Mal7
Insights
Banking architectureJune 29, 20266 min read

J.P. Morgan's digital twin of company cash flow retires the multi-year core rebuild

J.P. Morgan's liquidity twin shows how regulated institutions can simulate risk, surface mismatches and return measurable answers without waiting for a full core replacement.

By Mustafa Khider

Executive summary

  • J.P. Morgan's liquidity twin runs 4,000 overnight simulations against actual daily numbers and surfaces a 12% probability of covenant breach within 60 days
  • The important architectural point is that the twin works against existing systems, using live account balances and APIs rather than waiting for a rebuilt core
  • Treasury, AML and underwriting examples point to the same pattern: extract real-time data, simulate in a contained environment and return decisions to operations

So what

Bank CIOs and CFOs should treat digital twins as a budget-cycle alternative to multi-year core rebuilds when the immediate question is measurable decision evidence

The core signal is not speed for its own sake. It is a measured answer produced from the systems already running the bank.

4,000

overnight simulations

Run against the day's actual numbers in the J.P. Morgan example

12%

covenant-breach probability

Surfaced within a 60-day window against real balance-sheet data

14

hour reporting delay

The finance-system lag described across cash-flow records

USD 2.3M

foreign-exchange mismatch

Caught in the overnight workflow before quarter-close reconciliation

01 Signal

A 12% covenant risk surfaced from live cash data

On 25 June 2026, J.P. Morgan Payments published a strategic analysis introducing what the bank calls a digital twin of liquidity. The phrase means what it suggests. A digital twin is a live software replica of a real operation, in this case the money flowing through a corporation, that mirrors the real thing moment by moment, fed by actual balance-sheet data and direct feeds from the firm's banks. The treasurer runs what-if scenarios overnight against the digital twin without disturbing the live operation. In the example J.P. Morgan publishes, the twin runs 4,000 overnight simulations against the day's actual numbers and surfaces a 12% probability that within 60 days the company will breach its debt covenants, the financial promises attached to its loans that let the lender call in repayment if broken. That figure, specific, dated, anchored to real numbers rather than quarter-end forecasts, is the artefact worth reading.

02 Timing

Real-time payments have outrun overnight finance records

Why 25 June matters. Corporate treasury teams have spent the last three years watching the speed of real-time payment systems outrun the delay in the accounting records behind them. Cash moves in seconds; the records of where it sits update overnight. The mismatch is the operational reality of every chief financial officer trying to manage cash flow across 30 currencies and 40 banks while the finance system reports where the money sits with a 14-hour delay. The J.P. Morgan release lands inside the budget cycles where bank chief information officers and corporate chief financial officers are deciding whether the answer is a multi-year overhaul of the bank's central record-keeping platform, known in the industry as the core, or a thinner, faster digital twin overlay. The decision they make this quarter shapes capital allocation through 2028. The publication is timed to that decision.

Funding choice

Six months versus five years is now the budget choice

6 months

Digital twin overlay

A faster path running against existing infrastructure with a small senior team

5 years

Core rebuild

A multi-year overhaul of the central record-keeping platform

03 Rebuild case

The rebuild case still has weight

The prevailing view says the overlay cannot work without the rebuild. Real-time artificial intelligence in treasury, in fraud, in underwriting, requires data that is clean, current, and available in milliseconds, and that requires a modern core. The case has weight. The largest global banks run mainframe systems whose oldest components predate the readers of this article. Risk data sits in one system, cash-flow data in another, customer data in a third. Overnight processes reconcile them. Bolting an AI layer on top of those systems, the argument goes, produces fast nonsense. The institutions that have tried it have the scars. Core modernisation programmes at the largest European and North American banks have absorbed billions of euros and US dollars on the premise that the AI layer above can only work if the data layer below is rebuilt. The proponents of the rebuild include serious people inside the banks, inside the regulators, and inside the consulting firms that have spent two decades selling these programmes. Their argument is that any shortcut produces an answer that looks fast but is wrong, and a wrong answer routed into a treasury workflow at speed is more dangerous than a slow correct one. Bank supervisors have shared the worry, treating data integrity in the underlying systems as a precondition for any AI-driven decisions that touch the balance sheet.

04 Overlay proof

The twin works without a rebuilt core

The J.P. Morgan release punctures the consensus by demonstrating the opposite. The bank's twin works without a rebuilt core. It pulls live account balances through application programming interfaces, the technical connection points other software uses to read and write to a bank, and runs its simulations against the existing systems. The 12% covenant-breach probability is produced against the existing infrastructure while it continues to run. The overnight workflow J.P. Morgan highlights catches a USD 2.3 million foreign-exchange mismatch caused by a supplier silently changing its invoicing currency, a discrepancy that would not have surfaced in the overnight reconciliation until the quarter closed and the chief financial officer asked why the currency hedge looked wrong. The pattern reaches beyond treasury. Reform Technologies, a Belgian financial-crime regulatory technology firm that closed a EUR 1 million equity round and appointed former European Central Bank supervisor Elizabeth McCaul as chair on 25 June, builds the same digital twin construct for anti-money laundering (AML), this time mirroring the bank's compliance system rather than its cash flow. Compliance teams use the twin to simulate the effect of a rule change before deploying it into the live monitoring system. American International Group, a major US-headquartered commercial insurer, reported in its first-quarter results that General Insurance underwriting income more than tripled year on year to USD 774 million, attributing the move to underwriter-centric AI models scaled against the existing portfolio of policies rather than a rebuilt underwriting platform. Three different regulated workflows, three different overlays, the same architecture. Real-time data extraction from existing systems, simulation in a contained environment, decisions returned to operations without modification to live databases.

Operating pattern

The same architecture appears across three regulated workflows

01

Extract live data

Live account balances are pulled through application programming interfaces from existing systems

02

Simulate in the twin

The contained replica runs scenarios without disturbing the live treasury, compliance or underwriting workflow

03

Return decisions

Measured answers move back into operations without modifying the live databases underneath

Reference points

Treasury is not the only place the twin pattern shows up

J.P. Morgan Payments

Treasury

A liquidity twin runs simulations against real balance-sheet data and catches a USD 2.3 million FX mismatch

Reform Technologies

AML

A compliance twin lets teams simulate the effect of a rule change before deploying it into live monitoring

AIG

Underwriting

Underwriter-centric AI models scaled against the existing policy portfolio as underwriting income more than tripled year on year

05 Economics

The economics still favour the rebuild

This is where the category critique becomes load-bearing. The dominant consulting story of the last decade has been that AI in banking requires a core replacement first. That story has a structural reason for existing. The partner-light, junior-heavy delivery model the Big Four accountancies and global systems integrators run, where a small number of senior partners supervise a pyramid of junior consultants billed by the hour, produces its highest margins on the multi-year overhaul programme. The overlay model does not. A six-month digital twin deployment by a small team of senior practitioners cannot absorb 400 junior consultants, cannot run for five years, cannot sustain the partner economics. The advice from those firms therefore tilts, with structural consistency, towards the rebuild. The advice is sincere. The economics are aligned.

06 Budget choice

Production proof changes the funding choice

For the bank chief information officer in budget reviews this quarter, the funding choice has narrowed to two options: a six-month digital twin overlay or a five-year core rebuild. Each path now has visible reference points. The J.P. Morgan twin is one. Reform's twin is another. The chief financial officers reading AIG's first-quarter results have a third. Each represents a production system delivering measurable outcomes against existing infrastructure within a calendar quarter. The Financial Conduct Authority, the United Kingdom's conduct regulator for financial services firms, signalled in its 10 June Emerging Technology Horizon Scan that the same digital twin pattern is moving to the consumer side, where real-time data feeds will produce per-customer models that change how products are priced and sold. In the Gulf, the same architecture is finding a market. Omnix International, a Dubai-headquartered technology advisory, expanded its digital-twin advisory practice this month to address operating-model demand across regional asset portfolios, an offering aligned with the well-funded technology strategies of regional governments and sovereign investors.

07 Close

The next banking modernisation programme starts as a digital twin

The next banking modernisation programme starts as a digital twin running against the systems already in place, returning measurable answers in the same week the contract is signed. The firms still selling the five-year core rebuild will find the pipeline runs out before the pyramid does.

About the author

Mustafa Khider is a co-founder of Mal7, focused on enterprise AI, automation and production deployment for financial institutions, FinTechs and regulators.

Digital twin readiness

Use the systems you have to prove the decisions you need

Mal7 helps regulated institutions move AI into production with the data, controls and operating evidence needed to make measurable decisions against existing enterprise realities