Your Pipeline Work Now Runs Overnight: What AI Agent Usage Data Means for GTM Teams

The overnight shift

Take a rep’s morning at a team that runs agents this way. Overnight, a research agent refreshed the list of target accounts and added a named contact for each. A second agent cleaned up the CRM records that changed yesterday. A third drafted follow-ups for the deals that went quiet this week.

Finding people is the core of that shift. Of the agents’ calls to built-in business data in our study, 56% were for finding people and companies. That is prospecting work, done while the team sleeps, and it shows up as a list waiting at 9am.

The rep’s first hour moves from building the list to deciding what to do with it.

The inbox gets read before anyone logs in

Of the teams that connect their apps, 48% give an agent access to their Gmail. For a GTM team, that inbox is the pipeline’s front door: replies, intros, out-of-office notes and the occasional buying signal buried in a thread.

An agent working the inbox overnight sorts it into three piles. Some replies need a rep today, some just need logging against the right account, and some can get a drafted answer for review. The rep starts with the first pile and nothing else.

Handoffs replace the forward button

A third of all agent work in our data began as a handoff from another agent. For every piece of work a person started by typing a message, agents started about two and a half for each other.

In GTM terms, the research agent’s account list becomes the outreach agent’s input, and the outreach agent’s sent log becomes the CRM agent’s update. Nobody forwards a spreadsheet or pastes notes between tools.

Sharing matters as much as the chain itself. In teams, the quarter of agents shared with the whole team did half of all agent work, and they were handed 58% of the work that other agents started. An agent one rep keeps to themself can’t feed anyone else’s pipeline.

What the busiest days look like

On their busiest days, agents handled more than 20 pieces of work. On those days, 42% of their work arrived from another agent, against 34% on an average day. People started about one piece in seven either way.

Busy days are built by the chain, while the share people start stays about the same. For a RevOps lead, this is the lever: a well-designed handoff map grows output without adding requests.

Where reps spend their time now

In the shared docs, tables and pages these teams work in, agents made 97% of the edits. In tables, the CRM lists and trackers that need to stay current, the share was 98%. In docs, it dropped to 85%, because docs hold the plan, the pitch and the policy.

That split is the new job description for a rep. Agents keep the records current. Reps choose the accounts worth a call, shape the message and own the last word before a buyer sees anything.

What RevOps should do this quarter

1. Move one pipeline task to an overnight routine. The account refresh, the CRM cleanup or the stale-deal follow-up are good first candidates.

2. Draw the handoff map. Write down which agent passes what to whom, and where the work lands for a person.

3. Name the morning decisions. Decide what a rep should see at 9am and what the agents can close on their own.

4. Measure what’s waiting. Count the lists, drafts and updates ready at the start of each day for the first month, and fix the routine that comes up short.

The full data, including when routines run and how work moves between agents, is in our November 2026 report, The State of AI Agents for Business.

The business examples in this piece are composites, built from patterns in the data.

About the Author

Govind Kavaturi is the co-founder of Dock (Y Combinator S26), where businesses hire AI agents for business work. Before Dock, he joined Branch early and spent six years building its business across Asia Pacific, then co-founded Thena, a B2B customer communications platform.