audit logs and we have recording of who makes the decisions and at what time and based on what information. Where you measure the impact of AI is never about the decisions or the, growth of the company. You can use AI to do a lot of revenue generation, but the growth itself is usually very tightly coupled to the decisions that you made along the way. But on the operation side, right? On the execution side of it all, if you used to spend, 10 or 20 hours a week to put data into systems to go and execute stuff that you don't need to think about at all, and then you had maybe five or six hours where you needed to actually think and make a decision. What you want to see is you want to see the shift in hours, right? You want to see more and more people spending their time making decisions and thinking and being strategic and less and less of their time being spent on putting data into systems and, and executing and, and reviewing materials.
Tom Raftery:
And then what results tell you an AI operation project is genuinely working? You know what metrics matter more than adoption dashboards?
Elon Salfati:
Yeah, a hundred percent. That's a, that's a great, fantastic way to think about it. So the easiest entry point for AI is execution costs. Time spent on a task. And usually I'm trying to look at it not as a per task. I don't personally believe a lot on point solutions for AI. Where we see a lot of it happening is, especially when you have a process that is cross-functional or cross teams, it tends to break very often when you do like handouts between teams and when you need to wait for a decision and then someone needs to, at the very end of a hundred step process to do a review and then send it all the way back with, Hey, you haven't thought about like this thing, and then you start all over. Like all of that needs to go away. And the time that you spend, again, the time you spend to decrease that, and the time that it takes you from, I'll call it from, quote to cash in whatever process internally in the operation should be shrink down by at least 50%. And, and that's an actual metric that you can attach straight to the financial part of the organisation.
Tom Raftery:
Right, it's time now, Elon for the lightning round. So couple of quick questions, one sentence answers if possible. So, best first process to automate?
Elon Salfati:
Debt collection, financial operations.
Tom Raftery:
Worst reason to use AI?
Elon Salfati:
To make strategic decisions.
Tom Raftery:
Humans decide, or humans review?
Elon Salfati:
Both. It's part of the process. How can you make a decision without reviewing?
Tom Raftery:
Okay. Bigger risk, slow AI or reckless AI?
Elon Salfati:
Somewhere in the middle. Okay.
Tom Raftery:
Okay. What should leaders stop pretending?
Elon Salfati:
That using Claude enterprise or ChatGPT Enterprise is the thing that will make their organisation AI native.
Tom Raftery:
All right, and one phrase for resilient AI operations?
Elon Salfati:
Where I have to go with our army of agents, with the human, this is how you get resilient.
Tom Raftery:
Alright. And over the next three years, we're off the the lightning now, over the next three years, what's gonna change most in how operation teams use AI agents?
Elon Salfati:
So I think over the, the next three to five years, I would say, we would see more and more manual tasks going away. And where AI will start to shine is the ability to deduct insights. And I think there is a fine balance here that it's important for me to emphasise is there is a big difference between deducting insights based on information that exist and actual discovery. Thinking about new concepts. And I think AI is slightly starting to show early signs of the ability to truly deduct insights. So if you think about like general operations the way I, that I'm thinking about it is an engine that is running 24 7 and not only doing the operations, but also constantly optimising what it looks like and constantly suggesting new opportunities. And I think people need to start and think about it from, the operational perspective as it's not just about doing like pure operations and then, leaving all the strategic decision to the business side of the company. I think it becomes a true opportunity to change what the business look like. And I think it's important for operation people to be open to this.
Tom Raftery:
Alright. If people listening want to implement one idea from this conversation, where should they begin?
Elon Salfati:
if you haven't used Claude enterprise or ChatGPT enterprise, then definitely start there. But if you have implemented that, what I would recommend is to do, engage with people who kind of spend the time optimising that. And the easiest way to experience it in a meaningful way is to take a really, really small process and deploy an agent that can automate that process for you. you can do a lot of these days even within Claude Enterprise. So if you go to Claude Enterprise, go to Claude Code, they have like this concept called routines which are running on an interval that you give it, right? So every hour. You can automate a lot of your process, a lot of your day to day just by going through that. And I think this is where you get the magic of what's possible. You know, with food comes the appetite.
Tom Raftery:
Okay, A field question for you, Elon. If you could have any person or character, alive or dead, real or fictional as a champion for human led AI in operations, who would it be and why?
Elon Salfati:
Oh, Isaac Asimov.
Tom Raftery:
Brilliant. And why.
Elon Salfati:
Because I don't know if you'll, you'll remember, but he has this story. It's funny, I grew up on this story where he says that humans build a computer to try and fight a climate change. And they ask it how the human ask the computer, how they can revert climate change, and always replied that there is not enough information to answer this question. Always gave it more information. Always the same result. Eventually human got wiped out of earth because of climate change. And the computer, years later finally found the answer. But then he didn't have anyone to tell the answer to. So he created humans which is honestly, I'm so curious to know how Isaac Asimov came with the story, but like thinking it's, what is it like 20 years ago, if not more? Thinking about like this far away in an industry that is changing so fast,
Tom Raftery:
I
Elon Salfati:
think it's like an interesting skill.
Tom Raftery:
yeah, I mean, he invented the word robot and then went on to invent the three laws of robotics. I mean,
Elon Salfati:
Doesn't get better than that.
Tom Raftery:
it's amazing. It's amazing. The guy was incredible. Absolutely. Absolutely. Great. Super. We're coming towards the end of the podcast now, Elon. Is there any question I didn't ask that you wish I had, or any aspect of this we haven't touched on that you think it's important for people to be aware of?
Elon Salfati:
So I would say, although we're coming at it from a perspective of operations, I think that a big portion that is important to talk about is the security of AI. It's very easy to go and say, the impact of AI is so big, and I really want to get those results in. And I would go and bypass every kind of security measure that the company ever put in place because I can move so much faster. And you know, on the one side it's incredible. But on the flip side it's not that it's much harder to kind of understand the security risks of AI. It's a whole new surface that we have never discovered, which, you know, just talks to the level of how unfamiliar are we with what's possible these days on the security front. So, I would really urge a lot of people to stop and think about it. It's not to say that you should fear it to the point that you don't use AI, but think about it. Like, go about it in the right way. Talk to the right people.
Tom Raftery:
Yeah. Great. Elon, if people would like to know more about yourself or any of the things we discussed on the podcast today, where would you have me direct them?
Elon Salfati:
So anyone who wants to engage with us or talk to us or just explore, we love brainstorming with people. They can go to our website Salfati dot Group or search me on LinkedIn Elon Salfati. I'm more than happy to talk to anyone.
Tom Raftery:
Superb. Great. Elon, that's been fascinating. Thanks a million for coming on the podcast today.