Autonomous AI agents: Your next partner in corporate finance?
AI agents are moving from concept to desk-level utility. For corporate finance teams, that means automation across payments, receivables, cash positioning, and foreign exchange exposure-while humans step in only for approvals, exceptions, and policy-sensitive calls.
Sumitomo Mitsui Banking Corporation (SMBC) is launching a Singapore-based venture to build these agentic AI tools for internal use and client deployment. The focus: simplify heavy workflows, reduce manual touchpoints, and shorten decision cycles in the CFO office.
What SMBC is building
The new venture will sit in Singapore and develop agentic AI-systems that learn, reason, and act with defined guardrails. SMBC will be the first customer, then the tools roll out to clients, starting with CFO teams that manage cash, liquidity, and risk across complex entities.
It's part of a broader push by Sumitomo Mitsui Financial Group, which committed an initial 800 billion yen (about S$6.76 billion) to accelerate digital initiatives, with 50 billion yen earmarked for generative AI. Ahmed Jamil Mazhari, SMFG's groupwide AI transformation advisor, will hold a stake in the venture. The team plans to showcase accessibility and practical use cases at this year's Singapore FinTech Festival.
Why this matters to CFOs and treasurers
- Shorter cycle times: agents can prep, reconcile, and route transactions before human sign-off.
- Cleaner books: automated matching for payments and receivables, fewer aging surprises, clearer cash visibility.
- More accurate forecasts: continuous ingest of bank, ERP, and TMS data to refresh cash projections intraday.
- Tighter FX control: exposure tracking by entity and currency, with suggested hedges tied to policy.
- Audit-ready operations: event logs, versioned prompts, and approvals that meet internal control standards.
- Lower cost to serve: fewer manual touches and rework across routine tasks.
How agentic AI fits into your workflow
- Payables: ingest invoices, validate against POs, schedule payments within policy, escalate exceptions, and prep approval packages.
- Receivables: auto-match remittances and statements, chase missing references, and flag disputes with recommended next steps.
- Cash positioning: compile prior-day and intraday balances, project inflows/outflows, and suggest sweeps or short-term investments.
- FX exposure: track forecast accuracy by currency, recommend netting or hedge tickets aligned to limits, generate deal memos.
- Working capital analytics: monitor DSO/DPO/inventory turns, surface bottlenecks, and quantify impact on cash conversion cycles.
Guardrails you'll need
- Data permissions and segregation of duties mapped to roles and approval thresholds.
- Human-in-the-loop at risk-based checkpoints (value, counterparty, country, instrument).
- Model risk management: versioning, monitoring drift, and prompt/response logging.
- Scenario sandboxes before live deployment; clear rollback plans.
- Latency, uptime, and failover SLAs that won't stall payment runs or deal execution.
- Audit artifacts: immutable logs, evidence packs, and policy mapping.
- Change management and training for analysts, approvers, and control owners.
- KPIs with baselines: cycle time, touch rate, exception rate, forecast accuracy, and savings.
How to pilot in six weeks
- Week 1: Pick one process (e.g., invoice matching) and define a single metric to improve.
- Week 2: Map the workflow end-to-end, mark where human approvals must stay.
- Week 3: Secure a clean data slice; define access controls and logs.
- Week 4: Configure an agent with clear policies, thresholds, and escalation paths.
- Week 5: Run shadow mode against live data; compare to BAU results.
- Week 6: Move to limited production; measure impact and decide to scale or iterate.
What SMBC is signaling
Client conversations point to rising complexity across finance operations. SMBC's move suggests banks will deliver more than credit and advice-they'll embed AI agents alongside treasury and controller teams to keep processes moving, clean, and auditable.
The message is simple: experiment now, learn where agents actually remove friction, then expand with controls that regulators and auditors can live with.
See it live
SMBC plans to demo its approach at the Singapore FinTech Festival. For event details, check the official site at fintechfestival.sg. For policy context and AI guidance, visit the Monetary Authority of Singapore at mas.gov.sg.
Level up your team
If you're evaluating tools and training for finance teams adopting AI, browse a curated set of AI tools for finance here: AI tools for finance.
Your membership also unlocks:
