Collateral, now with AI at its heart

For more than a decade, ApPello’s Collateral Management System has run at the centre of collateral processes in leading European banks. Its strength has been the deep domain knowledge behind the business and proven best practice in modeling, valuation, eligibility, allocation and activation. 

Now the technology behind took a huge step. At our recent Collateral Management System Demo Day in Prague, we showed our clients from Erste Group, Česká spořitelna and BCR, the new generation of the solution: the same collateral concepts their teams know, rebuilt on ApPello’s AI-native, low-code Cloud Development Platform. As one of our presenters put it: same road, a new car. 

What we introduced 

Open around the clock: no more nightly lockout 

In the previous generation, the whole system went offline while the end-of-day batch ran. Now end of day runs per partner group, in parallel, in separate pods on Kubernetes, and only the partner group currently being processed is locked. Netbank, loan origination and other systems can send events and queries at any time, and a run resumes where it stopped after a restart. 

The concepts teams already know, each one redesigned 

• Partner groups: partners that share a common asset are modeled together. 

• One approved structure: every approved loan, limit, collateral and asset sits in one persisted layer, with snapshots at proposal and approval. 

• Central asset register: one record per asset for the whole bank, reused across deals and group companies and open to other systems. 

• Native credit lines: limits have their own approved, available and utilised amounts, and sub-loans inherit the collateral links automatically. 

• Insurance you can see: every policy carries a validity flag, and the effect of missing cover (ineligibility, haircut or both) is set by configuration. 

• One task inbox: a standard Camunda BPMN process engine with role-based task pools and four-eyes approval. 

Rules as visual models, not code 

Eligibility and loan risk value now run in the integrated Decision Engine: calculation trees built in a visual editor, tested on historical data before go-live, and versioned with who changed what and when. Business parameters are maintained once in a central business-administration module and propagated to every module. 

Configuration with version control 

Screens, data model, rules and processes remain configurable, but configuration now lives in Git. Every change is made in its own branch, tested, reviewed and merged, with one-click rollback. Processes are edited in the embedded Camunda BPMN editor, and document templates are maintained by the bank in Word or Jasper and changed in one place. 

AI at the heart of the platform 

The Hands-Off AI Assistant knows the platform’s interfaces, configuration and documentation, and works right where users work. In the live demos it: 

• created a complete collateral model (one loan, a cash deposit and a house) from a client’s PDF loan agreement; 

• produced a risk and collateralisation analysis as a downloadable PDF; 

• built an insurance-gap dashboard and remediation plans from a plain-language request, in its own reviewable branch.