Nanonets is an AI-powered platform that automates data extraction from financial documents to reduce manual work.
We built a reconciliation solution combining document extraction and AI to streamline tedious financial processes.
Design an AI-driven reconciliation experience to simplify workflows, accelerate growth, and challenge dominant incumbents like QuickBooks, Xero, Oracle, and BlackLine.
User research, usability testing, user flows, wireframes, prototyping, and impact tracking.
12 weeks
As reconciliation was a new domain for the team, we lacked a deep understanding of the specific requirements and challenges associated with the reconciliation process. To bridge this knowledge gap and gain valuable insights into the process, its pain points, and market gaps, we initiated a series of expert interviews. These interviews helped us understand the complexities of existing solutions, and identify areas where there was a clear opportunity to improve the user experience and streamline the process.
Automate Transaction Matching
Automatically match bank and ERP transactions from CSVs to remove manual reconciliation.
Extract Data from Physical Documents
Capture and structure data from paper records for direct comparison with digital files
Simplify Card Reconciliation
Manage variable fees and charge structures with accurate, automated workflows.
Handle Complex Reconciliation
Support many-to-one and rule-based matching beyond typical ERP limitations.
The AI-driven reconciliation market was still emerging, with few established competitors and limited publicly available information. Most traditional tools were only accessible via sales demos, making direct feature comparisons difficult, so we relied on customer insights to understand key pain points.
To address limited competitor insights, we adopted an iterative approach: rapidly prototyping and gathering feedback from early users. This helped us uncover key needs and refine the feature despite the lack of extensive market data.
To validate interview insights, we launched an MVP focused on speed and rapid feedback. It was integrated into the existing workflows product to test a targeted AI model in real scenarios.
Mapping the user flow established the foundation for the reconciliation feature, helping define the journey, structure the information architecture, anticipate edge cases, and align stakeholders early.
After launch, we monitored usage to identify improvement areas. The feature automated 85% of reconciliation, reducing manual effort and helping acquire three new customers, contributing to ARR growth. Here were the key opportunities identified:
Lack of Visualization
Users wanted clearer visual mapping between invoices and bank statements
Manual Verification
Despite automation, users needed the ability to review and correct mismatches.
Transaction Marking
Users requested an option to tag transactions as “reconciliation not required.”
Based on testing insights, we enhanced the feature with a matching view so users could approve, correct, or nullify reconciliations. We also introduced a dedicated sidebar entry to improve visibility and quick access.
Because this feature diverged from existing layouts, we explored multiple wireframe and IA options, iterating quickly to refine the structure based on feedback.
By acting on user feedback and improving visibility, visualization, and control, adoption of the reconciliation solution increased. Over 14 customers adopted the feature, contributing to a 12% ARR increase. We automated 95%+ of the process and reduced financial close time by 70%.
User empathy guided this project, especially without the resources for a full competitive audit. I learned to rely on iterative design, using session recordings and usability tests to improve continuously.
Setting clear metrics helped me measure impact and spot opportunities, while aligning product and business goals kept the work focused. I also realized the value of building a deep understanding of the domain to design more effective solutions.
I'd love to connect with you!