OpenAI’s latest Enterprise Signals indicates enterprise AI is shifting rapidly from assistance toward execution, with substantial variation in speed between organizations. Frontier firms (the top 10% by AI usage) now produce 8.3× as many output tokens per active user as typical firms — up from 2.6× in January. That widening gap suggests leading companies are not just using agents as helpers but tying them to company context and tools, delegating more substantive work, and making successful workflows easier to repeat.
The leadership challenge is to convert that depth into work people trust, can measure, and can improve, while preserving room for experimentation that may uncover value not obvious on first try.
Three startups — Basis, Clay, and Exa Labs — provide concrete examples where agents are embedded in employee onboarding, account management, and developer ecosystem growth. Their workflows differ, but the progression is instructive: teach an agent a stable process, give it persistent context as the work evolves, and then let it carry opportunities into tested action.
Basis: faster, more consistent onboarding
Basis, which builds AI agents for accounting firms, has shortened first-day onboarding from about two hours to roughly 30 minutes, freeing HR to focus more on culture and support.
New hires receive immediate access to Codex and a company-specific onboarding "skill" (a reusable set of instructions and resources for a workflow). Codex greets them, explains key company concepts, and completes integration setup on the employee’s computer in the background. When recurring questions or exceptions arise, HR updates the skill for the next cohort.
Basis demonstrated the onboarding process once and then converted it into a reusable skill with a clear trigger, known steps, access to the right tools, and a definition of "done." The process no longer depends on a single person’s availability, yet the team can intervene for exceptions or complex issues. Onboarding became more consistent, repeatable, and easier to improve, and new employees gain an immediate model for working with AI.
Clay: persistent account agents to keep deal context current
Clay builds a self-learning revenue engine for go-to-market teams and addresses a common sales problem: critical deal context is scattered across CRM records, email, Slack, calls, presentations, text messages, and internal conversations.
A Clay GTM engineer experimented with giving every account a persistent workspace and a dedicated subagent. Each subagent reviews primary sources and updates its deal folder overnight. Each morning, a coordinating agent turns those updates across all accounts into a short list of priority moves: answer a lingering customer question, fill a gap on the buying committee, or give a prospect a reason to re-engage.
Clay says the workflow saves that engineer roughly an hour of inbox triage each night. The daily priorities help ensure small actions get completed over long enterprise sales cycles. Supporting evidence stays close to each recommendation so sellers can inspect primary sources before acting. Enterprises could extend that shared context to account executives, BDRs, solutions engineers, and sales leaders, respecting existing account permissions.
Clay’s example shows what evolving work needs to scale: a consistent structure, a useful refresh cadence, shared evidence, and human judgment at the point of action.
Exa Labs: from discovery signal to tested integration
Exa Labs, which builds web search infrastructure for AI agents, aims to make its search API available wherever developers might use it — an objective the team calls "Exa everywhere." Previously, pursuing that goal required developer relations and account teams to monitor repositories and the broader ecosystem, identify promising integrations, gather context, and coordinate work across systems.
Although opportunities varied, the path from discovery to implementation followed a consistent sequence. Exa turned that sequence into a defined workflow for Codex, with clear priorities, access to necessary sources, and human review before anything ships.
Codex now monitors for high-priority integration opportunities, gathers relevant context, creates pull requests, runs tests, and prepares weekly updates using sources such as Slack and Notion. When appropriate, it can also draft the next step, including an initial announcement, for the team to review. The workflow carries an opportunity from signal to tested artifact while reducing handoffs across research, engineering, and communications.
People still decide which opportunities matter, what commitments Exa should make, and how to manage external relationships. Tests and review points make the agent’s work visible before it ships, and test results plus human feedback show where to adjust the workflow before the next run. As the work becomes more consequential, permissions, evidence, and decision rights become a larger part of workflow design.
Shared patterns and guidance for leaders
Together, Basis, Clay, and Exa translate the patterns in Enterprise Signals into operational practice. Common elements include:
- Converting proven processes into reusable skills (Basis).
- Providing agents with persistent context and refresh mechanisms so evolving work stays current (Clay).
- Adding tools, tests, and human review so agents can carry a signal into bounded execution (Exa).
All three make improvement part of the workflow: onboarding exceptions reveal where a skill needs refinement; new account activity and seller validation keep deal context current; tests and human review clarify boundaries for future execution. Each starts with a specific job and enough space to test; division of labor clarifies through use, and responsibility grows as the workflow proves itself.
With the frontier gap widening, enterprise leaders should give employees room to pilot consequential workflows, measure outcomes, and convert the strongest experiments into repeatable practice.
