Financial Services

As technology spend increased, it was unclear how much cost each application created, which systems were needed and how technical debt affected investment. Decisions rested on the views of different teams rather than shared data.

The application portfolio, total cost of ownership and technical debt were assessed together. A technology cost management solution made application, infrastructure, licence and support costs visible. Systems to retire and contracts to renegotiate were identified. In the first year the application portfolio was simplified by 15%; cost improvement initiatives delivered EUR 2.5 million in savings. Leadership began setting investment priorities on a more concrete financial basis.

Software entitlements were tracked manually, and as subscription models changed the link between real need and purchasing weakened. The organization bought new subscriptions without considering the unused licences it already held.

The advisory work matched contracts, usage data and existing licence entitlements. Through a software asset management and licence optimization solution, unused entitlements were reallocated and purchases were tied to verified need. A 10% reduction was identified across the five largest enterprise software costs. Internal reviews moved from annual to quarterly at no additional resource cost, reducing spend while increasing the frequency of control.

Using AI in critical functions required clarity across the organization on model ownership, approval, evaluation and monitoring. Different applications needed to be managed under common control.

An organization-specific AI lifecycle and continuous monitoring framework was established. An AI governance solution brought together responsibilities, model evaluations, version control and performance tracking. Automated oversight helped reduce manual control effort and enabled faster response to drift. No numerical financial result was disclosed; the value lies in scaling AI initiatives in a more controlled way and seeing risks early. Some generative AI integrations were still in testing when the work was completed.

Costs from multiple cloud providers were managed on separate screens. The central technology team absorbed the cost of services used by other units; resource ownership, budget forecasting and optimization opportunities were not visible enough.

A shared FinOps operating model was established with a cloud cost management solution. Oversized resources were rightsized, unused resources were shut down and suitable systems were stopped outside working hours. A 20% reduction in cloud spend within six months was reported. Tagging coverage that supports cost allocation to applications rose from below 30% to 80%. Forecast variance fell to about 3%. Finance, technology and business units began making decisions from the same cost view.

Teams involved in the financial close followed different procedures. Unclear handovers led to rework, and large, long meetings were held to resolve issues.

Process modelling and workflow automation made tasks, dependencies and handovers between teams visible. Progress and issues were tracked through a shared process. The number of participants needed for period-close meetings was halved. Savings of around USD 240,000 per quarter were calculated from reduced meeting time and attendance. The value came not only from automation but from turning unclear coordination into an orderly workflow.

Account opening, credit operations and customer information updates were document- and form-heavy. Re-entering the same data, correcting errors and manual checks lengthened processing times.

Processes were reviewed and steps suitable for automation identified. Robotic process automation was applied to document checks, data transfer, credit analysis and account opening. Processing times in the reviewed processes were reported to fall by up to 80% and costs by up to 20%. Employees could spend time on more meaningful work instead of repetitive transactions. These rates are not averages that can be generalized to all of the bank's operations.

Reviewing claim documents, re-entering data and switching between legacy systems limited staff capacity. Growing demand created more operational load under the existing way of working.

Workflows were mapped, and robotic process automation, workflow management and AI-assisted document processing were applied together. Data extraction from complex documents and repetitive steps were automated. Daily capacity was reported to rise from 15 claim files to as many as 288, with the processing cycle four times faster. No monetary savings were disclosed; the operational value is that existing staff can handle higher volume with fewer errors.