Name: Autonomous Sourcing and the Future of Procurement
Uploaded: 2026-05-14T16:16:14.174Z
Duration: 1 h 8 s
Description: Autonomous Sourcing and the Future of Procurement

Transcript for "Autonomous Sourcing and the Future of Procurement": Before Amazon Business, buying for work was chaotic. Now it's easy to find products from thousands of suppliers in one place. Save on every type of purchase from individual items to bulk orders. At any time, you can view your spending on pre built easy to use dashboards. Plus, you can free up cash flow if you choose to extend payment deadlines and view and approve your team's purchases easily. With Amazon Business, things just got a lot more manageable. Hello, everyone, and welcome to today's webinar in association with Amazon Business. Today, we'll be diving into autonomous sourcing and the future of procurement. Our session is part of an ongoing partnership with Amazon Business, where we explore the most important strategic shifts shaping the future of procurement and spend management. Procurement is entering a new phase of transformation, moving beyond simple digitization and automation into a space where autonomous agents can manage sourcing events, run negotiations, and evaluate suppliers with minimal human intervention. This is a fundamental change in how procurement operates, creates value, and supports the wider business. Today, we'll explore what this means in practice from self correcting sourcing cycles to the evolving role of procurement leaders as strategic partners driving innovation and resilience. I'm Ella Wilkinson, director of BizClick Studio. And joining me today are Ben Godfrey, director of BGA and former VP of procurement for Rolls Royce, and Abby Osh, assistant manager at Deloitte. I'm sorry. I've I've I've killed your surname there, but I I have obviously given you guys a very, very quick introduction there. Can you please tell me more about who you are and why you're joining us today? I'm gonna throw it to you, Abby, first so you can correct my terrible pronunciation. It's better than normally being corrected by word, which is normally corrects my name to hairdashery. So that's it's way better. But, Abby, I lead, our digital procurement team within Deloitte in The UK. Have over kind of ten years experience in implementing digital, procurement solutions, started off in category management, and have progressed into the tech space of things, itself. So from your SP suites and now the new kind of frontier of technology itself. And we're doing some really cool things, with a lot of cool clients. So I'm excited to kinda hear a little bit more from Ben as to what his thoughts are from a client perspective, industry perspective as well on how some of these technologies are shaping kind of your thinking, your targets internally, and those kind of things. So, yeah, excited to continue the conversation around Thomas Sourcing. Oh, well, thank you for joining us today. It's it's a pleasure to have you with us. And I will then throw to Ben. Thanks. Good to be here. Most of my work sits in high reliability and high impact supply chains. Think about environments where procurement decisions don't stay in the safety safety of the data system. They really show up on the shop floor and production and operational outcomes. So I think what's changing now isn't just automation. It's the shift towards systems that can make and adapt decisions, not just execute tasks. In my experience, procurement sits across three realities. There's a digital layer, which is data systems that increasingly autonomous agents, which we can talk about today. But there's still that physical layer defined by factory capacity, lead times, and constraints, and then really also not to forget a human layer of emotion emotion, judgment, and uncertainty, and they don't always clean the align. And I think the ability to sense to be able to orientate and act across those dimensions is key for procurement. So not question of how much or whether we automate, but how do we design systems that are able to stay aligned to that physical reality. And that's the shift, I think, that we're talking about where procurement's less of a linear functional process. It's not about persuading stakeholders or suppliers or just about persuading stakeholders and suppliers. It's an adaptive decision cycle that delivers your supply chain design and requires some level of strategic control. Well, it's it's a joy to have you with us as well. So I think we're in for a really good conversation today with all of the experience we have lined up here. So, obviously, we've seen automation play a role in procurement for years, but autonomous systems are really a one a one up from that, aren't they? So this is an open question to both of you. What distinguishes autonomous sourcing from those traditional automation methods? I guess maybe, I like to talk about it in, like, three levels of autonomy, is the concept that we tend to use, which is, you have kind of something that's, AI assisted, which still requires a level of kind of human dependency, but you have, AI enabled through kind of either with it'd be pricing, data enrichment, certain elements that are coming through. Or you have, I guess, AI augmented, which is it's doing certain elements of the task for you, whether it be, comparing, your RFPs or whatever it might be. But small tasks are almost always executed through, an AI agent or something. And then your third level is your full optimized autonomy where that full process is being done by the solution through AI and an agent in the background with the guardrails you've set. Now I think so that's, I think, the the general way that we look at how that's differentiated. So a lot of the initial solutions that came on the market when I think I first started in procurement were probably in that eight are now getting to that AI assisted where you'd pull in some information. They're starting to think about certain things that might be able to to be automated through AI, but it's how do you fully automate that through, an agentic AI process or technology, whether that be with enabled within the platform or externally through call, you know, AI platforms like Anthropic, OpenAI, all of these things, itself. So I that's, I think, the change we're seeing. Autonomous sorting to me is that last level of autonomy. So I guess the the next logical question is why now? Why why do you think we've got this big shift happening to to fully autonomous sourcing? I mean, maybe if I maybe if I jump in, because I I agree with the points on those levels. It's a nice way of thinking about it. I think clearly in every everybody's life, sort of AI is becoming increasingly embedded. And I think you can follow that logical flow that was just being described, but I think one thing many, you know, many conversations end up at a little bit of a level of being underwhelmed by the outcome. So maybe think about, well, why is that? And I invite you to think about it this way. You know, supply chains are permanently under pressure. Volatility is structural. It's multilayered. It's significant complexity. And in that, there's no finished state of optimization. I mean, the idea I have is that there is a best version of your supply chain. You'll never achieve it. You might get close to it fleetingly. So this is only ever about continuous improvement under those constraints. And if you try and plug and play a tool, into that kind of environment, I think that's partly what we see this sort of level of selling that's not matching in that sort of reality, because it's not embedded with a continuous improvement mindset in some instances. Now the positive is that I think automation in decision making is hugely powerful. It's a hugely powerful leverage, you know, in in the role of sort of working with supply chains and defining performance, And it can scale more significantly and bigger than perhaps human hands can. So I think there's a huge opportunity, but it does require perhaps more of a depth of understanding and clarity of thought to realize the benefit. Then I think in in a lot of instances, it's understood upfront. So do you think then people are adopting it in, not really not really doing that that future planning and just using it thinking it's gonna solve everything? Yeah. I think I think there's some I think there's some of that. I think, you know, there's a spectrum. I think that the businesses that are really, really sort of realizing the benefits, one, you know, it's not an overnight thing, but they are layering us on a level of maturity that kind of works with sort of learning mindsets, continuous improvement, problem solving skill sets. When you're able to apply, you know, a system and a tool in an area that can complement that, I think that's where you see the benefit. It's your question where where you sort of throw a tool at something and perhaps expect it's gonna fix something that you've not thought about the constraints and the decisions that you're trying to to think through or the complexity of the trade offs. That's where I think you see a level of that underwhelmingness that I that I described. I think I definitely agree. I think there's, a couple of kind of, drivers that are driving us to more automation. I guess, just as a general technology uplift in, like, our personal lives, in our personal productivity, and technologies as a whole. The availability of chat GPT, clause to your personal lives, in your phones, in your Copilot, on your teams, etcetera, makes a huge difference in being able to do your daily activities. And people now expect that, in kind of those enterprise solutions itself. Right? And it's the same conversation we were having, you know, a couple ten years ago when Amazon first launched and, you're like, well, we want the Amazon buying experience in our work life. And that is the the goalpost to get to. And I think it's that same innovation frontier that's driving a lot of that internal, corp in in the corporate world itself. And then I think the second part is these, as kind of Ben mentioned, these kind of technologies allow you to address various different elements that have been potentially untapped in the past. So, I guess, drive you know, being consistently under pressure a lot of the time, procurement has to focus on the strategic relationships. I think allowing some of the oh, like, some of these technologies allow you to focus on areas and spend areas that you haven't been able to tap into with the same amount of effort, and the resources itself that you have, but with additional manpower of an agent doing some of that activity and then getting, you know, more benefit for it. So it's really flipping, I think, some of the the work into that strategic kind of function rather than having to think about, okay, how do I manage my mid to tail spend and all of these, things that I haven't been able to get to where there is probably value because it's been untapped, but I don't have the capacity to to do so. Yeah. And I I think that's right. And I think the the point around sort of that shift is it moves from this is an interesting technology that can just plug and play and and fix to a a level of sort of understanding and the work involved actually to make it effective. If you speak to, you know, there are many providers of agents for everything you could think of in the whole procurement cycle. If you speak to them, they often underestimate the challenges of plugging into the datasets, of their potential customers. They said that we've got this we've got this agent. It can do these cool things. We say, yes. It can, but it won't work in one operating environment to another without some really careful thought and some tailoring. So I think I think that specificity, that's a word. I think that's one of the things that's come out in recent kind of learnings and engagements that you you can't hope to just repeat an outcome you've seen elsewhere even within your own organization without really getting under the skin of what it is that made it work and what variables you need to understand or or or perhaps control. So then how do you think organizations should be rethinking their workflows to make these tools successful? I think it's really what I think every all organizations have a little bit of a different prioritization, but I think it's really thinking about, okay, what are they trying to address? So we have there you know, as I mentioned, you've got some of the mid to tail spend that you can address. So if that's the element that you wanna address, then I think it's thinking about, okay, what are the use cases under that? What is the guardrails around that? What are the negotiation levers that I'm willing to use within my mid tail, like swapping out suppliers, reducing on delivery times. Whatever that might be, it's really thinking through some of the strategic decisions that might go behind some of that. Mhmm. But if you're looking to optimize some of your strategic spend, it's thinking about that in a completely different way because you can't apply the same generalized logics to your, you know, critical logistics supplier, for example, and really think about what the levers are in the different categories. I think we're now seeing I think in the past, we were able to apply a bit of a broad brush to sourcing, and I think the next iteration of that is way more detailed kind of at a category level or at a different kind of spend level what those, criterias and prerequisites, logics need to be, itself. Brilliant. So, I mean, obviously, historically, much of procurement's time has been spent on transactional and operational activity, whereas autonomous sourcing has has the potential to remove all of that. So with the introduction of autonomous agents, how has it changed the core KPIs of the procurement team? I think I think you you've gotta recognize that the shift starts to move from some of the administrative work becomes system managed. And therefore, the the sort of human role is thinking about the outcomes that are important to the business, and it's thinking about system architecture, some of the constraints. And therefore, your KPIs shift from some throughput speed or efficiency or tracking number of LTAs to the kind of things that support this sort of organizational health and its effectiveness. So how do you define resilience or responsiveness or continuity? You you see this shift into the baseline tracking becoming more standardized, more normal, more understood to some of those, perhaps considered higher value add type, the thought processes and activities. Abby, you're nodding away. I wanted to go ahead. I think, I I definitely agree. I think it's, I think it comes back to how does procurement then play a part when these technologies are doing some of that activity. And that is also a bit of a shift in thinking of the the capability of procurement and, I guess, what they bring to that autonomous optimize autonomous process. And we talk a lot about, you know, changing how AI is changing our jobs, but I think this is where it comes to fruition a little bit because actually on top of the AI, kind of optimized processes, I think there's the the new role will be how do we gatekeep those agents to do what they must that we want them to do, not overstep, not understep. What are those guardrails and do some of that gatekeeping activity, some of that control, like looking at AI agents as a control tower. Is it bringing you the productivity that you need it to do? Is it addressing all of the elements that you want it to address? What are the some of the error handling on these agents? So when they don't know they're stuck, what does that look like? And that's a completely different skill set, I think that procurement hasn't necessarily developed in a sourcing lens because they have been doing they've been doing what the, you know, the activity. And, so it's a real kind of shift in technology capability and literacy within some of these, within the sourcing teams itself. And I know that's a bit of a difference in terms of KPIs, but I think that's really then driving okay. If you are having all of these agents, then what because these agents still cost money, believe it or not. But, you know, they so what then does your ROI look and how quickly can you get these agents spun up? But then also how quickly can you improve on these agents to deliver more and more and because you can do that a lot quicker. So I think your KPIs are a lot more as they're evolving as well. So do you I I I agree with that. Sorry. Just to jump in, because I think if you were looking at your dashboard, you've you've got some of those traditional sort of process metrics. You still don't want to understand where you are with your number of contracts placed, delivering forms on time, how that contributes to the business. But we're also gonna be looking at quality and cleanliness of data inputs. You know, where are you seeing those gaps and, you know, what's what's that doing? You're gonna be thinking about, the quality of the external sensing, which sounds quite a sort of vague. But, you know, if you if you want to have an agent that's able to act independently, it's going to act on triggers. So where is it getting that information from and how reliable is that information? And then what would happen if you looked at a different source of information theoretically? You're gonna have something about, you know, the number of experiments you've run. You know, how many things didn't work as you expected because that's goodness. That's where you get this iterative learning loop from. And it's the treatment, if you like, to some what I was saying before about sort of plug these things in and expect them to work. These are the kind of metrics you need to have around you to understand how are they working in real time and how do you continue improving them? What are you changing? You know, what's working well? What's not working well? All of that's good learning, and you can imagine staring at that as your dashboard. Now as a procurement person says, I'm understanding what I'm doing with the function. I'm understanding what's country business. I'm understanding what this part of my resource, because that's how I see the agents. I'm understanding how they're working. And you can't have, you know, a traditional sort of performance review with those guys. You don't sit down and say at the end of the year, how did you do? You've gotta you've gotta build something so you get that feedback loop, and it becomes something that you can you can work with that is intuitive. So do you almost see the future of the procurement team as a team of data analysts, or do you think that the human art of negotiation is still gonna be that final mile? I think it was touched on a little bit earlier. I think it depends, which is the traditional answer. Right? But I think it does depend on kind of the category and the criticality to the business. I think the kind of technology and what we see at the moment is suited to, you know, kind of be more autonomous in certain categories. Maybe they're considered lower risk or or less critical to the business. That's where you would perhaps expect to see adoption. The way you where you're dealing with a highly constrained supply chain and that human layer I talked about in terms of emotion. I mean, there are there are agents and solutions out there that can negotiate. Mhmm. You know? Can they can they yet understand some of those nuances and some of those things? You know, there is this technology that can help you read body language over over a call and can support negotiation from that perspective. So I think in all instances, there is technology and capability that can be beneficial. But does a procurement team become purely data analyst? I think I think that undersells procurement team do and and can do. I think strategic thinking, sort of pattern recognition and understanding, I think, you know, commercial creativity. Some of those things, you know, still remain as being from the from a pure perspective in in my opinion. And I think this is where we it's that shift from tactical strategic to doing a lot more of that strategic, the business partnering, the the the supply management, some of leveraging some of the the the personable relationships within the business, within the supply chain itself that become highly critical skills. And I think the skills that kind of differentiate between the some of the tactical activities in the future or the the data analysis and things like that. Because I do think there is gonna be a more requirement for some of that analytical thinking, that engineering type thinking. But I think it puts more pressure to to think about how you are building those relationships, internally and externally, within your organizations. So, right, if you were hiring today then for for a procurement leader, what what are the three main skills that you'd want them to have? Oh, gosh. I guess I think it I mean, I think that the key thing is a future vision and being accepting of that future vision involving a different way that procure I guess, be okay with the fact that procurement may look different in the future. I think accepting that vision is, like, critical on my list. Mhmm. Because it really does make a difference, I think, in terms of the top down of how much you're able to bring your team, to deliver that. I think, the next for me is some of that strategic thinking, that business partnering, becoming hugely critical, how you'll risk mitigating using those relationships, whether that be, you know, political changes that are happening, sustainability changes, regulation changes. I think that's where this becomes more relevant. And then I think the last, element is really thinking about how procurement kind of embeds into the wider ecosystem of the the back office and finance functions. I think that's something that we're very conscious of. This technology is bringing a lot of the functions together, whether that be finance, procurement, GBS functions, HR functions, to think in a bit more of an enterprise architecture and way of working, itself rather than thinking just as to what may kind of benefit procurement in terms of whether that's be technology or KPIs, etcetera. It's thinking much wider, across the across the piece. Fabulous. Same question to you, Ben. I, I have the the slight disadvantage that the bandwidth dropped slightly. I think the question was what are the three characteristics you'd be hiring for. Yeah. Absolutely. It's a slight disadvantage, but got it in the end. So, maybe slightly different tack. I think I'm I'm looking for curiosity and a growth mindset, probably first and foremost. I think I'm then looking for problem solving and improvement continuous improvement skill set. And then finally, a commercial action bias. You know, how does how does those things combine, find opportunities, and turn it into value? Fabulous. So, right, let's move into actually applying autonomy in high volume, low touch areas. So one of the most practical and immediate applications is tail spend and sourcing execution. Obviously, we've mentioned earlier, tailspend has long been a challenge, high volume, low visibility, and difficult to manage effectively at scale. So how are autonomous agents and digital marketplaces helping actually bring greater control and visibility over tail spend? Yeah. I mean, I think you're right to identify it. I think it, you know, structurally suited for automation and autonomy. But something that was hammered into me very early in my career and stuck with me forever, you know, in a complex environment like building a jet engine, you need all the bits. So it can be easy from a human behavior perspective to think about those that considered strategic or those that make the biggest contribution to the p and l. But you need all of them. And that that's where this this can play. And I think one of the thing you know, often often there's a case with your procurement, you can run lots of scenarios, but it's quite flat in its analysis because you never quite know, had you done something different, would you have got a better or worst answer? And there's often lots of interesting debates around what's cost avoidance or what's a true saving or true benefit. I think what we're talking about here is the capability that can bring all of that conversation to life in this kind of supply chain. You know, you you're able to run multiple side by side experiments in, you know, populations relatively low risk, as was alluded to before, perhaps aren't aren't getting, you know, the same level of care and attention, as some other parts of your supply chain. You're able to run them. You're able to see what works. You're able to iterate, and then you're able to scale. And that that's a much richer conversation than I've not really looked at it. I think if I do this, I can say 5%, I'll get a better answer. We're able to say really practically at scale is I have run these changes. This is what the environment has responded with. Some bad, some indifferent. Can let's do that again and see how it works? So I think I think you are able to really get under the skin of. And then from there, you can start to scale, and it becomes less about those lower risk parts. And you start to develop an understanding, and you're able to apply the same methodology to, you know, parts of your supply chain that previously perhaps you felt, you know, you wouldn't want to go there with the technologies. Okay. I'm now I'm now confident, and I've got the ability and capability to to to start looking more broadly. Yeah. I think, echo all of the things that that Ben has said. I think in terms of the solutions giving more control and visibility, I get I think the the benefit is that you're not having to change a lot of the behavior manually. You're forcing the business to act in the way that you, I guess, want them to act, and it's a it's the the thoughtless user. It's not like they're being forced. It's not like they are doing anything that they don't want to. There's very little change management to it because, actually, the options that they're given are suitable for what they need and work in a really nice user experience way. So, automatically, if you're able to address some of the UX, the change management, you're gonna get greater adoption in any case, which then will allow you much more control across your tail where a lot of the users tend to go away and kind of figure that out themselves. If you have good data of your preferred suppliers, if you have good kind of agents set up with the right guardrails to do some of that tail sourcing, if you have the right marketplaces set up, you're giving a lot more options to the users to do the right thing rather than do the the the Maverick thing, that they might have done in the past, which will automatically give you a lot more auditability and control itself. So, obviously, like Ben Oh, so I was gonna just think it is probably just also a little bit of a note caution I'd apply here as well. Maybe the caution is about that critical thinking. The unintended consequence of, you know, overreliance or overoptimization can be, you know, you build in fragility to a supply chain because you've overoptimized for one particular outcome. So that understanding of what you're trying to achieve in the real nuances of it is something to consider. And I think there's also a case that, again, with clarity of thought can be addressed, but you've got to be eyes wide open that you don't lose the capability and the sort of, organizational sovereign strategic clarity that you are making your own decisions. You know, particularly if you think about this and you're thinking about outsourcing this capability to somebody else or somebody else owns the agent, there is a question about how well you understand, you know, what that agent is optimizing for. And are you able to be critical in doing that? And if you do that in an area that's been historically not looked at, I think the point, you know, kind of alluded to it, you get comfortable with a better answer. But, actually, the answer better in fullness, Perhaps it's better for the thing you specified. You know, I want a cheaper supply chain. Okay. It's cheaper. It may now be very fragile. You only see the cheaper bit and you you perhaps can get lazy. I think we're probably all all guilty of this at times, you know, rather than remembering something, you just ask chat GPT. I think there's a little bit of human behavior just to be, you know, eyes wide open to in this kind of engagement. And is that where the human comes in then, that final set of eyes going, actually, that you know, it may be cheaper, but it is only gonna last two months, and then we're gonna be back at square one. Yeah. I mean, I think lots of evidence suggests, and I'd be interested in sort of the of our opinions on this, but lots of evidence suggests that a sort of hybrid person and machine model out prefer outperforms pure automation, especially where it's sort of ambiguity heavy or where the objectives are poorly defined, which in the instance we're talking about would be that case. So I think I think it is from what I've seen and the the evidence and the research I've seen. I think that's where the thinking is that actually where there's complexity, ambiguity, and sort of objectives that haven't been performed as well. That that's where you you really do want to make sure that you do keep a very sort of man machine type and a model working. Yeah. And I I I agree. I think it's a it's a hybrid model, and I guess comes back to the three levels of autonomy and, I guess, what categories of spend or what your risk appetite as an organization in terms of which ones you're a what you know, you are able to bucket into each one, and that might be different. So for example, r and d as a category, you probably wouldn't have ever achieved an optimized autonomy on at the moment because, actually, that requires a huge amount of creative thinking, new initiatives, etcetera. Whereas a category such as, potentially kind of office furniture that is a little bit less risk kind of associated with it, you might be able to put into that optimized economy autonomy with the right guardrails. So it really starts to become a little bit more, I think, coming back to that that prerequisite, it needs to be way more granular in understanding the risk appetite, the level of spend that you're okay for a lot of this to go through, what channels you're okay for some of these to go through. You might want it to be autonomously sourced but not autonomously negotiated. In some of these, you might want it to be automatically purchased or not in certain scenarios. And I think that's where some of the different, I think, iterations and all the journey of the autonomy come in. And I don't think you ever get to a fully optimized autonomy across everything. I think the future is gonna be way more broken up even, as I said, in in sourcing. It's different use cases within within sourcing itself. So alright then. Let's have a real world example. Have we you know, have you have you got an example of a specific case when autonomy works perfectly because it was blended so well with the human interaction and and and with oversight? Yeah. I mean, there are there are examples where, you know, you have agents that are negotiating. You know, you keep the negotiation parameters relatively tight. So, you know, in any negotiation, you're not looking to optimize across several things at once. But, you know, in those instances, you don't you don't try and achieve 15 negotiation outcomes. You look for three. You know, you build some, you know, clear hierarchy between the three. You know, it's kind of quite simple if then type statements. So that that can happen. Contracting is another great example. You know, the the amount of time and effort that procurement and commercial and sales teams spend back and forth on contracts. I mean, everybody's working with a set of set of principles. You're not always seeking the same outcome, but you end up doing the same thing. And and you've seen, you know, great examples where, you know, drafting can be done 95%, you know, kind of by the machine. And you left just with the handful of things that are contentious that the the person, you know, kinda pick the phone up to the account and say, alright, we're stuck on these things. What what creative solutions can we come up with? Whereas your your agent, you know, won't have that real world experience, won't be able to necessarily understand something that it's not understood before. The good thing is that once you've worked through it, you can start to think about codifying, you know, and saying, okay. You now have another option. So if you run into this problem, you know, here's option three. So, yeah, there's a couple of real real world examples where that that could work very well. I think we're seeing more and more use cases work. Actually, I've been surprised as to how many I've seen across different types of clients as well. We've seen it in services. We've seen it in facilities, in, I guess, SKU type kind of consumables type, activities where you can do a thorough comparison of pricing, etcetera, and set the right guardrails around delivery times and all of that stuff. But, also, I was really intrigued to see it in services where, it's on a negotiation front. It's a little bit more complex, because you're, I guess, negotiating on delivery and quality of delivery. But, actually, that's where they've been able to use the kind of levels of autonomy where it's kind of AI assisted. So they've done the quality review by humans and then the pricing negotiation is purely done by the agent itself. So they set the, you know, their best and final internally or or whatever it might be as to their budget price and have used various different levers such as payment terms, working cap you know, contracting terms to negotiate the rest. And, again, it goes back to kind of what is your risk appetite to to allow for that to be done by an agent, and how comfortable are you negotiating payment term, allowing the the the negotiation agent to negotiate on working capital for you and all these lovely things. But it I I think there are proper use cases, at least I'm seeing on the negotiation front. And I think from an sourcing drafting perspective, I think some of the intake orchestration providers linking into some of the time of sourcing tools, also just general sourcing tools, they're able to pull that information, able to draft those RFPs a lot quicker than, I think, what we have seen removing some of the activity. You know, I think one of the biggest kind of barriers to entry of the sourcing tools has been to actually set up that RFP in the category manager, so enabling a lot of that to be done quicker. So I I I yes. I I've drank the Kool Aid, I guess. Definitely seen it working. I I think it's an interesting it's an interesting idea. I mean, I I haven't knowingly negotiated against an agent yet, having deployed agents kind of myself. So it was an interesting idea that some point that, you know, agent is talking to agent. Yeah. You know, without human in the loop in in some instances or, you know, with a level of human steering, and you start to see how this thing can sort of propagate and and grow. And and, actually, if you think about a complex system and sort of the point I was making quite early in the conversation about there's always room for sort of improvement in in those complex systems. This idea that, you know, across a system, not just you you are using an agent, but there's a whole series of agents through a value stream. How and who is setting those kind of overall parameters? Because it can be about well, actually, here, we're having that classic adage of, okay. There's there's lots of waste in the system. We're not negotiating against each other. We're working together to take waste out. And in doing that, everybody gets a better outcome. So you go from somewhere that would be an interesting experience to negotiate, you know, against an autonomous agent. And would you know? I'm not sure. Maybe you would, maybe you wouldn't. And then also say, well, okay. What what what happens when these things start layering and, you know, are are you able to define the outcomes and the objectives differently and and who, in fact, would be in a position to do that? So do you see an agent to agent negotiation as potentially fairer? I think it depends on what what what what they're what they're asked and Yeah. They're often programmed to do. I I think there is an opportunity to take out, you know, some of the the biases, and the sort of personal positions that people may take. But there's also the opportunity to build those in depending on how, you know, how the agent is trained and and sort of learn. So I think I think there is the opportunity for it to be an improvement and therefore drive an overall better outcome for everybody. Or, you know, those with the hypothesis that the agents can improve those negotiations in some instances, those with the better access to agents get better outcomes and, you know, others potentially get left behind. But it's also access to data and who have access to, I think, better datasets and, I guess, more internally cleanse data within those two agents for the agents to actually utilize some of that information that will play, I think, a huge part in how effective some of these agents are. I think the other aspect of, I guess, as Ben mentioned, you've got a lot of different agents across your value chain, and this is where I think the enterprise view comes into play is, I guess, where you have different technologies, you have different agents and different language models talking to one another. And that in its own right, I think, is gonna be a bit more of a challenging kind of, I guess, not challenging, but I guess it's more of a innovative way of thinking about it and has its own challenges, to think about. Whereas I think some of the enterprise having a value chain, a clear vision of what your agent strategy looks like, your AI strategy across your full source to pay suite is really critical because then you aren't as, I guess, thinking about the as much of the agent to agent conversation, them having hallucinations and disagreeing with each other and various different iterations that can occur, less error handling. Those are the kind of pros and cons that you start to think about within, I think, starting to think about the technology that underpins a lot of this itself. Well, you guys have been fabulous, and we are really rapidly running out of time. So I've got time for one more question, and I'm gonna ask you both to give me a thirty second answer. Cool. And it is, what does success look like for a procurement function that has fully embraced autonomous sourcing? I'll let Ben go fast. Thank you. I think I I go back to my point that says, you know, success is a is not a fixed point. It is a continuous sort of evolution. It's that cycle of improvement. You can refer to Deming with plan, do, check, apps, and cycle. I think a team that has embraced this autonomous sourcing approach, you know, is doing that with leverage and capabilities that they wouldn't have had before. But in my mind, they're still doing, you know, those those things. So they're much more focused on the the systems thinking, the constraints, and the continuous improvement piece with capabilities and tools and a breadth and a scale that they didn't have previously. Brilliant. Abby, same question to you. What is success? I think success is being able to drive more value, not just commercially, but, I guess, more on efficiency, time back, regulations, compliance, all of the value, you know, different types of value across utilizing across the source to pay value chain, but with similar with the same resources that you have today and being able to drive value through various different strategic outputs such as thinking about your your support, you know, your supplier relationship management, your strategic suppliers, your, your backup scenarios, a lot of things that we don't tend to get to focus on in procurement and being able to really focus on some of those, those factors to drive value. Fabulous. Thank you so much, Abby. I'm afraid we have, like, eight seconds left, so I'm gonna wrap up. But you guys have been amazing. That does bring us to the end of today's session. We've explored the shift from automation to autonomy, the changing role of procurement professionals, and how autonomous sourcing can unlock both efficiency and strategic value. Autonomous sourcing isn't about removing the human. It's about elevating it. A big thank you to our panelists for sharing your insights and to Amazon Business for partnering with us on today's session. And thank you to all of you for joining and to contributing to the discussion. We hope to see you again at a future webinar, so bye bye for now. Before Amazon Business, buying for work was chaotic. Now it's easy to find products from thousands of suppliers in one place. Save on every type of purchase, from individual items to bulk orders. At any time, you can view your spending on pre built, easy to use dashboards. Plus, you can free up cash flow if you choose to extend payment deadlines and view and approve your team's purchases easily. With Amazon Business, things just got a lot more manageable.