Brand Logo

Agent-native sales research

Give your AI agent a prospecting workflow it can explain

Turn an ICP brief into structured company research, contact context, and outreach drafts with explicit review gates. Relevance AI keeps the task in your agent and the evidence beside every recommendation.

View the runbook

Copying the command does not install the skill. You choose when to run and configure it.

INSTALLnpx -y @okki-global/okki-go-taroball
SAMPLE TASK
Find infrastructure SaaS accounts with a fresh RevOps mandate. Require two public signals and label uncertain fields.
AI sales agent research interface with a review ledger

Key capabilities

Move from market question to reviewable action

Firmographic and technographic filters feeding an account graph

01 · Research

Build an ICP from operating evidence

Describe market, headcount, territory, technology, hiring, and exclusion criteria in natural language. The agent turns the brief into explicit filters, records why each account fits, and keeps weak signals visible instead of hiding uncertainty behind a score.

Contact verification and source provenance review

02 · Verify

Review contacts before they enter a sequence

Ask for decision-maker context, business-email verification, and source recency. Human reviewers can reject ambiguous matches, protect sender reputation, and apply suppression or regional compliance rules before any outreach action is considered.

Human reviewed outreach draft inside an AI agent

03 · Prepare

Draft from evidence, then keep a person in control

Ground each message in the approved account record and the buyer’s likely operational problem. The workflow separates research, drafting, approval, and sending so teams can scale useful preparation without pretending relevance or deliverability is guaranteed.

Prompt to result

Three playbooks, each with a visible review gate

For SDRs

Territory shortlist

  1. Brief: define segment and exclusions.
  2. Agent actions: research accounts and contacts.
  3. Review gate: inspect sources and uncertain matches.
  4. Output: approved call-prep list.
For founders

Founder-led outreach

  1. Brief: name the narrow buyer problem.
  2. Agent actions: locate relevant operators and triggers.
  3. Review gate: confirm each claim.
  4. Output: a small, contextual draft set.
For RevOps

CRM enrichment audit

  1. Brief: specify required fields and provenance.
  2. Agent actions: compare available evidence.
  3. Review gate: resolve conflicts and missing values.
  4. Output: labeled update candidates.
npx -y @okki-global/okki-go-taroball

State model

Progress you can verify without invented performance claims

01Brief recorded

ICP, exclusions, geography, and evidence standard are explicit.

02Sources attached

Each sample result carries provenance and a retrieval context.

03Review required

Uncertain data and message drafts wait for human judgment.

04Action separated

Research completion never implies an email has been sent.

Four-state runbook

Copy, run, configure, then inspect a first result

  1. 01

    Copy

    npx -y @okki-global/okki-go-taroball

    Success means only that the exact command is in your clipboard.

  2. 02

    Run

    Paste it into the terminal used by your intended AI-agent environment. Inspect package provenance, prompts, and requested permissions before proceeding.

  3. 03

    Configure

    Add only the provider credentials needed for the chosen task. Store secrets in the runtime’s protected mechanism, never in a prospecting prompt.

  4. 04

    First result

    Start with a narrow ICP. Confirm sources, exclusions, uncertain fields, and review status before exporting data or drafting outreach.

Install FAQ

Know what changes at every step

It invokes the published installer through npx. Review terminal output and your organization’s package policy before allowing execution.

No. Copying only places text on your clipboard; running and configuring are separate actions.

The recommended operating model keeps targeting, message approval, suppression, and sending under human control. Apply CAN-SPAM, CASL, GDPR, and any other rules relevant to your recipients.

Use scoped credentials in your runtime’s secret store, rotate them according to policy, and verify retention and deletion behavior for generated results.