Automating Loan Application Verification for a Regional SME Foundation

At the core of the proposal is an AI bot that automatically verifies data in loan applications using government and commercial registries - eliminating manual labor and ensuring consistent, auditable verification across systems.

Browser-based operation

The bot operates within a web browser, enabling it to search government websites and other registries directly - including locations that have no public API. This is what allows verification to be automated even where formal integration isn't available.

Integration with 15 systems and databases

The proposed solution integrates with 15 key systems, including the Land and Mortgage Registers (KW), REGON, the National Court Register (KRS), BIG InfoMonitor, and the rejestr.io API - covering the registries that actually drive loan-eligibility decisions in Poland.

Automated behavior on websites

The bot performs sequences of actions automatically: logging into systems on behalf of the user where required, navigating to the right sections, filling input fields, checking and unchecking boxes, selecting options, and clicking buttons like "Search."

Screenshot generation and cataloging

During verification, the bot captures screenshots of relevant subpages and sections - critical for assessment - and saves them in dedicated folders, automatically cataloged. The applicant's file thus contains an auditable trail of every check.

Operations on Excel files

The bot can also open, process, and capture screenshots of Excel files containing verification results - extending automation into the spreadsheet-based parts of the workflow.

Clear results in the user interface

Verification results are presented to Foundation staff in a clear UI: a summary of key data, visualization of verification status (e.g. whether all required checks passed), and one-click access to the generated screenshots.

Automatic data entry into the central system

Once verification is complete, the results are written automatically into the Foundation's central system - eliminating manual data entry by staff and the errors it produces.

Error handling and resilience

The system is designed for real-world conditions: long response times from external registries, query blocks imposed by source systems, and inconsistencies in data format. The bot monitors errors, logs them, attempts to resolve them automatically, and escalates critical issues to an administrator.

A redesigned application form

Alongside the bot, Softwise.AI proposed a redesigned application form addressing the UX issues identified in Discovery: a simplified structure with clear sections, a progress indicator, and automatic data filling - significantly improving the ease and completion rate of the application process.

{Results}

An AI bot concept covering 15 registries and a redesigned application form

Because this engagement was a Discovery, the outcomes are concrete, decision-ready artifacts the Foundation now owns:

• A full analysis of current processes and systems, including the specific operational bottlenecks in loan verification and the UX issues degrading application submissions.

• A designed AI bot concept capable of browser-based verification across 15 government and commercial registries (KW, REGON, KRS, BIG InfoMonitor, rejestr.io and others), with screenshot capture, Excel handling, automated data entry, and error handling.

• A redesigned application form with clear sections, progress indicator, and automatic data carry-over from registration.

• Three solution variations differing in scope, cost, and timeline - giving the Foundation board real choices, not just a single recommendation.

• A detailed budget and timeline for the recommended path.

• A stakeholder-approved roadmap ready to move into implementation.

{Business impact}

Hours of manual verification replaced by automated, cross-registry checks

• Process automation. Real-time data verification in external systems and instant report generation replace multi-hour manual workflows.

• Major time savings. Automatic verification across 15 registries means staff stop logging into the same set of registers, one application at a time.

• Error reduction. Automated file naming and streamlined document archiving directly attack the inconsistencies and small errors that accumulate in manual processing.

• Better experience for SME applicants. A simpler, clearer application form raises completion rates and lowers the support burden on Foundation staff.

• Audit-ready verification trails. Every check is documented with an automatically cataloged screenshot - useful both for internal review and for the controls a public-funding institution must withstand.

{Key recommendations}

5 principles for automating verification-heavy public-funding processes

A few principles emerged from this engagement that apply to any organization automating multi-registry verification workflows under public scrutiny:

1. Start with Discovery, not tools. A clear map of the current process - and a real conversation with the people doing the work - is what separates an effective automation from an expensive miss.

2. Use browser-based automation where APIs don't exist. Many critical Polish public registries have no API. A bot that drives the website is often the only way to automate verification without waiting on someone else's roadmap.

3. Capture screenshots, not just data. In a regulated lending context, auditability is a feature. Storing a timestamped image of each verification check is what makes the system usable in oversight.

4. Offer the board variants, not a single answer. Three costed options differing in scope and timeline give stakeholders a real decision to make, not just an estimate to approve.

5. Fix the form, not just the back office. Automation behind the scenes only works if the data coming in is clean. Application UX is part of the verification system, not a separate problem.