Modelling changes before they go live
- Customer twin
- Credit pipeline twin
- Transaction flow twin
Solutions
DFT solutions are organised into 4 lines of work and 13 groups. Each group lists the tasks it covers. Data and pilot format for each of the 43 solutions are given in the catalogue.
Digital twins
A virtual model of the bank’s customers, processes, transactions and IT estate on which changes are tested before launch. DFT started with this line of work.
Release testing in a copy of production
- Environment twin
- Tester agents
- Load and stress testing
AI agents for business units
Agents gather data and prepare draft documents in AML, reporting, the back office and client work.
Automated investigations in AML and anti-fraud
- Alert investigation agent
- Linked-account network analysis
Analytics and reporting on DWH data
- “Ask your data” agent
- Regulatory and management reporting agent
- Credit analyst agent
Search, processing and review of bank documents
- Policy and NBU regulation assistant
- Incoming document processing agent
- Contract and internal policy review agent
Operational metrics, processes and service quality
- Department and branch KPI agent
- Process mining and workload planning
- Service quality agent
Task, time and budget accounting for IT projects
- Time tracking agent
- Project cost and budget agent
- Project status and risk agent
- Portfolio and resource twin
- Task triage and assignment agent
- Delivery analytics agent
Automated reconciliation, period close and budget control
- Procure-to-pay and invoice agent
- Period-close agent
- Budget control agent
- ERP assistant for employees
Meeting preparation and CRM data maintenance
- Relationship manager assistant
- CRM data quality agent
- Next best action agent
- Call and request summary agent
Control and audit
Model monitoring and a reproducible audit trail.
Monitoring, validation and audit of bank models
- Model registry and drift monitoring
- Challenger models and shadow scoring
- Audit trail for AI decisions
Platform inside the perimeter
Shared connectors and policies, the bank’s own GPUs and the applications developed on this foundation.
A shared platform for launching and controlling agents
- Agent orchestration platform
- Action approval loop and agent policies
- Agent evaluation and observability
Model deployment on the bank’s hardware
- Private Fabric: an AI platform inside the bank’s perimeter
- Private inference for open models
- Fine-tuning and federated learning
Custom PM and ERP application development
- PM system built around the bank’s processes
- ERP modules and workspaces
- AI features in existing applications
By role
Recommended solution groups for the bank’s departments
Recommended groups
If your task is not on the list, the use case is defined in a diagnostic session.
Discuss a pilotStarter packages
Solution combinations for a first pilot
- 01
Model validation and audit
Banks that already use scoring or anti-fraud models
- 02
Release testing
IT and digital channels ahead of a major update
- 03
Analytics for the board
Management board, finance and heads of business lines
- 04
Modelling product changes
Retail and lending ahead of rate or product changes
- 05
IT portfolio accounting
CIOs, PMOs and finance teams that collect hours, costs and project status by hand
- 06
CRM automation
Corporate banking, SME and the contact centre, where the CRM is filled in by hand
- 07
Accounting automation
Accounting and finance, where invoices and the month-end close are reconciled by hand
- 08
Change management system
CIOs and PMOs that need to track the path from request to committee decision
- 09
Private AI platform
Banks planning several AI use cases on one platform inside their own perimeter
Choosing the first use case
The use case and success metric are agreed with the bank in a diagnostic session.