The finance stack's real bill is integration, not licences

PYMNTS argues that finance technology has moved from a build-or-buy decision to an own, orchestrate or rent decision, because APIs and broader platforms now sell once-specialized capabilities as consumable services ^[1]. The consequence is budgetary rather than philosophical: the costly part of the modern stack is increasingly the integration between systems rather than the software, which forces CFOs to separate genuinely strategic systems from expensive but undifferentiated infrastructure ^[2].

The shape of the problem is familiar to anyone who has inherited a large back office. An ERP sits alongside treasury management software, accounts-payable and receivables platforms, fraud tools, bank-connectivity layers, payment orchestration, FX systems, reconciliation software and data infrastructure, with AI now being introduced across much of that architecture ^[5]. The bigger the company, the more sprawling the back office tends to be ^[3]. According to PYMNTS, the problem is no longer an absence of technology but its cumulative complexity ^[4].

The publication's illustration is an ordinary supplier payment: an accounts-payable platform approves the invoice, the ERP records the obligation, a treasury system folds the expected payment into its liquidity forecast, a payment platform sends the instruction to a bank, the bank executes, and reconciliation software confirms the transaction against the ledger ^[6]. That is six participants and five handoffs for one payment ^[7]. Each system may work properly; the company still maintains the interfaces between them, and those costs are easy to miss because they are spread across technology budgets, finance operations and control functions ^[8]. A cost with no single owner is a cost nobody defends in a budget review.

What has changed on the supply side is the unit of purchase. Cloud first weakened the assumption that buying software meant operating it, and APIs went further by letting companies consume specific capabilities without acquiring the complete systems that used to deliver them ^[9]. Payment network access without direct technical connections to every bank, account verification through third parties, FX pricing and execution embedded in treasury workflows, fraud detection bought transaction by transaction, bank data aggregated through connectivity providers: all are now consumption items ^[10]. PYMNTS notes that digital transformation and embedded services are table stakes, so the more consequential question is which parts of the infrastructure a company needs to own at all ^[11]. Its answer is that differentiation lives in the rules, not the pipes: when cash moves, where liquidity sits, which suppliers get early payment, when FX exposure is hedged, which transactions need a human, and how much risk is acceptable ^[12].

The precedent PYMNTS offers is a July development in Georgia's banking system, a shared Nasdaq Calypso deployment involving five of the country's largest commercial banks ^[14]. The argument is that if banks can share capital-markets infrastructure while retaining control, corporations may eventually share payments, KYC, fraud and treasury technology ^[13]. The caveats are stated plainly in the same piece: this is not wholesale outsourcing, concentrating payment connectivity, identity verification or treasury operations in fewer providers creates vendor dependency and operational concentration, and shared infrastructure demands data segregation, governance, portability and resilience ^[15].

Watch the framing shift in your own planning cycle. The last decade rewarded replacing manual processes with specialized software; PYMNTS expects the next to reward working out how few systems are needed for the same financial control ^[16]. The test of whether that has landed is whether integration and interface maintenance appear as a named line item rather than a rounding error distributed across three departments ^[8].