How The Washington Post Uses AI Agents to Eliminate Tax Overpayments

The Washington Post operates at the intersection of journalism and technology, serving readers globally while managing complex back-office finance operations across thousands of vendor relationships. As the company navigated evolving media economics and invested in building its own in-house ML, Data, and AI Analytics capabilities, it identified finance operations as a key area where intelligent automation could have immediate, measurable impact.

Before deploying AI agents, the Finance team faced a structural bottleneck: every vendor submitted invoices in a different format, and validating the tax billed on each one was simply beyond what the team could handle manually. With invoice volumes far exceeding human review capacity, tax errors went undetected, and overpayments were quietly accumulating.

The Post implemented an AI Agent built on Automation Anywhere’s Agentic Process Automation platform, powered by the company’s proprietary LLMs. The agent reads, extracts, and analyzes tax data across every invoice, validates calculations, and issues corrections where discrepancies are found—automatically and at scale. The implementation ran on existing AWS cloud infrastructure already in place at The Post, enabling rapid deployment without rebuilding the underlying technology stack.

The impact was immediate and unambiguous: 100% of invoices are now reviewed for tax accuracy, tax overpayments have been eliminated, and the Finance team generated $1 million in automation value in the first year. The CTO, Vineet Khosla, publicly declared the outcome a proof point for moving fast with generative AI in enterprise finance.

Fueled by the Finance win, The Post has committed to an “AI everywhere” strategy, identifying additional use cases including donor reconciliation, procurement KYC, audit compliance, bank reconciliation, and CRM and ERP updates. The company is also packaging its proprietary LLM capabilities to create new revenue streams by syndicating its AI knowledge base to external platforms—turning internal AI infrastructure into a business in its own right.