The $2.4M Sitting Idle in Your Treasury

Your treasury team holds buffers across 15 or more currency accounts. Each one is sized for the worst week, because the alternative, underfunding and delaying a settlement, is worse. The logic makes sense, but the cost adds up fast.

For a company processing $150M annually across multiple corridors, $2M to $3M or more sits idle at any given time. The buffers exist to absorb uncertainty in when money arrives, when it leaves, and how much each corridor needs on a given day. The root cause is the forecast itself: aggregate, monthly, backward-looking.

AI cash flow forecasting changes this by predicting inflows and outflows at the corridor and currency level with daily granularity. Instead of one blended number for your entire treasury, you get a probabilistic view of every corridor's cash needs, including when customers pay, when partners settle, and where volume is about to spike.

How corridor-level forecasting works

Cross-border cash flows are a portfolio of corridors, each with its own rhythm. EUR-USD has different volume patterns than GBP-NGN, Monday behaves differently from Friday, and month-end looks nothing like mid-month.

AI forecasting decomposes your cash position by corridor, currency, and settlement window. The model ingests not just historical transaction volumes but forward-looking signals: customer pipeline data that predicts upcoming large transfers, seasonal remittance patterns that shift corridor demand, macroeconomic indicators that affect FX volatility, and real-time payment status data that tracks where money is in the correspondent chain.

The result is a forecast that tells you how much capital each corridor needs, when, and with what confidence. This is the full cash flow picture: how much each corridor needs, when every dollar moves, and where your nostro pre-funding can tighten.

The system in five steps

1. Ingest demand signals by corridor. Transaction data, settlement schedules, customer pipeline information, seasonal patterns, FX volatility, and payment status feeds. Each corridor gets its own data view.
2. Train corridor-specific models. Each corridor gets its own forecast model that learns daily, weekly, and seasonal rhythms. A model trained on your USD-INR corridor would be useless for EUR-GBP.
3. Generate probabilistic forecasts. Confidence-weighted ranges by currency and time window, not point estimates. Treasury sees what is likely, what is possible, and what is unlikely for each corridor over the next 7 to 30 days.
4. Dynamically adjust liquidity positions. Raise or lower funding levels based on predicted demand. During a high-confidence quiet period, the system lowers the buffer. When a settlement spike approaches, it increases funding proactively.
5. Feed outcomes back into the model. Every actual settlement becomes training data. Forecast accuracy compounds over time as the model learns your corridors' real behavior.

Three mistakes that keep capital trapped

Forecasting at the aggregate level. A blended cash forecast across 15 currencies hides corridor-specific variance. You overfund quiet corridors and underfund volatile ones. The aggregate number looks fine while individual accounts are consistently wrong.

Using the wrong cadence. Monthly forecasts cannot capture day-of-week effects, month-end spikes, or partner settlement schedule changes. Cross-border settlement dynamics require daily or intraday granularity. PwC's 2025 Global Treasury Survey found that 38 to 52% of organizations still collect forecasting data manually, locking teams into slow update cycles that cannot keep pace with actual cash movement.¹

Ignoring leading indicators. Most teams forecast from historical volume alone. They miss forward-looking signals that predict cash needs before they arrive: enterprise customer pipeline data, seasonal remittance spikes, corridor-specific holidays, and macroeconomic shifts that redirect volume between corridors.

What the numbers look like

The results from companies that have made this shift are concrete. KPMG's 2025 report on AI in corporate treasury found that Bosch reduced its cash buffer by 30% after deploying predictive analytics for treasury.² ASML improved FX exposure forecasting accuracy from 70% to 96% using an internal AI approach.³ Strategic Treasurer's 2025 survey confirms the problem: 53% of treasury professionals now rate cash forecasting as "difficult," up from 39% in 2018, with 91% still relying on Excel.⁴

Here is what this looks like for a payments company:

You hold $15M across currency accounts, buffers calibrated for worst-case scenarios. Even a conservative 16% reduction through tighter forecasting frees $2.4M. At an 8% cost of capital, that is $192K per year in opportunity cost, capital that could fund corridor expansion, strengthen your next board update, or simply earn a return.

This connects directly to the broader working capital trapped across your payment stack. Cash flow forecasting is one of the highest-impact interventions because it attacks the uncertainty that forces every other buffer to stay oversized. The challenge is getting it into production.

Why most teams cannot build this internally

The concept is straightforward, but the execution breaks most internal teams.

Multi-signal feature engineering. Combining transaction history, pipeline data, seasonal patterns, macro indicators, and real-time payment status into a unified feature set per corridor requires data engineering most treasury teams do not have. Each signal has a different cadence, format, and reliability profile.

Probabilistic calibration. Generating point estimates is straightforward, but confidence-weighted ranges that treasury can trust for funding decisions require careful calibration and backtesting across each corridor. The difference matters: a model that says "$800K plus or minus $400K" gives you a number on a screen, while one that says "$800K with 90% confidence between $650K and $950K" gives you a funding decision.

Corridor-specific model management. Each corridor needs its own model. That means training, monitoring, and retraining 10 to 20 or more models in parallel, each with different data characteristics and drift patterns. When your GBP-NGN model degrades, you need to catch it before treasury makes a bad funding call.

Workflow integration. Forecasts in a dashboard do not change funding decisions. The system needs to feed directly into treasury workflows and settlement processes to actually reduce buffers.

What you can do in the next 60 days

Weeks 1 to 2: Audit balances versus actual utilization. Pull 90 days of daily balances and actual settlement volumes by corridor. Calculate the average idle percentage per account. This shows you exactly where capital is trapped and by how much.

Weeks 3 to 4: Map corridor-specific patterns. Take your top five corridors and chart day-of-week and month-end patterns. Identify which corridors have predictable demand and which are volatile. This tells you where forecasting will have the most impact.

Weeks 5 to 6: Build a simple rolling forecast. Start with your highest-volume corridor. Build a basic rolling forecast using the patterns you identified and measure accuracy against actual funding needs. Even a simple model reveals how much lift is possible.

Weeks 7 to 8: Quantify the idle capital cost. Compare your predicted funding needs to actual balances. The gap, multiplied by your cost of capital, is what you are paying for forecast uncertainty. That number builds the business case for AI forecasting.

How Devbrew builds this

At Devbrew, we build corridor-level cash flow forecasting systems for cross-border payments companies. Custom AI trained on your transaction data, your corridors, your settlement patterns.

The system includes multi-provider data connectors that consolidate balances from banking portals, SWIFT messages, and payment platforms into a single corridor-level view. Forecast models train per corridor and retrain automatically as new settlement data arrives. Funding recommendations feed directly into your existing treasury workflows, no rip-and-replace. The same forecast accuracy that frees idle capital also reduces FX hedging costs downstream.

Talk to us

If you are managing cash positions across multiple currencies and corridors, a 30-minute conversation can clarify where capital is sitting idle, what is at stake if it stays there, and whether AI forecasting fits your treasury.

Book a discovery call or reach out at joe@devbrew.ai.