Leading global banks are increasingly positioning artificial intelligence (AI) as a foundational capability in corporate liquidity and cash management, driven by corporate clients’ need for continuous visibility and optimisation of cash across multiple accounts, currencies and geographies.

Historically, liquidity management has been constrained by manual reconciliation across fragmented banking relationships, time-zone mismatches and reliance on outdated or static forecasts. Cross-border payment and foreign exchange (FX) complexity further compounds these challenges, often resulting in idle cash buffers, suboptimal investment decisionsand unexpected liquidity shortfalls.

AI is now being positioned as a structural solution to these inefficiencies by enabling real-time aggregation of cash positions across accounts, improving predictive accuracy of liquidity needsand automating payments and fund sweeps. AI shifts treasury operations from periodic, reactive processes toward continuous, data-driven liquidity optimisation, eventually supporting 24/7 corporate treasury.

However, AI maturity across key areas in liquidity management within select global and regional banks reveals that AI adoption is converging rapidly in forecasting and cash visibility areas but remains fragmented in execution-heavy areas such as funding optimisation and liquidity orchestration, that require transaction processing, balance movement, and system-level action rather than analytics or forecasting.

Rather than simply indicating uneven adoption across use cases, the findings point to a deeper structural split in how AI is being deployed in liquidity management. The ecosystem is increasingly bifurcating into two layers: high-maturity predictive intelligence, concentrated in forecasting and visibility use cases, and lower-to-mid maturity operational intelligence, centred on execution, funding decisions and end-to-end liquidity coordination.

This split is fundamental. AI is already transforming how liquidity is understood through forecasting and analytics across treasury platforms, but not yet how it is executed within payment, settlement, and liquidity management systems connected to underlying market infrastructure rails.

From predictive intelligence to execution bottlenecks

The heatmap of 11 global and regional banks highlights that AI maturity is significantly higher in predictive liquidity capabilities than in operational execution layers. The AI maturity table (Figure 1) shows that banks are converging in predictive cash intelligence, but execution-layer AI such as funding, sweeping and allocation across banking and payments infrastructure remains fragmented.

For example, real-time cash forecasting and liquidity visibility tools are the most advanced, with leading institutions such as JPMorgan, DBS Bankand Bank of America deploying enterprise-scale models that integrate ERP and banking data to deliver near-real-time liquidity views. These systems increasingly reduce manual forecasting effort and improve short-term cash visibility across currencies and accounts.

AI-driven forecasting systems can significantly improve accuracy by ingesting structured and unstructured data streams, continuously refining predictions as new transaction data becomes available.

However, beneath this predictive layer of forecasting and cash visibility, a different reality emerges in the execution of liquidity decisions across banking systems.

Forecasting is reaching structural maturity across leading banks

Cash forecasting is the most mature AI application in liquidity management. Banks are increasingly moving from static, spreadsheet-based models to dynamic machine learning systems that continuously update liquidity positions using real-time transaction data.

Industry examples show that AI-driven forecasting can extend visibility horizons from 30 to 90 days, reduce manual reconciliation effort by up to 50% and significantly improve forecast accuracy. JPMorgan’s internal tools, for example its Cash Flow Intelligence, have reduced manual forecasting workloads by nearly 90%, illustrating the scale of operational transformation already underway.

However, even as banks deploy machine learning models to improve forecasting accuracy, the underlying operating model remains partially manual and reactive, constrained by legacy payment rails, fragmented banking connectivity, and treasury execution systems that require human approval for liquidity movements and settlement actions.

Execution remains the key constraint in liquidity orchestration

Despite strong progress in forecasting, liquidity execution remains fragmented. Treasury teams still operate across multiple banking portals, ERP systems and manual workflows for payment preparation, validation and reconciliation.

Most AI systems today are still limited to analytical intelligence including summarising data, classifying transactions and generating insights rather than operational intelligence, which would autonomously execute liquidity actions such as fund sweeps, payment routing, or balance optimisation.

This pattern is consistent across all six AI liquidity management use cases in the heatmap, including real-time cash forecasting, liquidity orchestration, virtual treasury assistance, FX liquidity optimisation, tokenised cash management, and funding allocation, each of which shows higher maturity in predictive and visibility functions relative to execution-layer automation.

Funding and tokenised cash management remain structurally immature

The lowest maturity scores in the heatmap are concentrated in funding allocation and tokenised cash management. These areas require AI systems to interact directly with balance sheet structures, regulatory constraints and cross-bank liquidity frameworks. These conditions that are not yet fully digitised or API-accessible.

Even as banks invest in programmable money and tokenised deposits, these systems remain in early-stage deployment, reflecting both technological and regulatory complexity.

Towards a 24/7 liquidity intelligence for corporate treasury

For corporate treasuries, the immediate impact of AI is improved liquidity visibility and forecasting precision. Treasurers are gaining near-real-time insight into cash positions across accounts, currencies and geographies, enabling more informed short-term funding decisions and reduced reliance on idle cash buffers.

However, true 24/7 liquidity orchestration where AI systems autonomously manage funding, optimise balances and execute liquidity movements across banking networks remains an emerging capability rather than a fully realised operating model.

The result is a hybrid treasury environment: predictive intelligence is increasingly real-time and accurate, but execution remains largely human-controlled.

TABInsights AI liquidity management heatmap analysis points to a market in transition

Autonomous liquidity orchestration 24/7, where AI actively reallocates cash, optimises funding and manages FX exposure in real time remains unevenly developed.

While banks are converging rapidly on predictive cash intelligence, particularly in forecasting and visibility, the transition towards execution-layer AI capabilities, including funding automation, liquidity orchestration and tokenised cash management remains in its infancy. Sophisticated and mature visibility layers co-exist with still emerging autonomous liquidity systems.

For corporates, this means that while 24/7 liquidity optimisation is becoming technically feasible, it is not yet operationally embedded at scale across the banking ecosystem.