Expensify Wants Its AI to Eliminate Work, Not Create a New Interface to Learn

Expensify, a cloud-based software platform for managing expenses, announced expanded capabilities for Concierge AI on July 1, 2026, allowing customers to analyze spending, automate expense reports, configure workflows, and manage expenses using natural language across email, text message, and the Expensify app.

"AI should eliminate work, not create another interface to learn," said David Barrett, founder and CEO of Expensify. "With Concierge AI, customers can simply ask a question or make a request in plain English, whether they're replying to an email, sending a text, or using the app. Concierge can understand the request, complete the work, explain what it did, and keep improving over time."

The most operationally significant of the five new capability areas is conversational analytics, according to the press release. Users can ask Concierge for a quarterly flux analysis by category, merchant, and employee, or a simple monthly spend summary, and receive a visual chart as the response. 

The system covers any dimension of the data: submitter, category, tag, workspace, card, reporting period, or receipt status. If the chart does not answer the question precisely, a user can describe what is missing and receive a revised version without touching a single setting.

Barrett published a five-level AI capability framework alongside the launch, an adaptation of the self-driving car capability scale applied to expense management.

Level 0 is where Barrett says most of the expense management industry currently sits: rules-based automation and alerts that require manual configuration and manual debugging when they break. 

Level 1 introduces read-only AI insight. Level 2, where Concierge currently sits, is on-demand AI action, where users can ask the AI to perform any task they could do through the UI and it executes immediately, according to the press release.

The more significant claim is about Level 3. Barrett describes Level 3 as workflow autonomy, where AI operates independently within a predefined workflow, using large language model reasoning rather than simple if-then rules to evaluate unstructured documents and conversations, make judgment calls, and collaborate with other humans or agents without waiting for confirmation. 

Expensify's agent rules feature, now in beta, is what the company is positioning at this level, and Barrett describes it as the first in the expense management industry to operate there.

Level 4 goes further, a persistent virtual agent that operates as an extension of a human role, such as a virtual AP coordinator or expense compliance manager, deciding what work needs doing and handling routine exceptions within defined permissions. 

Expensify's custom agents support this today. Level 5, which Barrett describes as aspirational and not yet available in Expensify, is outcome autonomy, where a user gives the AI a high-level goal such as "reduce close time by 30%" and the AI independently determines what work needs to happen to get there.

Autonomous workflow agents can inspect, route, and process expenses, chase receipts through conversational email, apply exceptions, and escalate only the cases that genuinely require human judgment.