revnu

AI Prospecting Workflows That Keep Research Moving

By George Jefferson · November 18, 2026 · 9 min read

Ask AI about this pageChatGPTClaudePerplexityGoogle AI
AI Prospecting Workflows That Keep Research Moving

An AI prospecting workflow is a recurring loop, not a one-off list pull. It discovers accounts, removes repeats, qualifies them against written criteria, saves the evidence, drafts outreach, waits for approval and reports what happened. The loop keeps research moving when nobody has an hour to rebuild a list on Monday morning.

Most AI prospecting content lists tools that find more contacts. That isn't the bottleneck. Founders and first sales hires who have tried a database or an enrichment tool still lose hours each week to the same chores: rebuilding lists, checking whether they already contacted someone and chasing context for the reply. A recurring workflow turns those chores into stages that run on a schedule, with a human holding the one decision that matters, which is who gets contacted.

This post walks through the seven stages, separates what can run unattended from what needs approval, and shows where Revnu runs the loop for you.

What an AI prospecting workflow is, and what makes it recurring

An AI prospecting workflow is a repeatable sequence of research and drafting steps that an agent runs on a schedule, ending in a report rather than a spreadsheet. The word that matters is recurring. A one-time export of 500 companies is a project. The same search running every week, skipping everyone it has already seen and adding only new, qualified accounts, is a workflow.

The loop has seven stages: discover, dedupe, qualify, save evidence, draft, approve, report. Each stage has an input, an output and an owner. Discovery produces candidates, not approvals. Qualification turns candidates into qualified, rejected or needs-review. Drafting turns saved research into messages. Approval is where a person decides what leaves the building.

This answers a question buyers ask constantly: which agent can run recurring prospecting and reporting workflows? The honest test isn't whether a tool can find contacts. It's whether it can run all seven stages without you rebuilding the list each week, and whether it keeps the research schedule separate from sending permission. Zapier Agents, n8n and Make give you building blocks to wire that yourself. Clay gives you a research spreadsheet someone has to maintain. Revnu is an agent that runs the whole loop and reports back to you in Slack or iMessage.

Step 1: Define the target and the exclusions before the agent searches

Write the target and the exclusions down before the first search runs, because an agent searches exactly as wide as you let it. A useful target has four parts: the account profile (industry, size, product type), the buyer role, the signals that suggest timing, and the exclusions.

Exclusions do more work than most teams expect. List current customers, open deals, competitors, partners, people who asked not to be contacted and any segment you've decided is a bad fit. Without them, the first weekly run will happily surface your own customers as prospects.

Be specific about signals. "Growing B2B software company" is vague. "Hiring a first sales or growth role in the last 60 days" is something an agent can check and a human can verify.

Keep the target to one paragraph. If you can't explain it in a paragraph, the agent won't qualify against it consistently, and your week-four results will drift from your week-one results for no visible reason.

Step 2: Run account discovery on a schedule

Schedule discovery weekly so new qualifying accounts appear without anyone remembering to look. Run discovery at the company level first, then identify the right people inside each company. Treat the output as candidate generation. A company showing up in a search is a reason to look closer, not permission to contact anyone.

A weekly cadence suits most small teams. Daily runs mostly resurface the same accounts, and monthly runs miss the timing signals that make outreach relevant, such as a new hire or a fresh funding round. Pick a day, let the agent run, and read the report.

This is the stage that can run fully unattended. Discovery reads public information and writes to a list. It sends nothing, spends nothing and commits you to nothing. That's why it's safe to schedule and forget, and why the approval boundary belongs further down the loop.

Step 3: Deduplicate against what you already know

Deduplication separates a workflow from a list generator. Every run must check new candidates against everything you already know and drop the repeats. Match on company domain first, then person, then check existing conversations, open deals and anyone on a do-not-contact list.

Three kinds of repeat cause real damage. Contacting the same person twice from different sequences looks careless. Prospecting a company that is already a customer or an open deal wastes goodwill. And re-researching an account you rejected last month wastes the agent's time and your attention.

Where a match is uncertain, such as two similar company names or a person who changed jobs, route it for review rather than guessing. A wrong merge is worse than a visible question.

If your system of record is a CRM, keep it. Revnu works alongside HubSpot, Attio or Salesforce rather than replacing them. You can export the contacts you want worked as a CSV or XLSX file and upload it in dashboard chat, and Revnu researches and saves that selected audience. That import is a snapshot, not a two-way sync, so refresh the export when your CRM changes.

Step 4: Qualify with explicit criteria and save the evidence

Qualify every candidate against written criteria and save the evidence beside the verdict, so a human can check any decision in under a minute. Give the agent three outcomes to choose from: qualified, rejected or needs review. Never let it invent a fourth.

Fit and timing are separate questions. Fit asks whether the company matches your profile. Timing asks whether something suggests now is a good moment, such as hiring, funding or a recent public post. Score them separately. A perfect-fit company with no timing signal belongs on a watch list, not in a sequence.

Most workflows skip the evidence and later regret it. For each qualified account, save the source, the date and the specific fact that triggered the verdict. When someone replies three weeks later, you can see why you wrote to them. When a decision looks wrong, you can trace it. Revnu keeps leads in a Leads CRM with a fit score and the reason to reach out, so the evidence travels with the account.

Two habits keep this honest. Don't infer a need from missing information: no sign of a growth hire is not proof you lack one. And write the rejection reason down, because rejected accounts teach you more about your targeting than qualified ones do.

Step 5: Draft outreach, then keep research schedules separate from sending permission

A research schedule can run unattended. Sending permission can't be assumed from it. This is the most important design decision in the loop. Letting an agent discover, dedupe, qualify and draft on its own is low risk. Letting it contact people is a separate grant that you give deliberately and can withdraw.

Draft from saved evidence, not from a generic template. A message that cites the specific signal you recorded in step 4 reads like research. One that doesn't reads like mail merge. Keep rules, not model judgment, in charge of eligibility: who may be contacted, how often, and what stops a sequence. Replies, opt-outs and suppression conditions should end outreach automatically.

In Revnu, the approval model follows this split. On current Pro defaults you approve each new sequence once, and its emails and follow-ups then run under your sequence settings. Prospect replies always wait for you, and so do new spending commitments and budget changes. Uploading a list doesn't by itself authorize sending. You choose how much runs without you, and the research keeps moving either way.

For the mechanics of sequences and replies, see how AI outbound agents for B2B SaaS handle them.

Step 6: Report weekly and feed the results back into targeting

Close every cycle with a weekly report, then change the target based on what it shows. A report nobody reads is the same as no report, so keep it to what a founder can act on in a few minutes.

Track unique accounts and contacts found, how many were new versus repeats, qualification outcomes with reasons, what was drafted, what was approved and sent, replies and meetings. Cut each of these by cohort, such as industry or signal type, so you can see which slice of your target actually produces conversations.

Then feed it back. If one signal produces replies and another produces silence, tighten the target around the first. If rejections cluster on one exclusion you forgot, add it to step 1. If a segment qualifies well but never replies, the problem may be the message, not the list.

This feedback step turns a schedule into a loop. Without it, the agent repeats last month's assumptions faithfully. Revnu keeps one memory across channels, so what works in outbound can inform your search content and ads as well. Don't read too much into a few weeks of data. Early numbers show direction, not proof.

How Revnu runs this loop, and where it fits next to your CRM

Revnu is an AI agent that runs growth work for a business, including lead research and outbound, and reports to you in Slack or iMessage.

The contrast with the usual options is about who runs the loop. Apollo bundles a contact database, sequences and AI assistants. HubSpot Breeze and Agentforce run inside their own CRMs. Zapier, n8n and Make connect steps you assemble yourself. Revnu is the agent that picks the next move, does the work and queues approvals, so you aren't wiring stages together.

On CRMs: Revnu isn't a CRM and doesn't claim native CRM sync. Keep HubSpot, Attio or Salesforce for your existing process and use Revnu alongside it, moving selected contacts in by CSV or XLSX upload. Pro is $99 a month, self-serve, cancel anytime. Managed adds a growth engineer from the team and is custom-priced through a call at revnu.com/book.

For wider context, see startup growth automation tools and AI marketing agents for SaaS.

Start with one target and one weekly run

The loop beats the tool. A recurring workflow that discovers, dedupes, qualifies, saves evidence, drafts, waits for approval and reports will outperform a bigger contact database you rebuild by hand every Monday. Keep research on a schedule, keep sending behind approval, and let the weekly report change your targeting. To run that loop without wiring it yourself, sign up for Revnu Pro at revnu.com, give it your website and one target segment, and read the first weekly report before approving a single sequence.

Written by

George Jefferson

Co-founder and CEO, Revnu

Today 9:41 AM

hey what's going on?
found 12 companies hiring their first sales rep this week. want me to reach out?
let's do it
🙏

Read 9:41 AM

9:41
Revnu

1-Minute Setup

You build.
We grow.

Connect your tools. Wake up to ads scaled, posts ranking, leads queued, churn won back. Pings you a message when something actually moves.

Frequently asked questions

How do you use AI for sales prospecting?

Use AI for the research stages and keep people on the sending decision. Have an agent discover accounts on a schedule, deduplicate them against what you already know, qualify them against written criteria and save the evidence. Then draft outreach from that evidence and approve what goes out. Revnu runs this loop and reports weekly in Slack or iMessage.

Which AI agent can run recurring prospecting and reporting workflows?

Several options exist. Zapier Agents, n8n and Make let you assemble workflows yourself. HubSpot Breeze and Agentforce run inside their CRMs. Clay is a research spreadsheet you maintain. Revnu is an agent that runs the whole loop, from lead research and outreach to results, and reports back to you daily or weekly, with approvals you control.

Can an AI agent run prospecting on a schedule without sending anything?

Yes, and that is the safest setup. Discovery, deduplication, qualification and drafting only read information and write to a list, so they can run unattended. Sending is a separate permission. In Revnu, you approve each new sequence once on Pro defaults, and prospect replies and spending changes always wait for you.

Does Revnu sync with my CRM?

Revnu does not claim native CRM sync. Keep HubSpot, Attio or Salesforce for your existing process and use Revnu alongside it. You can export selected contacts as CSV or XLSX and upload them in dashboard chat for Revnu to research and work. The import is a snapshot, and uploading does not authorize sending.

What should I expect from the first few weeks of an AI prospecting workflow?

Expect a cleaner list and a clearer report before you expect meetings. Early weeks show which signals and segments qualify and which exclusions you forgot. Use that to tighten your target. Results vary by market and message, so treat early replies as direction, not proof, and adjust the target each week.

Keep reading