Weekly Exa lookalike outbound list from HubSpot closed-won deals

Every Monday at 8am, turn last week's closed-won customers into a fresh, deduped outbound target list for the SDR team. Use open-web research, not a firmographic database, so each pick comes with live signals from the candidate's own site.

Step 1. Pull last week's wins. In HubSpot, use Search Deals to find deals whose deal stage moved to closed-won in the past 7 days. For each matching deal, follow the association to the company and collect the company name, domain, industry, size or headcount, and any other useful firmographic properties. This is the seed set of customers we want more of.

If the week had no closed-won deals, skip Steps 2 through 5 and post a short Slack note saying there was nothing new to model from this week, then finish the run.

Step 2. Find lookalikes on the open web. For each seed customer, use Exa Search and Exa Find Similar together to surface 5 to 10 companies that look like that customer: same vertical, comparable size, similar product or buyer profile. Pull fresh content from each candidate's site so you have concrete signals (what they sell, who they sell to, recent positioning) to justify the match. Be conservative: if a candidate's site does not clearly support the match, drop it.

Step 3. Dedupe against HubSpot. For each surviving candidate domain, run HubSpot Search Companies on the domain. If a record already exists, drop the candidate. We only want net-new accounts.

Step 4. Apply a strict ICP filter and a volume cap. Skip candidates that look like agencies, consultancies, holding companies, parent shells, or otherwise off-profile. After ICP filtering, cap the total at roughly 25 net-new companies for the whole run so the SDR team can actually action the list. If you have more good candidates than slots, prioritize the strongest matches across customers, not the strongest matches for any single customer.

Step 5. Create the records. For each remaining lookalike, use HubSpot Create Company with the basics (name, domain, industry, size). In the description property, write a short Exa-sourced note in the form 'Lookalike of [customer name]: [one or two sentences explaining the match, drawn from the candidate's site].' Keep the note under ~300 characters so it stays scannable in the CRM.

Step 6. Post the digest. Use Slack Send a Message to post a single message to the sales channel (default #sales) summarizing the run. Group the picks by the source customer they were modeled on. For each pick include the company name with domain, a one-line rationale, and a link to the newly created HubSpot company record. Keep the message scannable: grouped headings and a short bulleted list per group, not a wall of text. If the cap was hit, mention how many additional candidates were considered but skipped.

Trigger: cron, every Monday at 08:00 in the workspace's local timezone. Integrations used: HubSpot (Search Deals, Search Companies, Create Company), Exa (Search, Find Similar), Slack (Send a Message).

Additional information

What does this prompt do?

Reads the previous week's closed-won deals in HubSpot, so each fresh outbound list is modeled on the customers you actually just signed.
Researches the open web for 5 to 10 lookalike companies per customer, pulling live signals from each candidate's site to justify the match.
Checks HubSpot before writing anything so existing accounts are never recreated, and skips weak matches to keep the list tight.
Creates up to 25 net-new companies in HubSpot with a short 'why this matches' note, then posts a grouped digest to your sales Slack with a link to each new record.

What do I need to use this?

A HubSpot login with permission to read deals and read or write companies.
An Exa account for the open-web research.
A Slack workspace and the channel where you want the Monday digest posted.

How can I customize it?

Change the schedule, the closed-won lookback window, or the weekly cap on new accounts.
Tighten what counts as a real lookalike: industry, region, headcount band, or buyer profile.
Switch the Slack channel, group picks differently in the digest, or route by territory.

Frequently asked questions

Will this duplicate accounts we already have in HubSpot?

No. Before creating anything, the workflow checks HubSpot by company domain and skips any candidate that already exists, so only true net-new accounts get added.

How many new companies does it add per week?

The default cap is roughly 25 per run, sized so an SDR team can actually work the list. You can raise or lower it any time.

What if last week had no closed-won deals?

The workflow skips the prospecting step and posts a short note in Slack saying there was nothing new to model from this week.

How are the lookalikes picked?

The agent uses semantic web search and a find-similar pass to surface companies that resemble each customer on vertical, size, and product or buyer profile, then reads each candidate's site to write a short rationale.

Can my reps see why a specific company was suggested?

Yes. Each new HubSpot company carries a short description note explaining the match, and the Slack digest lists picks grouped by the customer they were modeled after.