Banks can connect to enterprise systems. The harder question is what to do once they’re connected.
Open banking isn’t new anymore. APIs connecting banks with external platforms are table stakes at this point. But as these integrations get more sophisticated, something keeps concerning: is connectivity alone actually enough?
Here’s the thing most people miss: the real value isn’t in the API call itself―it’s in the workflow that surrounds it.
The Data Banks Can’t See
Most bank-to-enterprise integrations today focus on accounts receivable (AR) and accounts payable (AP), the financial records closest to the transaction layer. A Software-as-a-Service (SaaS) platform calls an API, a payment transaction gets initiated, and the bank executes it. It works. But that’s where it stops.
Years of experience working inside enterprise resource planning (ERP) environments reveals things most integrations miss. Sitting upstream of every AP entry is a purchase order. Behind every AR is a sales order. These documents live inside the enterprise’s ERP system, and the bank never sees them.
A purchase order is a future accounts payable. A sales order is a future accounts receivable. If a bank could see those, it could anticipate the cash flows, not just react to them.
Think about what that means in practice. A bank with visibility into purchase order data could spot that a client has committed to a 250,000 euro procurement coming due in 60 days. And that, based on current balances and projected inflows, the client will be short on cash. That’s not a payment processing problem. It’s a lending opportunity. It’s a cash management conversation. It’s a risk signal. All of this is sitting in one data point that’s currently locked inside an ERP system.
From Connectors to Orchestration
This is where it gets real. The enterprise technology landscape is fragmenting fast: SaaS tools, cloud applications, AI agents, and APIs are multiplying everywhere. The hard problem is no longer building connectors between systems―it’s orchestrating the workflows that span them.
For example, consider a single business scenario that touches multiple systems: a sales negotiation lives in the customer relationship management (CRM) system; the resulting purchase order sits in the ERP; the invoice and payment flow through a banking API; and if the counterparty is foreign, an FX conversion runs through a treasury or FX platform. Every step lives in a different system. Each system may have its own API, but without workflow logic that treats the chain as a single process, the bank executing the final payment has zero context about why that payment exists, what triggered it, or what comes next.
Banks, and the fintech platforms that serve them, need to stop thinking so much about connectivity and start thinking about context. You don’t get context without workflows.
The Opportunity Ahead
This isn’t theoretical. Financial institutions are already looking for ERP-native connections, not just to process payments but to power lending decisions, cash flow forecasting, and proactive financial services. The deeper they want to go, the more they run straight into these workflow challenges.
The industry is at an inflection point, and the plumbing is actually getting good. The APIs, the connectors, and the unified data models are maturing fast. Fintech vendors have made it dramatically easier to integrate with enterprise accounting systems. Open banking frameworks in Europe and evolving API standards in the U.S. keep expanding what banks can access.
The banks and fintech players that figure out how to orchestrate cross-system workflows, linking CRM data with ERP forecasts with treasury positions with banking services are the ones that will turn all this API infrastructure into something that differentiates them.
The AI Agent Paradox
AI agents make this picture more interesting- and more complicated. Agentic AI is changing how integrations get built. Right now, rather than writing integration code by hand, companies can increasingly describe what they want in plain language and have an AI agent to generate the software on the fly.
The problem is that an agent can only act on what it knows to look for.
If nobody tells the agent that a purchase order ledger contains predictive cash flow data, it will hardly make that connection on its own. AI agents are great at execution, but they still need humans to define the workflows, map out which data sources matter, define how they connect, and include what actions to trigger. Agents learn from existing patterns. Inventing new workflows? That’s still a human job.
A purchase order sitting in an ERP system today is just a row in a database. But wire it into the right workflow, and suddenly it’s a lending signal, a cash management trigger, a reason to pick up the phone with a client. The API is what makes the connection possible. The workflow is what makes it worth something.
There’s a real paradox buried in the “service as software” shift. As agents get better at executing tasks, the human work of figuring out which tasks to build becomes more valuable, not less. The bottleneck isn’t implementation anymore―it’s imagination.
The next wave of value in banking won’t come from building more connectors. It’ll come from figuring out what to actually do with all those connections.
If you’re ready to move beyond connectivity and start orchestrating the workflows that drive value, send me a quick note to [email protected].
