Mor Cohen-Tal

How Do You Switch From a Legacy Procure-to-Pay System to AI Without a Second Migration?

A checked gear with upward arrows on a teal honeycomb background, showing a switch from a legacy procure-to-pay system to AI without a second migration

Picture a controller on the second day of month-end close. The ERP balances, as it always does, but the numbers behind it took a week of chasing: a purchase approved in a chat thread, an invoice that arrived before anyone raised a PO, a commitment nobody logged until the bill landed.

Then someone in a planning meeting suggests the fix is a new procure-to-pay suite, and the room goes quiet, because everyone remembers the last migration. The ledger is not the problem, and replacing it again feels like paying twice for the same lesson. What finance, IT and procurement actually want is AI in the flow of work, without starting over.

There is a way to get there that begins with the systems you already have.

You switch from a legacy procure-to-pay system to AI by adding an orchestration layer above the ERP and CLM you already run, instead of replacing them. Intake, approvals, purchase orders and invoice matching move into one AI-driven flow, while your ERP stays the financial system of record, which is why the switch does not require a second migration.

Mor Cohen-Tal, Co-Founder and CTO of Opstream
By Mor Cohen-Tal, Co-Founder and CTO
Co-Founder and CTO of Opstream, previously Cloud CTO at Turbonomic (acq. IBM for nearly $2B) and holds 13 patents in cloud and AI infrastructure.
View LinkedIn profile →

Why do legacy P2P replacements turn into a second migration?

Most teams that want AI in procure-to-pay hear the same answer from the large suites: replace your P2P. On paper that sounds like modernization. In practice it means a new data model, a new vendor master, re-implemented approval chains and a migration of open purchase orders and contracts, and for many companies it is the second time in a decade they have moved that data.

That would be easier to accept if large replatforming projects reliably delivered what they promised, but the track record is not encouraging. Gartner research published in May 2024 predicts that by 2027, more than 70% of recently implemented ERP initiatives will fail to fully meet their original business case goals. The same research cites a 2023 Gartner survey in which 87% of respondents from organizations with implemented ERP applications planned to replace or upgrade them within three years. Read together, the two findings suggest that another replatforming project is a risky road to AI.

The deeper issue is that the ledger is rarely what is broken. What breaks is everything around it: requests that start in email and Slack, approvals that live in someone’s inbox, and contracts that sit in a separate tool. Deloitte’s 2025 Global Chief Procurement Officer Survey found that siloed ways of working (57%) were the barrier CPOs ranked most often as preventing value delivery. A new suite does not fix silos if finance, legal and IT still work outside it, so the more useful question is not which system to replace but what should sit on top of the systems you keep.

The problem is rarely the ledger. It is everything around it.

What does an orchestration layer above your ERP actually do?

An orchestration layer sits between the people who ask for things and the systems that record them. It does not replace the ERP; it feeds the ERP clean, approved data. In Opstream, that layer handles four connected jobs, and they make the most sense when you follow a single request through them.

Every request starts at one front door. Adaptive Intake turns plain-language requests into structured, policy-compliant submissions by pre-filling fields from multiple data sources, so requesters do not have to know which form to use and approvers do not have to chase missing details. From there, rules route each request by spend threshold, vendor risk, category and compliance requirement, with legal, security and finance reviewing in parallel rather than waiting on one another.

Once a request is approved, it becomes a purchase order with its financial context attached. When the invoice arrives, it is three-way matched and scored by match confidence, so clean invoices move on and exceptions go to a reviewer. Throughout the flow, validated transactions sync into the ERP through native connectors for SAP, Coupa, NetSuite, Workday, Microsoft Dynamics and JD Edwards, with bi-directional sync and connector monitoring owned by Opstream’s engineering team. For more on why that integration depth now decides most evaluations, see why ERP integration is now a deal-breaker.

The difference from a rip-and-replace project is easiest to see side by side.

QuestionRip and replace a suiteOrchestration layer above your ERP
What happens to the ERP?Often re-implemented or heavily reconfiguredStays as the financial system of record
Data migrationVendor master, open POs and contracts moveData stays where it is and syncs both ways
Who changes workflows?Usually IT or an implementation partnerBusiness teams, with no-code configuration
Where legal and IT workOften in separate toolsIn the same request and approval flow
Time to first valueTied to the full migrationTypically live within 30-45 days

A layer that sits above the ERP raises an obvious question from finance: if the work no longer happens in the ERP, is it still the record?

Which system stays the financial system of record?

Your ERP does, and that is the point of the design. The orchestration layer prepares and validates transactions before they post, so the general ledger receives structured, consistent data instead of entries that need manual correction. Finance also gains something most legacy P2P tools never offered: a live view of commitments from the moment a request is approved, rather than weeks later when the invoice posts.

Where purchase orders are issued is a configuration choice rather than a fixed rule. Opstream can create the PO and push it to the ERP, for example creating a PO in NetSuite and sending the contract to DocuSign from a single approval. Whichever option you choose, decide it per entity during setup and write it down, so audit knows which system owns which record. With the role of each system settled, the remaining question is practical: how do you get from here to there without disrupting the business?

How do you switch from a legacy P2P system to AI, step by step?

The switch works best as a sequence in which each step makes the next one easier. Mapping comes first because it tells you what to build, connecting comes second because it removes the need to migrate, and retiring the old tools comes last, once the new flow has proven itself.

Step 1: Map where requests start and where they stall

List every way a purchase starts today, whether that is email, chat, a legacy intake portal, a ticket or a spreadsheet, and then mark where requests wait longest. The map does double duty: it shows you where the delays are, and it becomes your first set of intake forms and routing rules.

Step 2: Keep the ERP and CLM, and connect them

With the map in hand, connect the systems you already run instead of moving their data. ERP, CLM and eSignature tools (such as DocuSign, Ironclad and Adobe Sign) and compliance tools (such as OneTrust and AuditBoard) plug into one data layer, so vendor data is unified across ERPs, CLMs, HRIS and compliance tools. The integrations page lists what connects today.

Step 3: Put one intake front door in front of everything

Once the systems are connected, route all new requests through one entry point. Requesters see only the fields that matter for their category, and approvers receive complete, structured data, which removes much of the back-and-forth that slows legacy approvals. If intake is new to your team, this guide to intake management covers the basics.

Step 4: Move approvals and policy into rules

Next, rebuild your approval chains as rules. Low-risk, low-value requests can be auto-approved, high-value ones can be routed by spend tier and risk, and legal, security or GRC review can be triggered only when the criteria are met. Because Opstream’s configuration is no-code, the people who own the policy can change it themselves instead of waiting on an engineering ticket.

Step 5: Turn on PO and invoice automation in phases

Start with one entity or category. Turn on PO creation and confidence-based invoice matching there, check the postings in the ERP, and expand only when the results look right. Starting small keeps the risk contained and gives finance evidence before the rollout widens.

Step 6: Retire the old front end, not the ledger

When requests, approvals and matching are running through the new flow, switch off the legacy intake portal and approval tool. The ERP keeps doing what it does well, and the second migration never happens.

That covers the mechanics. The more persuasive case is what changes for the people who use the system every day.

What does each team get from the switch?

Because the orchestration layer touches every function that buys, reviews or pays, the gains land differently for each team.

Finance

Budget checks happen at intake, before money is committed. Commitments, partial billing and remaining exposure are tracked continuously, so period close and budget reviews start from current numbers, and multi-entity and multi-currency structures are supported.

Contracts are linked to the request, vendor and approval they belong to. AI Document Comparison puts a vendor draft next to your standard terms and surfaces clause-level differences, and legal sees only the requests that meet the review criteria, so its time goes to the agreements that need judgment.

IT and security

Vendor security reviews trigger when a vendor will touch sensitive data, and SOC 2 reports and certificates are tracked by expiry. Opstream itself is SOC 2 Type 2 audited and supports SSO and SCIM provisioning.

Operations and procurement

Everyone uses one front door, and status is visible without chasing anyone for an update. Opstream reports 73% faster request handling and an 81% cut in shadow procurement. All of this leads to the question every project sponsor asks next, which is how long it takes.

How long does a switch to AI procure-to-pay take?

Opstream is typically live within 30-45 days, because the ERP and CLM stay in place and the work is configuration rather than migration. After go-live, plan a phased rollout with one category or entity first and the rest to follow. The bigger dependency is usually data rather than software. Gartner analysts reported in 2025 that 74% of procurement leaders say their data is not AI-ready, so it is worth cleaning up vendor and category data early; why clean data is the real AI readiness test goes deeper on that work. Timing and data readiness also shape how you should evaluate vendors.

What should you check before you sign?

Before you commit to any platform, test it against the questions that separate an orchestration layer from another migration. Ask each vendor to answer them in writing:

Before-you-sign checklist
  • Does the platform keep your ERP as the system of record, or does it require a re-implementation?
  • Are connectors bi-directional, and who monitors them when something breaks?
  • Can business teams change approval rules without engineering?
  • Can legal, finance and IT work in the same flow as procurement?
  • What has to be live in the first 30 days, and what can come later?
Key Takeaways
• Keep the ERP as the financial system of record and add an orchestration layer above it.
• Move intake, approvals, PO creation and invoice matching into one flow, with bi-directional sync to the ERP.
• Roll out by entity or category, and retire the legacy front end only after the new flow is proven.
• Start data cleanup early, because data readiness is usually the bigger dependency.

For quick reference, here are short answers to the core questions about making this switch.

Frequently asked questions

How do I switch from a legacy procure-to-pay system to AI?

Keep your ERP as the financial system of record and add an AI orchestration layer above it. Move intake, approvals, PO creation and invoice matching into that layer, connect your ERP and CLM with bi-directional sync, and roll out one category or entity at a time.

Do we have to replace our ERP to use AI in procure-to-pay?

No. An orchestration layer works with the ERP you have. Opstream connects natively to SAP, Coupa, NetSuite, Workday, Microsoft Dynamics and JD Edwards and syncs validated data into them.

Where are purchase orders issued after the switch?

You choose during setup. Opstream can create the PO from an approved request and push it to the ERP, while the ERP remains the financial system of record.

How long does a switch to AI procure-to-pay take?

Opstream is typically live within 30-45 days. Most teams then expand in phases by category or entity.

What is an agentic AI platform to automate procure to pay?

It is a platform where AI acts on live data and rules, not just suggests. It routes approvals, matches invoices by confidence, triggers reviews and flags exceptions across the procure-to-pay flow, with every action logged.

A different kind of month-end

For the controller in the opening, the change is not a new ledger. It is a month-end that starts from numbers everyone already trusts, because every purchase entered through one front door and reached the ERP clean.

Opstream runs intake, approvals, POs and invoice matching on top of the ERP you already have. See how it works at opstream.ai

Book a Meeting

Or tell us more:

About the Author
Mor Cohen-Tal, Co-Founder and CTO of Opstream
Mor Cohen-Tal
Co-Founder and CTO, Opstream

Mor Cohen-Tal is a visionary technology leader and the Co-Founder and Chief Technology Officer of Opstream, an intelligent procurement orchestration platform that is transforming the way companies buy.

With a career marked by a relentless pursuit of innovation, Mor has earned 13 patents for her groundbreaking work. Notably, Mor was the Cloud CTO at Turbonomic, where she spearheaded the company’s successful transition from a datacenter-focused business to a cloud-centric model. Turbonomic was acquired by IBM for nearly $2B in 2021.

As a leading thought leader in cloud and AI, Mor plays a critical role in cultivating partnerships with leading cloud providers such as AWS and Microsoft Azure, and has presented and keynoted at conferences around the world, including Microsoft Ignite and AWS re:Invent.

Mor holds an M.Eng from Cornell University and a B.Sc from the Hebrew University.

Connect with Mor on LinkedIn →

Sources

  1. Gartner, “What IT Leaders Must Do to Avoid Disappointing ERP Initiatives” (Denis Torii), published 10 May 2024. https://www.gartner.com/en/information-technology/insights/what-it-leaders-must-do-to-avoid-disappointing-erp-initiatives (70% by 2027; 87% figure from the 2023 Gartner ERP, Procurement, HCM and Finance Apps Survey, in the full report PDF https://emt.gartnerweb.com/ngw/globalassets/en/enterprise-architecture/ea-reskin-for-erp-research.pdf)
  2. Deloitte, “Procurement at the Tipping Point: Deloitte’s 2025 Chief Procurement Officer Survey” (press release), 19 Aug 2025. https://www.deloitte.com/us/en/about/press-room/2025-chief-procurement-officer-survey.html (siloed ways of working 57%)
  3. Procurement Magazine, “Gartner Explores How CPOs Can Lead Through 2025” (Sophie Rice), 26 Mar 2025. https://procurementmag.com/news/gartner-cpos-lead-through-2025 (Gartner: 74% of procurement leaders say their data isn’t AI-ready)

Want to see how it works?

Book a demo with our team or reach out at support@opstream.ai