Introduction: From Design System to Presentation Workflow

Okay so I wasn't planning on talking about clod design when I came in the room but then a couple of people have been chatting about stuff so there are not many clod design front end experts because it's been live for 10 days so I might be an expert in this but so essentially what I'm going to one of the reasons why I'm sort of showing this is because the whole presentation that I've done has been based off a clod design river design system and essentially I took some

some consulting output from one of our customers, dropped it in, put the UKII branding in and built a presentation, didn't like the way that the speaker notes worked, so created my own system for that, so I have speaker notes here for me, which actually move with the conversation as well, so it's quite a useful thing to do.

So within this design system, essentially, I'm just going to quickly show you it, but you know, actually, do you know what? This is going to be really difficult with a microphone phone in my hand can we still hear if I'm just generally talking like this is that okay yeah so

just really quickly you know in a design system if I I should really have a mouse here but no I'm not going to do it hang on this is me being terrible with an actual yeah yeah exactly exactly so I'm just going to move straight on so essentially what what the design system is and And in fact, I'll be better off just explaining it.

But you construct a design system and Anthony was saying, I can't do anything with Figma. I dropped my Figma design into the design system. So that's pretty useful, which gave me a basis

of everything that we've been designed, has been designed by in Figma by our marketing agent, dropped it in and now I can create presentations in our style, in anyone else's brand colors,

drop my own consultancy notes into it, creates the entire presentation for me this one took an hour so that's the that's the definition of being able to move quickly um but i'll just uh go go into this anyway but that so i was going to

give a little bit of a demo but it's me messing around with a trackpad i hate trackpads always have done for all my life and i forgot my mouse today um so let me just make sure this is working

Case Study Context: AI for High-Speed Procurement

yeah okay so what we so one of the things that we oh thank you that's really and one of the things that we have worked with and I cannot say the client name however it may be on the video that has been done which is one of our guys has done and so please shut your eyes and don't mention

it again but what this is what we're talking about is how modern work is going to happen happen.

This is an AI system built for a race team. It's like the future of work will be compressed.

So what we're looking to do is give humans plus AI. There's some very, very expensive people that work in certain places. And what we want them doing is actually the

human value work that makes them amazing. And what we don't want them doing is things that don't, that are, you know, what us humans call work, but it's not really work.

It's inputting stuff into stuff that is not something that we want to do so you know and i suppose in in a in a very fast -paced environment it's actually quite a good use case to be teaching everybody

The Real Bottleneck: Trapped Intelligence and Manual Work

so essentially so the problem is not a lack of intelligence all the intelligence that they have to work with all of the problems is there it's just trapped it's trapped behind copy and paste 78 emails, status chasing, re -keying data, 12 tabs open or in my case 372 tabs open and some manual approvals etc.

So the experts within the system already exist. They're paid for judgment.

Nobody pays me to type an email. Nobody pays me for that. People pay me for the 20 years, 30 years of experience I've got of actually delivering solutions. The actual email itself, the document, the sales proposal, nobody pays me for that.

So, how do we utilize systems to make sure that we're taking away all of that work, which we call work which isn't really work, and actually turn it into a system. So, in

this case, speed is a competitive advantage. The moment you get a part to track is the moment it's in a testing window and that could mean championship championship points.

Goal: Human + AI to Create Speed and Capacity

So we're not going into an organization going, oh, how do we get rid of some people? We're going, how do we make your organization better? How do we look at the strategic advantages that you can get from speed, delivery, utilizing your people better? What can we do? And there are many advantages.

But a lot of companies that we have spoken to, one of the the first, I need AI is the first thing, but what? But not just I need AI, I need AI to replace these 10 people. Okay, that's nice. Okay, brilliant, but not nice really at all. I don't like that conversation, but it's about giving people capability and time back.

We are time generators and from that time we can generate some pretty mean KPIs. So essentially it's about getting

things faster for the race weekend so the official system is there but it's never the whole truth so we've got ERP system CRM purchase orders contracts all in this got what we call a golden data set so within the golden data set and

they are incredibly important data is fundamentally important but what happens where actually work actually happens and when you know I get apologies from people who we talk to and go oh we use this spreadsheet brilliant you're using a spreadsheet that means your system of your system of truth isn't giving you what you need and where work

actually happens is in those conversations those judgment calls those the memory of someone who has done something and it's a lot of this is in unstructured detail in email and actually the The only way of memorizing that is actually search your emails or potentially, you know, go and talk to someone who's worked on it. And actually, all of this intelligence can be captured at point and still leave the judgment calls to the actual buyers themselves.

So, you know, and there are a load of things, exceptions that happen last minute. A supplier doesn't deliver. All of this intelligence is lost. It's never in the official truth. It's never in the golden data set.

So this is where we look to implement AI into an organization is actually, it's not just going, let's do work like we used to do. How are we going to improve the intelligence of your organization?

The Agentic Approach: “Ralph” as Chief of Staff

So this one is called Ralph. I do like to name my chief of staff agents. So that is the lead agent that is looking over all of the other agents.

There will be many other agents. There's a whole table of agency is what we call it underneath this. and they will be doing multiple things, which I'll come on to, but Ralph is going to run the organization.

He will be the person that the buyer would speak to via Teams or email or however it is that they want to consume this. We can create a chatbot if you really want to, but how AI is best consumed is how people use it, so the design needs to be fundamentally put in front of people,

And if actually your suppliers use email and you use email, you need to be able to interact in the tool of choice that you're using. So in this instance, email was one of the key things and Teams is also one of the key things for helping with Ralph. But, you know, he's got to be dependable, wise counsel. That's one of the reasons why we chose that name.

Core Capabilities: RFQs, Quotes, and Comms Automation

But one of the key things is it wants to draft, send and chase RFQs. All RFQs were in emails. There was no process to capture them.

they need to be built and drafted within the system you know all those quotes coming back off those RFQs they are actually all part of the system as well you know it's all completely auditable and this is where we come into what we call AI orchestration layers and control planes

so many of you may not know what that that is right now but essentially it's the layer of of control between you and your agents and then further extrapolated your LLMs and what they are doing because you absolutely need a governance and audit trail, especially in a large organization

like this one, to make sure that decisions that are being made are fully auditable, fully governable and are not going to cause that organization some trouble if stroke when when absolutely something goes wrong.

So, and the key, one of the key factor is

human plus AI is fundamentally the thing that we press forward with the most. It's that human in the loop.

The buyer must keep the call. He must be able to keep the approvals. He must make that call.

So, and I think that's one of the fundamental key parts of what we're building.

Architecture: Orchestration Layer, Control Plane, and Governance

So, when I talk about the orchestration layer, this is this middle layer.

So essentially, we have all the inbound systems, the SAPs, the SharePoints, the GDrives, the 3DXs, the Teams, the M365. They are our data sources. They're also places where things can work or work happens.

We built three intelligence zones. So how do we create a set of agents which build buyer intelligence?

The data store that remembers everything and we want to create demand intelligence so one of the things which

Three Intelligence Zones and the BOM Extractor

was quite key was the bomb extractor what they used to do is wait right till the end and then they'd order the parts post the after extracting data from the bomb we've brought the bomb all the way to the front because it really never changed before approval so actually they can start ordering parts probably three to four weeks earlier than than actually when they were ordering parts before

that's a massive time saving they thought it was a boring task i thought it was a we thought it was a fundamental task for making an ai system work and one of the massive key things is about any ai system is how the buyer how profiling a user so in this system buyers don't just use it finance people from finance use it so how they and that data needs to be presented back to them in the

the same way so what we're doing is taking multiple data sources utilizing skill these are probably you could probably say they're agents or even skills some of them are ai some of them are just technology and workflow and and the thing that you find with most deterministic stroke non -deterministic systems obviously ai being non -deterministic workflow being deterministic is it's about 85 to 15 is what we see and when people say we want AI what they actually a lot of organizations want is just a better system and we use AI to improve that system fundamentally

and then we've got comms intelligence so this is you know the things that comms intelligence is chase I'm gonna lose it no surely not yeah we go fundamentally comms intelligence is very much about getting rid of those pesky tasks like who wants to who wants to send an email that chases someone down for a specific part none of that needs to happen so but so all of these intelligence

zones are part of the orchestration layer which then fundamentally creates an intelligence layer

now stock orders and suppliers already existed but what we have created on top of stock orders and suppliers is this sort of bomb summary which is what happens when drawings come through which builds intelligence for pre -stock orders the rfq process which was which wasn't there and quotes

so with quotes you can manage things like then you have no idea how many quotes have come through where um people had quoted late or not quoted at all um you can actually get quite a quite some quite good intelligence about who your best quotas are and how fast they are

And actually, some of these quotes were varying in price as well from one quote to the next because they were maybe trying to, you know, big organization, let's pop our prices up. Now we can see it.

Demo Walkthrough: River AI Agent Dashboard

And then actually what we're going to just quickly, what I'm quickly going to demo, thanks to Lee over there, is the River AI agent dashboard for this specific tool. It is not live yet. yet. It's fundamentally working on the test data.

So, obviously, we've got to go through the big system integration parts of it.

But fundamentally, what we've done is actually we can extrapolate this entire agent system away from a horrible data source and actually build an agentic system, which is still going to help them. And we don't even have to plug

it in. Apart from the email side, that's the only place where we have to plug in.

So what what I'm going to show you is the dashboard, but a lot of the interactions are through Teams chat and outbound email.

The actual AI part of a lot of this is quite straightforward.

And one of the

things that is fundamentally difficult about any AI system and any organizational AI system is integrations, is Teams. We literally had a conversation today about, well, what does our

Integration Reality: Process Redesign Over “Just Adding AI”

sap team need to do and actually we just say oh we we need sign off from the sap team so i don't need to do anything just send these reports to here and then we can build we can do anything from here and i think one of the fundamental things about working with these guys was they didn't realize that literally we're just redesigning your entire process about how you work and that's

been one of the fun one of the fun things so it's like sort of one store many agents but and three types of intelligence all being surfaced into one but I'm gonna I think the next one is so yeah so buyers actually stopped becoming coordinators because that's what they were they were just shifting pots from one place to another to a to b they actually become specialists again and that

how enthused those guys are to actually do their job and talk to suppliers and not have to do some Some of the things, you know, that I've popped on here, you know, that they'll be using their judgment all the time. Their job doesn't shrink. It becomes better.

It becomes doing the things that they originally wanted to be, being in purchasing. You know, it's actually a lot better than, you know, we're actually improving their jobs. There was, there's always worry of people going, and we always go, what would you do if you did have time? What can you improve?

What can you do better? you know what projects can you get involved in and they were like oh we've got loads like we'll do that then you know don't worry about losing your job worry you know get excited about what your

job's going to become that's where we need to be and all this searching and retrieval and drafting and re -keying that take all that away what ai is going to be fundamentally brilliant at is taking away crap work pardon my French and and that is going to be fundamentally a great change and then just rethinking about what we can do with our

time is going to be great and so this is another a good a good example of Claude Zhang because this this demo wasn't built in five minutes before I did this so hopefully this works but it nicely slapped the video straight in the middle of it so let's hope it does um so let me just get on here oh no i don't know that there we go

can i press play hopefully the audio is on welcome to the mercedes formula one's procurement intelligence platform a system engineered to bring speed precision and ai powered efficiency to formula one supply chain operations obviously we had to do it in ai team members securely access the platform, where every workflow is designed to support Mercedes' mission to chase every millisecond.

Once inside the dashboard, Mercedes gains a complete real -time overview of its procurement ecosystem. Front and center is the countdown to the Miami Grand Prix, just five days away, reinforcing the team's philosophy to chase every millisecond.

Beneath this, key operational metrics are surfaced instantly, including total RFQs, active supplier quotes, procurement human spend, and AI -powered BOM extractions, giving decision makers immediate visibility into performance.

The RFQ and quote trend analysis provides a clear picture of sourcing momentum over time, while the needs attention panel highlights urgent supplier risks, delayed responses, and cost anomalies before they impact performance.

Quick actions streamline critical workflows, allowing teams to rapidly create RFQs, launch BOM extractions, or review supplier quotes with minimal friction.

Finally, the recent activity feed ensures every procurement movement is tracked, from quote submissions to AI extraction results, creating a centralized command center for Formula One supply chain precision.

This dashboard is designed not just for oversight but for speed intelligence and competitive advantage along the left -hand side the platform is organized into the key operational areas that power mercedes procurement workflow the dashboard acts as the team's central command center delivering a live overview of procurement performance supplier engagement and race critical operational priorities

RFQs manage the full Request for Quotation lifecycle. Within this module, teams can oversee all RFQs, monitor drafts, track sent requests, review incoming supplier quotes, identify RFQs that are ready for action, and rapidly create new sourcing requests.

The BOM Extractor is Mercedes' AI -powered document intelligence tool, enabling users to launch new bill of materials extractions from technical documents while maintaining access to historical extractions for traceability and analysis.

Quotes serves as the supplier quotation hub, where procurement teams can manage incoming quote submissions and perform detailed quote comparisons to identify the best commercial and technical outcomes.

Suppliers provides full oversight of the vendor network, ensuring supplier relationships, capabilities and sourcing options remain optimized across the supply chain.

Parts delivers detailed access to component -level data, helping the team maintain precision over specifications, availability, and procurement requirements.

Purchase Orders controls final procurement execution, with Draft Purchase Orders allowing teams to prepare, validate, and manage orders before deployment.

Together, these modules form a fully integrated procurement platform, designed to increase speed, improve visibility and ensure Mercedes can operate with the precision required to chase every millisecond.

Here we arrive at one of the most powerful capabilities within the platform, Mercedes' AI driven BOM extractor. This is where engineering documentation is transformed into procurement intelligence.

Teams can upload technical drawings, BOM tables or specification documents directly into the system, where advanced vision and OCR models immediately begin passing complex engineering data.

The AI then performs structured extraction, identifying critical information such as part numbers, quantities, materials and manufacturing processes, converting unstructured documents into clean, actionable JSON data.

But this goes goes beyond extraction alone. The platform also performs supplier matching, cross -referencing extracted components against Mercedes' supplier network to rapidly identify suitable sourcing partners based on capability and category alignment.

What traditionally required hours of manual review can now be completed in moments, dramatically accelerating procurement speed, reducing human error, and creating an intelligent bridge bridge between engineering and supply chain operations.

For Mercedes, this represents more than automation. It is AI deployed with purpose, delivering the speed, precision and competitive edge required to chase every millisecond.

Okay, so thank you for that. And thanks to Lee for putting that together.

What the Product Really Is: A Unified Window and Audit Trail

It is a window. Two minutes, yeah. It is a window. So what we've rebuilt is a window.

Everything, the product that AI brings is the audit trail. The audit trail of decisions, of the humor interactions, of bringing that all together to be one record.

We are not replacing SAP. We are not replacing email. This is the window where it comes together and that you can make sensible sourcing decisions.

When we've spoken to people, it has been, what does it not do? What can you do? A lot of the time, and we have done this, we've gone, well, what does your current toolset do that you aren't using?

Because, you know, I spoke to an e -commerce company and basically they just weren't using Shopify properly. So, you know, don't engage with us until you're using that properly and then we'll build something on top of that because actually a lot of their functionality within Shopify well we're not going to rebuild that you know there's plenty of things that we can

build that are not part of products so this is just a purchasing window which is like really every every record everything is a record every override every order and and and and it builds a history so you know and and it gives you permission to move quickly because you you understand it um and so just finally i think

Conclusion: The Future of Work as Augmentation

the future of work this this we are augmenting these guys we're giving them more capacity the thing the things that they will do to they will speak to customers more often they will speak to suppliers more often they will do deals they will get parts to parts to track more often what they

They will do what humans do best, monitor anti -patterns, deal with errors and frustrations and problems. That's what we are brilliant at.

We are not particularly brilliant at typing or re -keying data or doing boring tasks. We're not here for that. We didn't get put here to be boring. We want to do all the exciting stuff.

This boring stuff can be done by someone someone else and yep and so that's it we've reached the checkered flag thank you very much so cheers