From today, every Lusha user who opens the platform sees a ranked list of accounts and contacts, refreshed daily, with the reason each one is there. Pick a row, pull the buying group, reveal the right person, and open the call with a dated reason instead of “just checking in”. This is Deep Intelligence working as one product, in the Lusha platform, in the API, and inside Claude or ChatGPT.
This post explains what shipped, how the scoring works, what it does not do, and four ways to put it to work this week.
Why a ranked list, not a bigger database
Every rep starts the day with the same question: who do I work first? A search box does not answer it. A saved list does not answer it either, because every row looks the same until someone reads it. So reps decide by habit, by whoever replied last, or by whoever is loudest in the CRM.
Deep Intelligence answers the question before the rep asks it. It takes four kinds of context that already exist in your account and uses them to rank the market against your business, every day.
What shipped on October 19
In April, Lusha introduced Deep Intelligence as an idea: a second layer on top of the Search layer, unique to each customer. The pieces existed, but many were early, separate, or open to a few accounts. This release connects them into one product a rep uses every morning.
| Capability | What it does | Where it runs |
|---|---|---|
| Recommendations | A ranked list of contacts and companies to work, refreshed daily, with the reason each one is there. | Platform, MCP |
| ICP Hub | Where you define who you want to reach, as criteria or as a sentence. It drives recommendations, not Search. | Platform |
| ICP Fit Score | How closely a company or contact matches your Target Profile, from 0 to 1. | Platform, in recommendations |
| Signals | The why now: job moves, promotions, hiring surges, funding, M&A. They feed the scores and give the rep an opening line. | Platform, API, MCP |
| Buying Group (beta) | The committee inside an account: decision maker, potential champion and end user, each with a role and a score. | Platform, API, MCP |
| Predictive Scoring | How likely a record is to become an opportunity, from Lusha network patterns plus your own HubSpot or Salesforce pipeline. | Platform, MCP, accounts with access |
| Lookalikes | Give it seed accounts or contacts you know are good, get similar ones back on demand. | Platform, API, MCP |
| Website Visits (beta) | Companies visiting your site, scored 0 to 100 and rated Hot, Warm or Cold, with employees one click away. | Platform, API, MCP · Pro and above |
The core loop: no setup required
The design principle is that the ICP is a booster, not a gate. Nothing stands between a new user and a first ranked list.
Step 1: Lusha builds the list
Every day, contacts and companies are scored and ranked on Home with the latest signals. New signal builds land about twice a week and a full re ranking runs weekly. For eligible users, the list is built even with nothing configured, from the account’s own search and reveal history, and labeled “Lusha generated”. Eligibility is simple: activity in the last 90 days, a completed onboarding, or an active Target Profile.
Step 2: every row shows why it is there
Each row carries a fit score from 0 to 1 and, where one fired, the signal: a new role, a promotion, hiring growth, a funding round, a match with a record in your CRM. The signal is a business event, dated, and it is the first line of the rep’s message.
Step 3: the rep works a row
From the row, the rep pulls the buying group for that account, reveals the contacts they need, and uses the signal as the reason to reach out today. Score and contact live in one place. The rep never leaves the list to go find the person.
Sharpening the list: three switches, mostly for admins
The list works on day one. These three steps make it yours.
- Define a Target Profile in ICP Hub. Account criteria (industry, size, geography), persona criteria (titles, seniority, department), or a sentence describing who you sell to. The list re ranks within minutes.
- Share the profile account wide. An admin shares one Target Profile and every rep, including the one who started this week, works the same ICP. Reps can still have their own.
- Connect HubSpot or Salesforce. Once the admin gives account level AI consent, Lusha stops recommending accounts your team already owns. For accounts with Predictive Scoring access, your won and lost deals also train the score.
How the scoring works
Three scores can appear on a row. They answer different questions, and only one of them uses a 0 to 100 scale.
ICP Fit Score
Fit compares each company or contact with your Target Profile and returns a number from 0 to 1. It needs an active Target Profile. Without one, the “Lusha generated” list still ranks on account history and signals, but it will not show a fit score.
Predictive Scoring, and the report that shows its work
For accounts with access, Predictive Scoring estimates how likely each record is to become an opportunity. It learns from two places: patterns across the Lusha network, and your own won and lost deals in HubSpot or Salesforce. Then it does something scoring tools rarely do. The Opportunity conversion report shows how accurate the scores were on your own pipeline, so a RevOps lead can check the model against reality before asking reps to trust it.
Signals
Signals are dated business events: a position filled or vacated, a promotion, a hiring surge in a department, a funding round, a company announcement. They feed the ranking and they give the rep an opening line that is about the account, not about Lusha.
Buying Group: who decides, inside the account
A ranked account tells you where to spend the morning. It does not tell you who to call. Buying Group does. For any account, it maps the committee: the decision maker, a potential champion and an end user, each with a role and a score. You can ask for one role only, for example decision makers, and the rep reveals the right person from the same list.
Buying Group is in beta on the platform, the API and MCP. Reveals use credits from your plan’s shared wallet. Because it is beta, we are not publishing coverage figures for it yet.
Lookalikes and Website Visits sharpen the list
Lookalikes: more accounts like your best ones
Seed Lookalikes with the accounts or contacts you already know are good, for example your ten best closed won deals this year, and get similar ones back on demand. Results land in a Workspace table you can enrich, score and export. Lookalikes feed the daily list too, so the accounts that resemble your winners show up in tomorrow’s ranking without anyone building a search.
Website Visits: the companies already looking at you
Add your domain, set your high intent pages, install and verify the pixel. Visiting companies then appear rated Hot, Warm or Cold on a 0 to 100 score, with their employees one click away. Website Visits identifies companies, not individuals. It is GDPR and CCPA aligned, in beta, and available on Pro and above.
It runs where you work: platform, API, Claude and ChatGPT
The same intelligence runs in three places, and a team can mix them. A rep works Home in the platform. A RevOps engineer routes fit, signals and buying groups into the CRM through the API. A seller who lives in Claude or ChatGPT asks for today’s list in the conversation and reveals the contacts there, through the Lusha connector.
Try it in Claude: connect the Lusha connector, then ask: “Pull today’s recommended companies for my account, ranked by fit, and for the top three show me the buying group.” Lusha returns the ranked list, the roles and the credit cost of revealing each contact before it reveals anything.
What it is not
Honest limits, stated before the first list rather than after.
- It is not discovery. Deep Intelligence ranks and explains. Search still finds. The two work together; neither replaces the other.
- Target Profiles do not filter Search. They shape recommendations only.
- Website Visits is company level. It identifies the organization visiting your site, then shows you the employees Lusha already knows at that company. It does not identify the individual visitor.
- Fit scores need a Target Profile. No profile, no fit score, though the list still ranks.
- Coverage is deepest in North America and thinner in EMEA and APAC today.
- Buying Group is beta. Expect gaps, and tell us where you find them.
What’s included in your plan today
| Capability | Availability |
|---|---|
| Daily ranked list, ICP Hub, ICP Fit Score | Included in your plan today |
| Lookalikes | Platform, API and MCP |
| Buying Group | Beta. Reveals use credits |
| Website Visits | Beta. Pro and above |
| Public API | Pro and above |
| Predictive Scoring | Accounts with access, HubSpot or Salesforce connected |
Plan details are on lusha.com/pricing. Capability status per surface is on docs.lusha.com.
Four ways to start this week
Each one is a Campus play with a prompt you can run in Claude or a flow you can follow in the platform.
- The morning queue. Open Home, work the top five rows, pull the buying group for each, and open every message with the signal on the row. Run the play
- Warm inbound. Check which companies hit your pricing page this week, pull the committee for each Hot account, and reach them before they fill a form. Run the play
- Scale what’s working. Seed Lookalikes with your ten best closed won accounts, build the list in a Workspace table, and let it feed tomorrow’s ranking. Run the play
- Why now outreach. Open with their hiring surge, funding round or new VP, not your product. A past buyer who moves to a new company becomes a warm way into a new account. Run the play
Warm inbound → Website Visits → employees → Signals
Scale what’s working → Lookalikes → Workspace table → Signals
Why now outreach → Signals → Buying Group → first line
FAQ
Do I need to set up ICP Hub before I see recommendations?
No. For eligible users, Lusha builds the first list from your account’s own search and reveal history and labels it “Lusha generated”. A Target Profile makes the list yours and adds fit scores, but nothing waits on it.
Does a Target Profile change what Search returns?
No. Target Profiles shape recommendations only. Manual Search still returns everything that matches your filters.
What is the difference between ICP Fit Score and Predictive Scoring?
Fit asks whether an account looks like who you sell to, measured against your Target Profile. Predictive asks how likely a record is to become an opportunity, learned from your own won and lost deals in HubSpot or Salesforce plus Lusha network patterns. Both use a 0 to 1 scale. Predictive Scoring is available to accounts with access.
Does Website Visits tell me which person visited my site?
No. Website Visits identifies the company, scores its engagement from 0 to 100 and rates it Hot, Warm or Cold. From there you see the employees Lusha already knows at that company. It is GDPR and CCPA aligned.
Is Buying Group live?
Yes, in beta, on the platform, the API and MCP. Reveals use credits from your plan’s shared wallet. We are not publishing coverage figures while it is in beta.
Can I use all of this inside Claude or ChatGPT?
Recommendations, Buying Group, Lookalikes, Website Visits and Signals run through the Lusha connector in Claude and ChatGPT, with reveals in the same conversation. The live list per surface is on docs.lusha.com.
