Which workflows will move from copilot to genuine labor and budget substitution first?

Evidence(8)

1. 

   Microsoft publishes randomized field-experiment findings on generative-AI work patterns

   [7]

   Apr 2025

   Microsoft Research published findings from a six-month, cross-industry randomized field experiment. On an intent-to-treat basis, email time fell by 1.4 hours per week; documents appeared moderately faster to complete, while meeting time did not change significantly. The result supports copilot augmentation and individually reclaimed time, not demonstrated payroll or budget reduction.

   1

   This is the report's clearest evidence that individual productivity gains do not automatically become labor-budget substitution. The experiment found behavior changes in independently adjustable work, while coordinated meeting time did not materially change and no staffing reduction was reported.

   See how this bears on the question
2. 

   Gartner surveys customer-service leaders on AI-driven staffing changes

   [3]

   Oct 2025

   Gartner's October 2025 survey of 321 customer-service and support leaders found that 20% reported reducing agent staffing because of AI, while 55% maintained staffing despite higher customer volumes and 42% were hiring specialized AI-related roles. These are respondent shares, not percentages of jobs eliminated.

   See how this bears on the question
3. 

   NBER coding-productivity study observes AI-tool effects through late 2025

   [6]

   Dec 2025

   The NBER matched event study analyzed coding-tool adoption windows that extended through December 2025 and covered more than 500,000 GitHub developers. The report uses its attenuation from commits to projects and releases to argue that maintenance, testing, documentation, migrations and other bounded, testable coding tasks may move earlier than broad developer replacement.

   See how this bears on the question
4. 

   UiPath closes its first quarter of fiscal 2027

   [11]

   30 Apr 2026

   UiPath's first quarter of fiscal 2027 ended April 30, 2026. The company reported $418 million of revenue and $1.901 billion of ARR. In the supplied report, this is evidence of commercial momentum in automation and agentic products, but not proof of customer labor substitution.

   See how this bears on the question
5. 

   Accenture closes its third quarter of fiscal 2026

   [12]

   31 May 2026

   Accenture's third quarter of fiscal 2026 ended May 31, 2026. The company reported approximately $19.3 billion in bookings and $18.7 billion in revenue. The report notes that these are not AI-only bookings: implementation gains can coexist with pressure on managed-services labor.

   See how this bears on the question
6. 

   ServiceNow publishes its 2026 Enterprise AI Maturity Index

   [1]

   9 Jun 2026

   ServiceNow published its 2026 Enterprise AI Maturity Index on June 9, 2026. The supplied report uses the index's enterprise-readiness findings to qualify the substitution thesis: permissions, reliable records, audit trails, integration and governance are prerequisites for turning agent activity into lower purchased labor.

   See how this bears on the question
7. 

   Salesforce reports second-quarter fiscal 2027 results

   [10]

   31 Jul 2026

   Salesforce's second quarter of fiscal 2027 ended July 31, 2026. The report cites the quarter's Agentforce/Data 360 ARR and Agentic Work Units as monetization and usage indicators. It cautions that bundled product definitions and usage volumes are not evidence that customers removed jobs or reduced labor budgets.

   See how this bears on the question
8. 

   GameStop reports AP automation results

   [14]

   Date unavailable

   Coupa's GameStop case describes a previously manual process involving 750,000 invoices entered into the ERP each year. After automation, GameStop reported a 20% reduction in AP headcount while processing 20% more invoice volume, a 70% reduction in average processing time and an 82% increase in first-time matching. The case says resources were redirected to strategic initiatives, so it supports functional displacement and redeployment rather than isolating a net enterprise payroll decline.

   See how this bears on the question