Smartsheet launches AI planning app to flag budget overruns and project delays

Smartsheet introduced vsActual on November 4 at Engage 2026, giving executives a standalone AI application designed to spot when company plans slip off course. Separate from its core work-management software, the app connects to business systems to flag budget overruns, delays and staffing conflicts, then recommend changes. It is available in early access.

General availability is planned for early 2027. Smartsheet confirmed to Computerworld that billing will be based on usage, but did not disclose rates. The application is aimed at senior executives, particularly chief financial officers and chief operating officers overseeing budgets, capacity and delivery.

From strategy documents to corrective action

The workflow starts with material companies already use to describe their plans. An intake agent reads documents such as presentation decks and spreadsheets to help create a plan of record: a structured account of goals, deadlines, budgets and staffing limits. A planning agent supports capacity planning and models alternative scenarios when those plans need to change.

Rather than require a company to move all its work into Smartsheet, vsActual draws information from connected applications. These include Oracle, Salesforce, SAP, Workday and Smartsheet itself. It can also analyze Slack conversations for signs that execution is diverging from the plan, adding workplace discussion to the operational data it examines.

A drift agent continuously looks for trouble, including spending above budget, schedule slippage and resource conflicts. Its recommendations can include adding staff to a project facing a predicted delay. Other agents handle data quality and reporting, keeping users informed about changes across projects. Agents can respond to plain-language requests or run in the background.

Connecting a delayed task to a company goal

Behind those agents is Smartsheet’s Execution Graph, a knowledge graph—a map of relationships among tasks, projects, programs and strategic goals. Smartsheet says it helps the application trace how a disruption in one part of a business affects others, rather than treating each late task as an isolated problem.

The company says the graph draws on two decades of workflow data covering millions of projects and nine petabytes of information. Smartsheet is not training its own large language models. Instead, it selects among multiple foundation models according to the work being done and its cost, chief product and technology officer Pratima Arora told SiliconANGLE.

If there is a delay in a Jira ticket that has impact five levels up, we can detect it and tell you what is happening,

Pratima Arora, Smartsheet chief product and technology officer, speaking to SiliconANGLE

A new buyer—and a data-quality test

The agents are not limited to giving advice. They can update data in third-party applications when granted the relevant permissions, although human approval is required by default. That gives organizations a choice about whether the system only proposes a response or also carries it out through connected software.

Liz Miller, vice president and principal analyst at Constellation Research, told Computerworld that output accuracy will depend on the quality of customers’ operational and financial data. She urged organizations to test and ground agents before assigning autonomous tasks, and to establish processes that allow human judgment and intervention when needed.

There is also a sales challenge. Futurum Group research director Keith Kirkpatrick described vsActual as a shift from Smartsheet’s bottom-up adoption among project managers and operations teams toward executives and strategy offices. His assessment was that the concept is sound, but it asks Smartsheet to sell an unfamiliar kind of product to a new buyer.