Automate and Secure Your Client Document Onboarding

Many accounting firms rely on a combination of email, Google Drive, and a client portal like SmartVault or Canopy. These portals provide a secure place for clients to upload files, but they do not use AI to read the documents. An accountant still has to open a PDF bank statement, manually key transaction data into workpapers, and verify totals. The workflow logic lives in a separate spreadsheet, disconnected from the documents themselves.

A typical onboarding scenario involves an admin emailing a checklist of 7-10 required documents to a new client. The client sends back a mix of files, forgets two, and scans three others into a single 50-page PDF. The admin then spends 30 minutes downloading, renaming, and saving files to a specific folder structure, then updates their tracking spreadsheet. They follow up with the client about the missing W-9, starting a cycle of emails that can delay the start of actual accounting work by days.

Even practice management software with document features, like Karbon or TaxDome, treats files as attachments to tasks. They can tell you a file was uploaded, but not what is inside it. You can check off 'Received Bank Statements', but you cannot automatically confirm that all 12 months are present or that the ending balance on one statement matches the beginning balance on the next.

The structural problem is that these tools separate the document from its data. The file is a static object, and the information inside is invisible to the system. This forces your team to act as the bridge, performing low-value, error-prone data entry that creates bottlenecks and security risks from handling sensitive information manually.

The first step is a workflow audit. Syntora would map every document you collect during onboarding, identify the key data points in each, and define the rules for a 'complete' package. We would also review your security and compliance needs to establish the data handling requirements for sensitive information like Taxpayer Identification Numbers (TINs).

The technical approach uses a FastAPI backend to provide clients with a secure, dedicated upload portal. For security, files are streamed directly to an encrypted AWS S3 bucket, never resting on the application server. An AWS Lambda function, triggered on upload, uses the Claude API to perform data extraction. For example, the function would read the client name from an engagement letter, the TIN from a W-9, and transaction lines from a bank statement. All extracted data is stored in a structured Supabase (PostgreSQL) database where Pydantic models enforce data validation.

The delivered system is a central dashboard showing the onboarding status for every new client. The dashboard would flag missing documents or incomplete data and can be configured to send automated reminders. Instead of spending 30 minutes on manual data entry, your team would spend under 5 minutes verifying AI-extracted information, which can be formatted for direct import into your accounting software.