Docsumo

Docsumo is an intelligent document processing (IDP) platform that converts unstructured and semi-structured business documents into structured data for downstream systems. The product is organised around AI agents that handle successive stages of a document process: an email/collection agent that pulls attachments from shared inboxes, portals and APIs and chases missing files; a classification agent that splits and labels mixed multi-page uploads; an extraction agent that captures key-value fields and table rows with per-field confidence scores; a verification agent that checks documents against each other and against customer rules; and an analysis agent that produces figures such as cash-flow, revenue trends and risk flags. Output is posted to customer systems via API, webhooks, Excel, Salesforce or MCP, with only exceptions routed to human reviewers.

The platform groups three capabilities: Document AI (LLM-based classification, splitting and extraction, a pre-trained AI Models Hub covering bank statements, pay stubs, tax forms, ACORD forms, invoices, bills of lading and IDs, plus custom model training), Case Management (case types and checklists, cross-document validation, a Copilot for case questions and summaries, and branded request forms for missing documents), and Document Workflows (a canvas-based Agents Builder in which a workflow can be described in plain English and then tested and published). Data Tables present each processed document as a row and each extracted field as a column, with search, filtering and export. Docsumo serves industries including insurance, banking, lending, logistics and healthcare, and states SOC 2 Type 2 audited controls, ISO 27001 certification, and HIPAA and GDPR compliance, with customer documents not used to train shared or third-party models.

Founding story

Docsumo's founders, Rushabh Sheth and Bikram Dahal, encountered the document data capture problem while working on APIs for a consulting project — described variously as work in 2018 on document data capture APIs and as a project for a Canadian bank — and after seeing a finance head at a previous employer spend days reconciling invoices with bank statements. Bikram had earlier worked as a data entry clerk, and the pair had backgrounds in data science and business intelligence at Rocket Internet in Asia. Concluding that the problem was widespread and poorly served, they built Docsumo initially to automate extraction of data from invoices for accounting, starting the company in January 2019 with data capture APIs.

Business model

Docsumo sells software-as-a-service through a free 14-day trial (up to 1,000 pages, 10 user licences, pre-trained models, field and table extraction, integrations and API/webhook access) and two quoted commercial tiers. The Business plan adds auto-classification and splitting, AI-transform validations, custom document pipelines, permissions, a test environment, audit logging, custom integrations and a dedicated account manager and solutions engineer. The Enterprise plan adds automated workflows, AI-led case management, cross-document validations, real-time document analytics, master data lookup and priority support.

Subscription pricing on Business and Enterprise plans is quoted on the volume of pages processed over a month or year, plus the workflow steps required (classification, master data lookups, cross-document checks, automated workflows) and the payment term; annual plans carry a lower rate with credits issued up front, multi-year deals receive further discounts, and setup fees are charged separately depending on document complexity and support needs.

Traction

Reported milestones include PayU as the first enterprise customer in 2019, ten enterprise customers by October 2020, a 6x revenue increase over the 12 months preceding March 2022, a team of 30+ by April 2021, more than 5 million documents processed annually, and 10,000+ businesses on the platform. Customers span the United States, European Union and Asia.

Latest developments

Current platform positioning centres on a "Document AI Layer for Enterprise Workflows" built from seven AI agents spanning collection, classification, extraction, verification, case management and analysis, including a workflow agent that builds document workflows on a canvas from a plain-English prompt, Data Tables for reviewing captured fields at scale, and a Copilot for case review. Security posture includes SOC 2 Type 2, ISO 27001, HIPAA and GDPR.