AI-native firms transform workflows into operational capabilities

Enterprise AI moves from assistance to execution

OpenAI’s Enterprise Signals report shows that AI use in large firms is shifting from supportive tools to direct execution of work.

The data indicate that “frontier” firms—those in the top decile of AI usage—produce 8.3 times more output tokens per active user than average companies, up from 2.6 times in January.

This widening gap signals that leading organizations are linking AI agents to internal data and tools, delegating more substantive tasks, and making successful workflows easier to replicate.

The report advises leaders to convert this depth of usage into work that can be trusted, measured, and refined, while still allowing space for experimental use cases.

Case studies: Basis, Clay, and Exa Labs

Basis, a startup that builds AI agents for accounting firms, has embedded an agent into its employee onboarding process.

The company reports that first‑day onboarding now lasts 30 minutes instead of two hours, freeing HR staff to focus on culture and support.

New hires receive instant access to Codex and a company‑specific onboarding skill—a reusable set of instructions and resources for a defined workflow.

Codex greets the employee, explains core company concepts, and configures integration tools on the user’s computer in the background.

When recurring questions or exceptions arise, HR can update the skill before the next cohort arrives.

Basis demonstrated the process once, then codified it as a reusable skill with a clear trigger, defined steps, appropriate tool access, and an explicit “done” condition.

The onboarding flow no longer depends on a single person’s availability, though the team can still intervene for complex issues.

“Onboarding is now more consistent, repeatable, and easier to improve, and new employees gain an immediate model for working with AI,” the company says.

Clay builds a self‑learning revenue engine for go‑to‑market teams and faced the classic sales problem of scattered deal context across multiple platforms.

To centralize that information, Clay created a persistent workspace and assigned a dedicated sub‑agent to each account.

Each sub‑agent reviews primary sources such as CRM records, email, Slack, calls, and presentations, then updates its account folder overnight.

In the morning, a coordinating agent aggregates the nightly updates across all accounts and produces a short list of priority actions.

The recommended actions might include answering a lingering customer question, filling a gap in the buying committee, or providing a prospect with a new engagement hook.

Clay reports that the workflow saves roughly an hour of inbox triage each night for the engineer who piloted it.

Because the supporting evidence stays attached to each recommendation, sellers can inspect the original sources before acting.

The model can be extended to account executives, business development reps, solutions engineers, and sales leaders, respecting existing account permissions.

Clay’s approach illustrates that scaling evolving work requires a consistent structure, a reliable refresh cadence, shared evidence, and human judgment at the point of action.

Exa Labs applies AI agents to streamline developer ecosystem growth, integrating them into account management and related workflows.

By embedding agents, Exa aims to make developer onboarding and integration tasks more repeatable and measurable, mirroring the patterns seen at Basis and Clay.

Implications for enterprise leaders

The three examples demonstrate a common progression: teach an agent a stable process, give it persistent context as work evolves, then let it pursue opportunities autonomously.

For leaders, the challenge is to replicate this progression while ensuring that AI‑driven actions remain auditable and aligned with business goals.

Measuring output tokens per active user, as OpenAI does, provides a quantitative benchmark for tracking how deeply AI is embedded in daily work.

Companies that can consistently attach agents to core business context may achieve faster cycle times, reduced manual effort, and more predictable outcomes.

At the same time, organizations must retain human oversight for exceptions, complex queries, and decisions that require nuanced judgment.

Maintaining a “room for experimentation” allows firms to discover use cases whose value may not be evident on the first attempt, as the report advises.

By treating AI agents as operating capabilities rather than optional tools, enterprises can turn ad‑hoc assistance into repeatable, measurable processes.

Why This Matters: Companies that embed AI agents into core workflows can cut onboarding time, streamline sales coordination, and accelerate developer integration, creating measurable efficiency gains.