Workflows

Build prospecting as a chain of reviewable agent tasks

Persana AI organizes discovery, enrichment, contact research, message preparation, and signal tracking without collapsing them into a black-box automation. Each handoff has an input, an output, and a human decision.

AI prospecting task chain

Four tasks an agent can coordinate

Account discovery

Translate a natural-language ideal customer profile into firmographic, technographic, geographic, and exclusion criteria. Return candidate domains with match rationale rather than an unexplained list.

Contact research

Map buying functions to likely responsibilities, then verify professional contact data. A title match remains a hypothesis until responsibility and relevance have been reviewed.

Signal context

Observe dated hiring, leadership, technology, or company events while keeping the event separate from a buying-intent inference. Recency windows belong in the task specification.

Outreach preparation

Draft concise context from approved evidence, then stop at a human review gate. Sending policy, suppression, opt-out, and regional rules remain operator responsibilities.

Reproducible workflow demonstrations

RevOps enrichment demonstration

RevOps cohort cleanup

Resolve identity before updating the CRM

Start with a controlled cohort containing redirects, subsidiaries, renamed companies, and shared brands. The agent compares domain, operating name, legal name, location, and dated evidence. It proposes changes but leaves conflicting values unresolved.

The operator records accepted source, rejected alternative, observation date, and rollback path. This demonstration measures correction work and explainability; it does not claim a universal match rate.

SDR account brief demonstration

SDR account brief

Prepare relevance without authorizing a send

Use an accepted account to assemble current business context, a role hypothesis, source dates, and open questions. Contact enrichment follows only after the account rationale passes review.

A message draft can then reference approved facts while avoiding sensitive or speculative personalization. SPF, DKIM, DMARC, sender reputation, suppression, and opt-out controls remain separate prerequisites for outreach.

4 task statesDiscover · enrich · review · approve
3 evidence labelsObserved · inferred · unresolved
1 human gateBefore connected outreach
0 outcome promisesCoverage and replies vary

Install the workflow inside your agent

Copy the exact command, run it deliberately, configure minimum-scope credentials, and validate the first result against companies you already know.

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