A Wake-Up Call for Legacy S2P: The Market Has Already Moved On
Source-to-Pay isn’t going away, but the center of gravity is shifting from monolithic suites to an AI-native experience + orchestration + data layer. The winners will be the ones that feel like a consumer app on the front, run on agents in the middle, and plug into anything at the back.
1. Macro shift: from “suite” to “AI-orchestrated ecosystem”
A few clear trends have crystallised in the last 18–24 months:
Intake & orchestration as the new “front door”. Players have grown fast by owning intake-to-pay: a single front door for any employee request, then orchestrating the right process across ERP, CLM, AP, risk, etc. (CPOstrategy). Their 2024–25 positioning is explicitly about procurement orchestration rather than just P2P. (Business Wire)
Agentic AI as the execution engine Instead of passive analytics, we now see AI agents validating PRs, matching invoices, parsing quotes and even driving sourcing events.
Predictive / prescriptive sourcing uses AI + game theory + behavioural science to predict the best suppliers and prices and pre-populate events, reporting ~18–19% average cost reductions and attracting fresh funding in 2025. (Arkestro)
AI-native S2P positioning from the big suites themselves SAP is literally marketing “next-gen SAP Ariba” as the first AI-native source-to-pay suite, tightly coupled with its Joule copilot. (SAP) Coupa is pushing its own AI agents (Navi, Contract Intelligence, etc.). (Procurement Magazine) Ivalua’s roadmap is now heavily AI-themed as well. (CPORising)
So the “legacy vs AI-native” divide is already blurring: the traditional S2P players are trying to reinvent themselves, while new entrants are starting from AI, UX and integration as first principles.
2. What an AI-native S2P stack will actually look like
If you project 3–5 years ahead, the reference architecture looks more like this:
a) A single conversational intake layer
Any user raises “I need X / I want to work with vendor Y” in natural language.
A gen-AI layer: This is exactly the space Zip and similar platforms occupy today, and what many AI guides describe: gen-AI classifying intake and auto-generating the approvals path. (Spendflo)
b) AI agents running processes end-to-end
Think of small, specialised “colleagues”:
PR validation agents: checking GLs, budgets, policies, preferred suppliers. (GEP)
P2P agents: 3-way matching, compliance checks, exceptions handling, some vendors already market specific “procure-to-pay agents” that perform full invoice–PO reconciliation. (V7 Labs)
Sourcing agents: reading demand, suggesting event design, inviting suppliers, scoring bids, flagging risk (Arkestro, and early work by various suites). (PR Newswire)
The human’s role becomes "setting guardrails and resolving edge cases", not driving every transaction.
c) Predictive & prescriptive analytics by default
Future S2P will assume:
Continuous price / risk / ESG benchmarking from multiple data sources (networks, external data, supplier docs).
“Next-best-action” everywhere: which category to source, which supplier to renegotiate with, which payment term to change, which risk issue to escalate.
Vendors like Arkestro and a growing set of AI tools are already selling this as “predictive procurement orchestration.” (Arkestro)
d) Embedded copilots in every module
Instead of separate “AI modules”, you get copilot bars across the suite:
“Draft an RFP based on last year’s event, but incorporate ESG KPIs X/Y.”
“Summarise this contract and highlight non-standard clauses vs playbook.”
“Show me why Marketing’s spend is above budget and propose mitigations.”
SAP Joule, Coupa’s Navi, and Ivalua’s AI roadmap are all moving in this direction. (SAP)
e) Network & data flywheel
The most defensible advantage is becoming:
Data fabric (clean vendor & spend data from ERPs, networks, third parties).
Network effects (Ariba’s network for supplier discovery + AI to match and assess risk, for example). (Lumi AI)
This is where legacy players have a structural edge if they modernise the UX and AI layer fast enough.
3. What this means specifically for legacy platforms ?
They are not “dead tech” – but they face real pressure.
Strengths:
Full S2P depth (complex categories, multi-ERP, governance, auditability).
Global customer base and rich transaction history.
Strong partner ecosystems, implementation know-how and SI familiarity.
Pressure points:
UX & time-to-value: new AI-native tools show that intake workflows can be deployed in weeks, not months, with consumer-grade UX and low-code configuration. (CPOstrategy)
Composability: customers increasingly want “anchor + best-of-breed” (e.g., keep Ivalua/Coupa core, but use specialist agents for sourcing, intake, CLM, etc.).
AI differentiation: if everyone has a copilot, the real differentiator becomes:
Expect:
More emphasis on platform / iPaaS capabilities (events, webhooks, open APIs, marketplaces).
Acquisitions of AI-native players (intake, predictive sourcing, risk analytics).
“AI tax” in pricing models – but also more value-based or outcome pricing (savings-share, working-capital improvement, etc.).
4. The challenger pattern: AI-native, UX-first, narrowly opinionated
If you look at the disruptors:
Intake & orchestration: AI-powered intake-to-pay, strong UX, and now an AI Agent Builder for no-code creation of procurement agents.
Predictive sourcing: predictive procurement layer that you put on top of existing ERPs/S2P, rather than rip-and-replace.
Specialised agents: vendors and even horizontal AI companies launching specific “P2P agents” and “sourcing agents” that integrate via APIs into whatever S2P you already have.
These challengers push the market to be more client-centric: quicker wins, nicer UX, less process worship, more business outcome language.
5. From an end-user point of view: what “good” will feel like
For the business user in 3–5 years, an S2P environment that’s “state-of-the-art” will feel like:
One simple entry point (web, mobile, Slack/Teams) to ask for what they need in natural language.
Minimal visible process: policies applied in the background; the system tells them what’s allowed and offers guided choices rather than a maze of forms.
Fast cycle times driven by agents doing the grunt work (3-way match, chasing suppliers, consolidating quotes).
Continuous guidance for buyers & category managers via copilots suggesting events, negotiation levers, and risk mitigations.
Real-time cockpit for leadership: AI explaining variances (“why did Marketing spend jump?”, “which contracts are value-eroding?”) and simulating scenarios.
If your current S2P stack doesn’t move materially towards that experience, it will feel legacy regardless of the logo.
6. So what should a CPO / CFO / CIO actually do with this?
Here’s a practical way to position a roadmap:
a) Think “Anchor + Edge”
Anchor = your core S2P (Ivalua, Coupa, Ariba, GEP, Jaggaer, etc.), where you get control, compliance, and scale.
Edges = AI-native capabilities you can add or swap:
This lets you modernise user experience and AI without immediately ripping out the core.
b) Change your RFP lens
Instead of just function checklists, evaluate vendors on:
AI operating model
Orchestration & openness
Experience & adoption
Data strategy
c) Design for “replaceability”
Whatever you pick now, assume:
You’ll want to swap components (e.g., today’s orchestration layer for tomorrow’s better one).
Some AI agents will be built in-house or by your SI using open platforms.
So: insist on open standards, APIs, and data portability upfront.