Account intent evidence workspace

Account intelligence inside your agent

Demandbase research that separates signal from intent

Turn an ICP and an account question into a reviewable shortlist. Web activity, campaign engagement, technology change, and hiring become dated observations—not automatic proof that a buying committee is active.

Open installation runbook

Account research specification

Identity resolution

Input
Domain, CRM account, business unit, region.
Output
Resolved account with conflicts retained.

Signal interpretation

Input
Dated first-party and connected observations.
Output
Evidence summary separated from an intent hypothesis.

Buying committee

Input
Required functions, seniority, and problem context.
Output
Candidate roles with verification state.

Activation boundary

Input
Approved account evidence and suppression policy.
Output
Human-reviewed next action; no automatic sending.

An auditable account result

SAMPLE

Account candidate

Atlas Security · identity unresolved sample

  • ICP match: provisional
  • Observed topic: dated evidence required
  • Committee role: verify responsibility

Six controls before an account moves

Resolve the domain

Prevent subsidiaries, redirects, and shared brands from merging silently.

Date every signal

An old visit or hiring post should not appear current without a timestamp.

Separate fact and inference

Preserve the observation independently from the buying-intent interpretation.

Map role responsibility

A title is a clue; verify actual function and seniority before selection.

Record exclusions

Agencies, existing customers, competitors, and restricted regions need explicit handling.

Require review

Relevance, contact status, compliance, and message claims pass a human gate.

Copy, run, configure, first result

  1. 01 Copynpx -y @okki-global/okki-go-taroball

    Clipboard success is not installation success.

  2. 02 Run

    Execute in a supported terminal and inspect package prompts.

  3. 03 Configure

    Keep connected credentials in environment secrets with minimum scope.

  4. 04 First result

    Test known accounts and inspect identity, recency, provenance, and exclusions.

Questions for account-based teams

Does activity prove purchase intent?

No. It is an observation whose meaning depends on identity resolution, topic relevance, recency, baseline behavior, and other plausible explanations.

Can the workflow identify anonymous visitors perfectly?

No. Account identification varies by network, device, geography, consent context, and provider method. Treat matches as evidence to review.

Does it send outreach?

No. Research, drafting, approval, and sending remain separate. Human review and regional requirements still apply.

Are match or reply rates guaranteed?

No. Establish representative cohorts and measure your own acceptance, bounce, suppression, and response outcomes.