ZoomInfo
HackerOne
Apollo
Clay
Sumo Logic
G2
Chrome Web Store
AppSumo
Updated October 8, 2026
Prospecting automation, defined
Prospecting automation is software doing the repeatable part of prospecting on a schedule: search for people who match your ideal customer profile, drop the ones you already have, find and verify their contact details, and hand the list to whoever works it. The judgement stays with your team: who to contact, what to say and when.
- Search: people who match your ICP today, by title, seniority, function and company.
- Dedupe: leave out anyone delivered before and, if you choose, anyone in your CRM.
- Verify: a work email for each person, checked before anyone sends to it.
- Deliver: one list where the team works: Slack, email, a sequence or the CRM.
In Cleanlist AI one agent runs all four steps every week. You describe the job to Clu once; the agent finds the week's new people, checks their emails and posts them.
What's the best AI tool to build a lead list from a prompt?
The best tool turns your sentence into search filters you can read, sizes the result before you spend anything, and returns people with verified contact data. Check the filters before you trust the list: a prompt such as “sales leaders at mid-size SaaS companies” can mean a dozen different searches, and the filters show which one ran.
What to look for:
- It shows the filters it used, so you can fix the one it got wrong.
- It sizes the audience for free: a count and a sample before a credit is spent.
- It removes people you already have: past lists and CRM contacts.
- It charges for results: a missed email costs nothing.
- It runs again: next week's list holds only new people.
In Cleanlist AI you type the sentence to Clu. Clu turns it into People Search filters from your company profile, sizes the audience for free and shows a sample. Once you accept the play and turn the agent on, the search runs every week and delivers only people it has not sent you before.
For a one-off list, 12 prompts translated into filters shows how a sentence becomes a search.
The weekly ICP leads agent, step by step
The agent is five steps: a people search on a schedule, an email search, a second email check, a branch that keeps valid emails, and a Slack post. Two settings decide what you get each Monday: what counts as a new person, and how many the week can hold.
What counts as new
New means this play has not delivered the person before: the agent keeps a record of everyone each play has sent. Skip leads already in CRM adds a check against Salesforce or HubSpot contacts and leads, narrowed if you like by Only if the record has an owner or Only if recently active. On Salesforce it needs a LinkedIn URL field on Contacts or Leads; if the CRM cannot be read, the run stops.
How many a week
Set Leads per week from 1 to 5,000. The agent works through the week's number in batches and sends one post for the period. If fewer people match than you asked for, Clu gives you both numbers before it proposes the play.
Write your ICP once: the filters a people search takes
A people search in Cleanlist AI filters on the person and on their employer: job title, seniority, function and location for the person; industry, headcount and headquarters location for the company. Funding stage and revenue are company facts, so for those the agent finds the companies first and then the people at each.
| Filter | What it takes | Example |
|---|---|---|
| Job title | Titles to include or exclude | VP of Sales, Head of Sales |
| Seniority | From Owner / Partner and CXO down to Entry and Training | Vice President, Director |
| Function | The person's department | Sales, Marketing |
| Person location | Country, region or city | United States |
| Industry | The employer's industry, or a group such as Software & IT | Software & IT |
| Company headcount | Size bands from 1-10 to 10,001+ | 51-200, 201-500 |
| Company HQ location | Where the employer is based | United States, Canada |
| Funding stage, revenue | A company search first, then Find People | Series B |
Clu writes these from your company profile. Change them with Edit search filters, or ask Clu in the agent's chat. For the profile itself, see how to build an ICP.
For an ICP defined by funding or revenue, Clu builds the play from New companies match your ICP with those filters, then a Find People step with the titles. The weekly number then counts the people found.
How to source your own leads?
Source your own leads by writing down who buys from you, searching for people who match, removing anyone you already have, finding and verifying a work email for each, and saving the list where your team works. Then repeat it every week without re-adding last week's names. That last part is where sourcing by hand usually slips.
- Write the ICP as filters: titles, seniority, industry, company size and location.
- Search: LinkedIn Sales Navigator or a people search. Export the matches.
- Remove what you have: past lists, customers, open deals and contacts a colleague owns.
- Find and verify emails: a finder for the address, a verifier before anyone sends.
- Save it where it gets worked: a CRM view, a sequence, a shared sheet.
- Repeat next week, with a record of everyone you already sent.
None of it needs a purchased database: a people search, an email finder and a verifier cover it. The manual method in detail: how to build a B2B lead list, what a good prospect list looks like and list building.
Only verified emails reach the post, and a miss costs nothing
The play checks each email twice. Enrich keeps an address only when it comes back verified or likely valid, and charges 1 credit only then. Validate checks each kept address again for 0.5 credits, and the Branch passes only the ones that come back valid. A person with no email found costs nothing at either step.
| Status after Validate | What it means | What the play does |
|---|---|---|
| valid | The address passed the deliverability check | Posted |
| catch_all | The domain accepts any address, so this one cannot be confirmed | Stops at the Branch, unless you add it to the arm |
| risky | The check found warning signs | Stops at the Branch |
| invalid | The address will not receive mail | Stops at the Branch |
| unknown | The check could not decide | Stops at the Branch |
Without the two extra steps, the list still holds only addresses Enrich found verified or likely valid. Add them when the list feeds a sequence directly. More on catch-all emails and email verification.
From the list to a sequence
Send the week's people to a sequence in one of three ways: approve them from the Slack post, add an Add to sequence step to the play, or sync them to Salesforce or HubSpot and let the cadence your team already runs pick them up. The first keeps a person in the loop; the other two run on their own.
- Approve from Slack: set Approve to sequence on Notify and the post gets an Approve button that enrolls the listed people. It expires after seven days.
- Add to sequence: each verified person goes into a Cleanlist sequence, a lemlist campaign or a Salesloft cadence.
- CRM Action: each person is synced with your CRM's record type, owner and field mapping; someone already there is matched and updated under your write rules.
Cleanlist Sequences send email and LinkedIn steps from your own accounts, a reply stops the sequence, and sending costs no credits. For timing and examples, see sales sequences.
Clu has no tool that sends email by itself. Email goes out only through a sequence you set up, or through an agent you accepted and turned on.
How can AI be used to generate sales leads?
AI helps at every step before the conversation: it turns a description of your buyer into search filters, researches accounts, scores fit against your ICP, finds and checks contact data, drafts first lines and watches for changes such as a new leader at a target account. People still decide who to contact and run the call.
This page covers the weekly list. For the wider build, see how to automate lead generation and the AI lead generation tools we compared.
Prospecting automation tools compared
Prospecting automation tools differ in how much they do without a person. All-in-one platforms hold the data and the sending. Workflow builders let you assemble your own. Autonomous agents find, write and send with little review. CRM-native agents work inside the CRM you have. Pick by how much of the outreach you want a person to approve.
| Kind of tool | What it does | Examples | Best when |
|---|---|---|---|
| All-in-one sales platforms | A contact database, sequences and AI that builds lists and writes outreach | Apollo, Amplemarket | You want data and sending from one vendor |
| Workflow builders | Data from many vendors, AI research agents and plays you build on a canvas | Clay | Someone on the team maintains the workflows |
| Autonomous outbound agents | Find prospects, write and send the emails, answer replies, book meetings | Artisan (Ava, which Artisan calls an AI BDR) | You want outbound run end to end |
| CRM-native agents | Research accounts, suggest contacts and draft outreach inside the CRM | HubSpot's prospecting agent | Your team lives in HubSpot |
| Scheduled list agents | Run your ICP search every week, verify the emails and post the new people; sending goes through a sequence you choose | Cleanlist AI | You want a verified list each week and decide what gets sent |
From each vendor's own pages, read October 8, 2026: Apollo describes AI agents for research, list building, outreach and follow-ups; Amplemarket's Duo builds email, call and social sequences; Clay offers 200+ data and AI vendors and Claygent research agents; Artisan calls Ava an autonomous AI BDR. HubSpot's agent finds new contacts through one connected data source, such as Apollo or ZoomInfo, and sends after review or automatically (our review).
Where Cleanlist AI fits: the agent does the list half of the job every week, with emails from 25+ providers, charged only when found. It sends only through a sequence you set up, automatically or after you approve people from Slack. For the wider market, see the AI prospecting tools we tested.
What 100 new leads a week costs, and which plan runs it
At 100 new people a week, the play on this page costs at most 150 credits a week: 1 credit for each deliverable email found and 0.5 for each check. The search, the Branch and the Slack post cost nothing, and a person with no email found costs nothing.
| What | Credits | Charged when |
|---|---|---|
| People search (the trigger) | 0 | Never |
| Work email | 1 | Only when a deliverable email is found |
| Email check (Validate) | 0.5 | For each person who has an email |
| Phone number (optional) | 10 | Only when a number is found |
| Branch, Notify, Add to sequence | 0 | Never |
That is at most about 650 credits a month. Each Pro seat adds 1,500 credits a month to one shared balance, and unused credits roll over.
Running the agent needs Pro (from $89 a seat a month, $67 billed annually), which includes 5 active agents across the workspace, or Enterprise for unlimited agents. The 14-day Pro trial lets you build the agent and review its play; testing it and turning it on need a paid plan. Every agent Clu builds carries two limits, records per run and a credit limit counted across each week's runs, and runs stop when the agent reaches it.
