Context for the article – two years ago
In May 2023, I published an article in CPOstrategy on ChatGPT’s impending impact in procurement – my first accurate forecast after nearly a decade of AI “predictions gone wrong.” In 2023, I predicted that generative AI would open a new frontier for data processing, content creation, and decision support, and that procurement leaders must balance its opportunities against its risks.
This article, two years on, is split into two parts:
- Part 1 (“Setting the Scene”) published in June 2025 revisited that May 2023 article and explained the significant AI technological advances across GenAI since.
- Part 2 (“AI-Enabled Source-to-Pay”) this article assesses how those capabilities are impacting the S2P markets, with real-world examples and looks forward another five-10 years.
Two years on – Setting the scene…
Key takeaways from part one for the Chief Procurement Officers (CPOs) summarised from this article:
- Strategic impact of AI: AI will significantly transform procurement functions and fundamentally alter the categories procurement manages. Categories a company procures and how suppliers sell will evolve significantly. Products and services will merge, traditional service models will transform dramatically, and some categories may entirely disappear due to automation and technological innovation.
CPOs should actively monitor these shifts, develop new sourcing strategies to manage emerging categories, engage closely with suppliers to understand their evolving business models, and proactively prepare their organisations for this extensive reshaping of markets and supplier relationships.
- AI technology changes: Rapid advances in AI technology are driving significant improvements in AI performance, notably through increased GPU power, captured by Huang’s Law, which significantly reduces costs and boosts processing speed.
Modern AI now integrates multiple capabilities – perception (recognising patterns and speech), cognition (reasoning, problem-solving), and action (physical autonomy) – resulting in highly autonomous systems. This evolution supports increasingly sophisticated tasks within procurement, enhancing efficiency and effectiveness by automating routine operations, offering real-time insights, and enabling smarter decision-making.
Procurement leaders should prepare for accelerated AI adoption by adapting their technology strategies and aligning AI deployments with their operational and strategic objectives.
- Impact on the supply base: AI technology is dramatically reshaping the supply base by fostering innovative business models and merging traditionally separate categories of goods and services.
Suppliers increasingly leverage AI for product development, service automation, customer engagement and sales, leading to the convergence of physical and digital offerings. Traditional service-oriented sectors will see substantial disruption or potential elimination, replaced by automated and AI-driven alternatives.
Procurement leaders must maintain strong engagement with suppliers, anticipate market evolution, and align procurement strategies and resources accordingly to capitalise on these shifts. Close collaboration and strategic alignment with innovative suppliers will be critical for businesses to remain competitive and responsive to market disruptions.
Quote: “AI will transform procurement — not just how it operates, but what it manages”
Business has access to so much AI…. so what?
As AI moves from novelty to necessity, CPOs face a pivotal moment. This article bridges the strategic horizon set in our last piece with a pragmatic roadmap for action. It helps Chief Procurement Officers translate business strategy into procurement priorities – by identifying where AI can make the most material impact.
The article outlines a CPO-specific method to build an action plan for the next 24 months, followed by a forward-looking view of how AI capabilities are likely to evolve across the procurement value chain over the next five to 10 years. The goal: to equip CPOs with an approach, tools, and confidence to lead AI-enabled transformation, not just in procurement, but across the enterprise.
The first step is to scan the current Source-to-Pay (S2P) technology footprint and introduce the concept of a “platform architecture” to clarify which systems are most critical for your company.
Understand the market – Source-to-Pay capabilities
Before we assess how AI is being ingrained into procurement, it is useful to consider the capabilities in Source-to-Pay by looking at the Source-to-Pay technology landscape as of 2025.
The starting point for any future AI journey is to map the as-is architecture and ensure that you are aware which functions you have covered. Many CPOs will have a well-documented architecture supporting their strategic, tactical and operational capabilities but some will not.
The tables below should enable you to understand what technologies you have and don’t have by capability area (e.g. category management). These tables are a quick scan designed to be illustrative. For reliable view, I recommend tools like ProcureTech or Spend Matters.
Over the last four years, ProcureTech took the time to analyse over 4,000 procurement applications and selected the top 100. Rather than a formal assessment, I have relied on my experience and knowledge, supported by a quick market scan using Google and GPT-based searches to illustrate capabilities that are available in enterprise systems like SAP, Oracle and Microsoft Dynamics. This is an illustrative scan at the time of writing the article and should not replace a formal application selection process.
From this table it is possible to see that the breadth and depth of functionality across ERPs and then deep within individual capabilities through start-ups is significant. The challenge is connecting this together in a meaningful way where the workflow and process and data can be governed meaningfully to create efficiency, effectiveness and insight.
CPOs should map their current architecture and then try to decide which capabilities and processes are most critical, including those systems such as Master Data Governance or Process Orchestration that may be currently managed outside your function.
Setting the right foundation
Map your platform – Introducing the Platform Architecture
When I am designing the technology layer of my client’s to-be operating models, I use tools such as Gartner Pace Layered architecture model to design a Platform Architecture which maps the business criticality and interconnectivity of systems.
The Gartner Pace-Layered Application Strategy provides a structured approach to enterprise architecture by segmenting systems into layers based on their rate of change, business value, and purpose. These systems work at different speeds (pace) depending on the criticality with the top systems having the fastest clock speeds, often real time and the bottom system being slower, transaction by transaction.
The layers shown in the diagram are:
- Systems of innovation – These sit at the top, supporting experimentation and rapid development of new capabilities, either for the business or for procurement. Examples could include experimentation in AI agents, supply risk or sustainability. These systems are agile, insight-driven, and help test or pilot new business models and new technologies.
- Systems of differentiation / transaction – Systems of differentiation provide competitive edge and are tailored to support role-based workflows. They are purpose-fit applications that help shape distinctive business experiences, such as AI agents or supplier collaboration portals. These are supported by SaaS applications and ProcureTech App. This is where most of your companies differentiation in procurement will sit. Systems of transaction are where the process flows sit, in either ERP systems or SaaS applications. Examples include eSourcing, eProcurement and Category Management.
- Master data and integration – This core layer governs the flow of data between systems, ensuring business-critical information (e.g. supplier or item master data) is accurate, consistent, and shared across platforms.
- Systems of analytics and insight – These aggregate and govern structured and unstructured data to generate actionable intelligence. They enable visibility, compliance, and performance improvement through advanced analytics.
- Systems of record – These are foundational systems like ERPs that ensure transactional and financial data integrity for core operations and reporting. This is where the orders and invoices sit that form the basis for tax and financial reporting.
- Infrastructure security and access control – At the base, this layer supports the secure operation of all systems above through resilient, well-governed infrastructure and cybersecurity protocols.
- Cross-cutting all layers is workflow and orchestration, connecting people and processes to ensure seamless, integrated operations. This is often forgotten, or left to a workflow email managed through an email or pop up in an app, which does not connect to people, process and data.
[see attached ppt]
YOUR NEXT MOVE
Document your own company’s Source-to-Pay application landscape, and map this into the pace layered model to create your as is Platform Architecture and then work out where the major gaps are given your business and procurement strategy and your 3-year targets.
Identify the critical master data items that need to be of exquisite quality to ensure your strategy can be executed, as I guarantee that most data is of mixed quality and poorly governed.
Pick the top three areas and obtain funding to experiment in those three areas, ideally including master data and process orchestration.
Quote: “Even the smartest AI won’t deliver value without a connected platform to support it”
If we have all these systems and £ms of investment, why is it still so hard?
Despite years of investment in ERP platforms and a growing wave of ProcureTech innovation, procurement still struggles with core execution challenges. Reliable spend insight remains hard to achieve. Transaction automation is patchy. And workflow orchestration – between users, suppliers, systems, and now AI agents – is often broken or incomplete. These issues are not just technical; they are also structural.
At the heart of the problem lies fragmented and poorly maintained master data and transaction data. Without a clean, connected foundation, even the best AI tools will struggle to deliver value. There is also no true orchestration layer – no connective tissue that spans people, processes, systems, and now autonomous agents. In many cases procurement lacks clearly defined user journeys to design experiences that integrate data, workflows, and decisions coherently.
Now, people are racing to deploy AI agents to automate tasks, answer queries, and drive action without fixing the foundation – data, orchestration, and design. These agents will be operating in silos, perpetuating the same fragmentation they were meant to solve.
Key challenges in summary:
- The strategic elements of Source-to-Pay are knowledge based – which traditional systems and struggle to manage.
- Poor structure and governance of master and poor transaction data – partly as supplier invoice data is not captured and cleansed.
- Poorly designed procurement architecture with an absence of process flow – systems are designed technically, not with business process in mind, and around the “happy day” scenario and not real life.
- Lack of an orchestration layer to connect people, processes, and systems – Tools like Zip and ORO are starting to address this.
- Absence of well-defined user journeys for process integration – even if you have a well-considered architecture and an orchestration layer, unless you design by Person/real user journey, usage is likely to be poor, and you risk missing the other elements such as policy and change management.
Identify your priorities
How AI can help solve these fundamental challenges?
1. The strategic elements of Source-to-Pay are knowledge-based – traditional systems struggle to manage this effectively.
How AI helps:
Some tools (as listed in Table 1) already support parts of strategic procurement – like category planning or supplier risk – but they are often siloed. AI co-pilots help connect these systems and bridge gaps where coverage is weak. For instance, an AI agent can combine sourcing data, internal strategies, and supplier insights to draft a category plan or negotiation brief. They also capture and reuse tacit knowledge from senior managers, making it more accessible and durable.
In short, AI augments existing tools and fills blind spots, helping teams tackle strategic work with more speed, consistency, and context. This should be more than your category managers writing strategies or analysing data with Copilot.
2. Poor structure and governance of master and transaction data (especially invoices)
How AI helps:
AI models, especially those trained on financial documents, can extract, clean, and standardise data from unstructured invoices, emails, and legacy systems. Some specialist Accounts Payable packages like Basware move you a long way forward extracting that data but there is still more to do to get the full level of source data in the systems of transaction or record.
ML algorithms can detect master data duplicates, classify suppliers, and identify anomalies in spend data. Over time, AI can automate parts of data stewardship, helping improve master data quality continuously, rather than relying on periodic clean-ups.
This is the low-hanging fruit that will unlock everything else. If I were a CPO, I’d start by using AI to fix master data and extract the transaction data we already have.
Quote: “If I were a CPO, I would prioritise AI to fix my data”
3. Procurement architectures lack true process flow and are built around technical assumptions
How AI helps:
Tools like Celonis and AI agents can highlight and resolve process friction by analysing user behaviour, exception rates, and bottlenecks across the S2P lifecycle. Large language models can simulate real-world ‘unhappy day’ scenarios to test resilience and spot design flaws and allow you to manage them on an ongoing basis. Over time, this enables more iterative, user-informed design rather than rigid technical blueprints.
Beyond this there is no alternative but to implement the core components required across the Source-to-Pay lifecycle. Some believe that agents will fully circumvent the need for Source-to-Pay packages, but I think this may still be five years away – we still need applications to manage access, data and integration. Increasingly through systems of differentiation and transaction will move into systems of record and some differentiation will move into the agentic or orchestration later.
4. No orchestration layer connecting people, processes, and systems
How AI helps:
AI agents can serve as a connective fabric across fragmented systems, acting as orchestrators of micro-tasks – routing information, prompting next actions, and maintaining flow across different tools. Platforms like Zip and ORO are embedding AI to automate intake, triage, approvals, and supplier onboarding in a coordinated, responsive way.
AI agents, which are autonomous systems that initiate and perform tasks or decisions on autonomously or supervised on a user’s behalf.
5. No well-defined user journeys – hurting adoption and change
How AI helps:
AI tools can help map and simulate user journeys based on behavioural data, identifying friction points and usage drop-off. AI-driven design assistants can also support rapid prototyping of procurement interfaces and workflows tailored to specific personas (e.g., requester, budget holder, category lead). With the right prompts, AI can even generate policy summaries or training content to reinforce process adoption.
All that said, there’s no substitute for well-run design thinking sessions—bringing users together to identify issues, explore solutions, and align across both logic and emotion. AI can provide inputs, and draft great personas, but the well facilitated design thinking will allow you to align on logical rational (what should we do?) and political and emotional (why should we care and will we bother?).
So, you have a sense of where you are, what your priorities are and where you need to focus. You should have a sense of where you need to build the foundation, such as data or extend applications or invest in AI.
YOUR NEXT MOVE
Given your existing footprint and architecture, your priorities and the gaps you wish to close out you can now built a to-be architecture and transformation plan such as the one below to close out the gaps.
Figure 1: Example Transformation Map
That’s the next two years taken care of, but what about 5-10 years away?
Procurement in the future
Five years out
In five years, AI will fundamentally reshape the foundations and flow of procurement. AI agents will be native and human and autonomous communication hard to differentiate. Machine to machine collaboration will happen at scale, hopefully with the right guardrails.
How AI Will Help:
Master data will shift from static, error-prone records to continuously updated, self-healing datasets. AI models will continuously detect inconsistencies, enrich attributes, and reconcile inputs across systems in real time.
Over time, this will evolve into a continuous sensing and real-time response capability. These will synthesise internal data, market intelligence, and supplier insights to develop category strategies, evaluate risks, and generate negotiation playbooks. This shift will free up teams to focus on judgement and stakeholder influence.
Strategic procurement will be transformed through AI co-pilots that synthesise internal data, market intelligence, and supplier insights. Over time, this will evolve into continuous sensing and real-time adaptation.
Transactional procurement will become near touchless. AI agents will triage intake requests, validate POs, auto-match invoices, and route approvals based on context, not rigid rules. These agents will interface directly with both internal systems and external supplier platforms, enabling seamless, multi-agent orchestration across organisational boundaries.
Tomorrow’s insights won’t be extracted – they’ll be embedded and automatic. AI will analyse activity, exceptions, and trends continuously, surfacing risks, savings, and opportunities as prompts – tailored to each role, process, or decision point.
Those companies that do not respond these challenges may face fundamental disruption from competition and new entrants.
Quote: “Insight will be embedded, not extracted”
Supplier-side AI agents will play a key role. Acting on behalf of vendors, they’ll negotiate, confirm orders, and provide real-time updates – creating a dynamic, responsive ecosystem of machine-to-machine collaboration. These future demands readiness: cleaner data, better design, and trusted orchestration to make it real.
10 years out
10 years from now, procurement will be radically transformed – fluid, autonomous, and deeply embedded in business value creation.
To understand the scale of change, think about the world pre- and post-Amazon, Uber, or ChatGPT. Now imagine all those shifts happening at once. That’s where procurement is heading. I believe that 10 years from now, we’ll be living in a post-transformation world – one with five times the technology, twice as smart, and twice as effective.
Quote: “Procurement is shifting — from human-led coordination to AI-fuelled orchestration”
How AI Will Help:
Master data as we know it will disappear. In its place, AI agents will maintain a dynamic, real-time digital twin of the supplier ecosystem—drawing on contracts, transactions, third-party data, and direct machine-to-machine signals.
Strategic procurement will evolve into continuous sensing and response. AI agents will monitor markets, regulations, supplier health, and innovation trends 24/7—generating options, not just reports. Strategic procurement will evolve into continuous sensing and response. AI agents will monitor markets, regulations, supplier health, and innovation trends 24/7 – generating options, not just reports. Category strategies will be living frameworks, adjusted in real time by autonomous systems collaborating with human oversight.
Transactional procurement will be fully autonomous. Buyers will define outcomes, and AI agents – both buyer and supplier-side – will negotiate, place orders, manage compliance, and resolve exceptions through digital dialogue, without manual touchpoints. Approvals will be trust-based, outcome-led, and exception-driven.
Insight won’t just be surfaced – it will be predictive, prescriptive, and self-acting. AI will not only tell you what to do, but initiate action unless told otherwise.
Supplier-side AI agents will be indistinguishable from their human counterparts – negotiating, customising, and troubleshooting live in digital environments. The procurement landscape will shift from human-centric coordination to AI-led orchestration, demanding new skills, governance models, and trust infrastructure to thrive.
In conclusion
Two years ago, when I first wrote about ChatGPT’s potential impact on procurement, I called it a new frontier. At the time, many of us were still asking whether AI would truly change the way we work. Today, we know the answer. It already has.
What I have seen since that first article has only reinforced my belief that procurement is on the edge of a profound transformation. AI is moving faster than expected, and the capabilities we once imagined for the next decade are already starting to appear. But the pace of progress is not the only story. The real challenge is what we choose to do with it. Many CPOs are still at the starting line, grappling with foundational technology issues that predate the AI wave.
This article has outlined a practical path for Chief Procurement Officers to follow. Map your technology landscape. Improve your data quality. Establish orchestration across systems. Prioritise investments with a clear line to business value. Most importantly, do not delay. The future of procurement will not be built solely on tools or technology. It will be driven by vision, leadership, and the willingness to rethink how procurement creates value.
If 2023 was the year to ask what AI might do, and 2025 is the year to decide how to act, then I believe 2027 will be the year AI-first procurement operating models start to scale. Procurement teams will begin to see AI agents working across buyer and supplier ecosystems. Autonomous sourcing cycles will become common. Strategic decisions will be supported by co-pilots drawing on structured data, supplier intelligence, and real-time risk signals. Category strategies will be dynamic, constantly updated in response to market and performance data. Procurement will become less about managing steps and more about delivering outcomes.
If you are not a CPO but work in procurement, this is still your call to action. Start learning everything you can about AI today. Understand how it works, where it applies, and how it fits into your role. This shift is already redefining the profession. The gap between those who are AI-literate and those who are not will widen quickly. There is no time to wait.
The groundwork for the future must be laid now.
If this perspective resonates with you, or if you are looking for someone to help you shape the first steps, feel free to reach out. I am always open to a conversation.
Let’s lead the next chapter of procurement together.