# How many AI agents can safely contact the same account?

**There is no safe number of agents; there is a safe total of touches, and
almost nobody can see theirs.** Every agent, sequence and person reaching the
same buyer counts against one total, and the limit that matters is the one
on that total — set by you, per person and per opportunity.

Two agents working one account do not know about each other. Neither does
the marketing sequence, the event follow-up or the account executive. Each
stays within its own limit, and the buyer receives the sum.

## Count the person, limit the opportunity, watch the account

| level | what it is for |
|---|---|
| **the person** | always counted, because it is where the damage lands |
| **the opportunity or territory** | where limits are enforced, because that is where touches genuinely compete |
| **the account** | seen, not enforced — the *do they think we're disorganised* view |

**Enforcing at the account is the tempting mistake.** It lets activity in one
region starve another that is working a different deal with different
people. Support is outside the count altogether: limiting a reply to
somebody's problem is absurd.

## Who sets the numbers

You do. Limits differ by motion, by relationship stage, by channel and by
seniority, and the right figures depend on your market. What you need from a
tool is the dimensions and the total across every system, because the total
is the thing no single system has.

**A few limits are not yours to set.** Mailbox providers publish complaint
and bounce thresholds, and DNS has its own hard limit on who may send as your
domain. No setting overrides those; you only choose what happens as you
approach them.

## What *safe* needs that most teams lack

A total that includes everything, or one that says what it is missing. A
count built only from your agents' own traffic, presented as the whole, is
worse than no count: it says *quiet* while marketing is mid-sequence. So the
total says which sources reported, and until you name what exists, it says
the picture is partial.

</how-often-should-an-ai-agent-email-the-same-person.md> covers the limits
in detail.