Skip to content

Blog

Marketing & Growth

AI Lead Scoring in Your CRM: Start With the Rules You Already Trust

Add AI lead scoring by writing down the rules sales already trusts, cleaning CRM data, then testing the AI score against those rules before reps use it.

7 min read
Portrait of Omer Faruk Ilhan.
Omer Faruk Ilhan
Business Analyst
Dithered illustration of a fragmented marble face looking down at a jeweller's tray where gems are sorted by size, the largest stone glowing amber under a brass lamp.
logo of Facebook
logo of X
logo of LinkedIn
logo of Reddit
logo of Telegram
Contents
No headings found on page

A B2B company should add AI lead scoring by first writing down the rules its sales team already trusts, cleaning the CRM fields those rules depend on, and then running the AI score next to the rule-based score for a few weeks. Let reps act on the AI score only once it agrees with closed deals at least as well as the rules do.

Picture a software company selling to logistics firms. Its sales lead knows a good lead when she sees one: a company with more than a handful of trucks, a job title with "operations" in it, and a demo request rather than an ebook download. None of that is written down. Marketing switches on an AI score in the CRM, and within a week reps are calling students who downloaded three whitepapers, while a fleet manager who asked for pricing sits at the bottom of the list. The team stops looking at the score, and it never recovers their trust.

What is AI lead scoring?

AI lead scoring is a score in your CRM that a machine-learning model calculates by comparing each new lead with the leads that converted, or failed to convert, in the past. A rule-based score, by contrast, adds and subtracts points for criteria a person chose.

Both kinds answer the same question: which leads should sales call first? The difference is who decides what matters. With rules, your team decides and can read every criterion. With AI, the model finds the pattern in your history, which can surface signals nobody thought of, but it also inherits every gap and bias in that history.

How do HubSpot and Salesforce handle AI lead scoring?

HubSpot can draft a contact score with AI that you then review and edit like a rule-based score, while Salesforce Einstein Lead Scoring builds a predictive model from your converted leads and shows reps which fields pushed each score up or down. Both depend on your past conversions being recorded properly.

Here is how the two compare, from each vendor's own help pages (checked November 2026):

  • HubSpot lead scoring: What it is: a lead scoring tool with fit scores (based on property values), engagement scores (based on actions) and combined scores. AI option: contact scores built with AI, which evaluate contacts that moved between two lifecycle stages you choose and recommend criteria and points. Control: you edit the recommended criteria before turning the score on. Plan: contact scores need Marketing Hub Professional or Enterprise; AI-built contact scores need Marketing Hub Enterprise.

  • Salesforce Einstein Lead Scoring: What it is: a predictive model that analyses past converted leads, including custom fields and activity data, and scores current leads by how closely they match. Data needed: at least 1,000 leads created in the last 200 days, with at least 120 of them converted, per Salesforce's setup considerations; with less, it uses a global model built from anonymous data from many Salesforce customers. Control: reps see the top positive and negative factors for each score. Plan: requires Sales Cloud Einstein, included in Performance and Unlimited editions and available at extra cost in Enterprise.

The practical difference: HubSpot's AI hands you a set of rules you can read and change. Einstein gives you a model you can inspect through its top factors but not edit criterion by criterion. If you are still choosing a CRM, our comparison of HubSpot and GoHighLevel for service businesses covers the wider trade-offs.

Why start with the rules you already trust?

Rules you already trust give you a baseline to measure the AI against and a score reps will use from day one. Without a baseline, you cannot tell whether the AI score is better, only that it is different.

A marble hand tracing the carved marks on a weathered milestone beside an old Roman road at dusk, the deepest mark glowing amber

Most sales teams already score leads in their heads. Writing those rules down does three things:

  1. It exposes disagreement. Sales and marketing often have different definitions of a good lead. Settle that before any model learns from the CRM.

  2. It shows which fields matter. If the rules depend on company size and job title, those fields must be filled in and consistent, for the rules and for the AI.

  3. It gives the AI something to beat. A model that cannot outperform a page of rules your team wrote in an afternoon is not worth the cost or the opacity.

A good first rule set is short. Use a few fit criteria, such as industry, company size and role, and a few engagement criteria, such as a demo request, a pricing page visit or a reply to a sales email. Add negative points for clear disqualifiers: students, competitors, personal email domains for a B2B product.

What CRM data does AI lead scoring need?

AI lead scoring needs a reliable record of which leads converted and which did not, plus consistent fit and engagement fields on those leads. If conversions are logged late, inconsistently or not at all, the model learns the wrong pattern.

What to check before you switch it on

  • Conversion is defined once. Pick the event that counts, such as a lifecycle stage change to sales-qualified lead, an opportunity created or a deal won, and make sure everyone records it the same way.

  • Duplicates are merged. One buyer stored three times looks like three leads, two of which never converted. Our post on cleaning CRM data before adding AI covers duplicates, record grain and field ownership.

  • Source data survives. Engagement signals are only as good as the tracking behind them. If campaign parameters are lost between the ad, the form and the CRM, the model cannot learn which channels bring buyers; see why UTM parameters get lost between tools.

  • Old leads are not mislabelled. Leads that were never worked are not the same as leads that were worked and lost. If your CRM cannot tell them apart, the model will treat ignored leads as bad leads.

How do you test an AI lead score before reps rely on it?

Run the AI score and your rule-based score side by side on the same leads, without changing who gets called, and compare both against what actually converts. Switch routing to the AI score only when it ranks the leads that closed higher than the rules did.

Two identical seed trays on a glasshouse bench, a marble hand watering only one, its tallest seedling glowing amber

A simple test plan:

  1. Turn on both scores and keep routing on the existing process.

  2. Wait for outcomes. Choose a window long enough for a typical lead to reach your conversion event.

  3. Compare the top of each list. Of the leads each score put in its highest band, which share converted? Do the same for the lowest band: a good score keeps buyers out of it.

  4. Read the disagreements. Look at leads the AI rated high and the rules rated low, and the reverse. Ask a rep whether the AI's picks make sense. This is where you find new signals, or find that the model is rewarding noise.

  5. Decide and document. Move to the AI score, keep the rules, or fold what the AI found into the rules.

Both CRMs give you something to work with here. HubSpot's score history and performance views show the distribution of scores and the events that changed a record's score. Einstein shows the top factors on each lead, which makes the disagreement review faster.

What goes wrong with AI lead scoring?

The common failures are training on a biased history, rewarding activity instead of intent, and a score nobody owns. Each one makes reps stop trusting the number, and an ignored score is worse than none.

  • Biased history: if your team only ever worked one segment, the model learns that only that segment converts, and it will keep starving the others.

  • Activity over intent: a lead who opens every newsletter scores high while the buyer who sent one email asking for a quote scores low. Weight high-intent actions heavily, and use score decay (HubSpot lets you decay engagement points every 1, 3, 6 or 12 months) so old activity fades.

  • No owner: markets change, products change and the CRM changes. Someone must review the score on a schedule, check it against recent closed deals and adjust it. Einstein reanalyses leads roughly every 10 days, according to Salesforce, but retraining is not the same as someone checking the result.

  • Hidden reasoning: if reps cannot see why a lead scored high, they will not trust it. Show the top factors or the matched criteria on the record.

How we approach it at Vizio AI

We start with the rules, not the model. In a short workshop with sales and marketing we write down what a good lead looks like, agree on the conversion event, and check that the CRM fields behind those rules are filled in and consistent.

Then we build the rule-based score in the CRM the client already uses, fix the data problems the rules expose, and connect the tracking that feeds engagement signals. Only after that do we turn on an AI score next to it, run the side-by-side test and review the disagreements with the reps who will use it. The client ends up with a score their team understands, a record of why it was chosen and a named owner who reviews it.

This work sits in our Growth Studio, alongside CRM setup, attribution and outbound.

Sources

FAQs.

01

02

03

04

05

Read more.

Dithered illustration of two marble hands pulling a mason's line taut between stakes across bare ground at dawn, the first course of stones laid along it and a pink knot at the nearest stake.
Your First 30 Days With a Product Studio: What Happens Week by Week
Tarik Gonen
7 min

The first 30 days with a product studio: week one sets access and goals, week two tests the riskiest assumption, weeks three and four ship a first slice.

Dithered illustration of a fragmented marble face looking down at a jeweller's tray where gems are sorted by size, the largest stone glowing amber under a brass lamp.
AI Lead Scoring in Your CRM: Start With the Rules You Already Trust
Omer Faruk Ilhan
7 min

Add AI lead scoring by writing down the rules sales already trusts, cleaning CRM data, then testing the AI score against those rules before reps use it.

Dithered illustration of a marble hand holding an oil lamp out across a threshold, its pink flame lighting the first stones of a mosaic path.
Onboarding for AI Products: Getting Users to Their First Good Answer
Umut Durgut
8 min

Onboard AI product users by stating what it can and can't do, offering example prompts and sample data, and reaching a useful answer in the first session.

Think AI-first.
Move future-fast.

Let’s build what’s next! From product ideas and intelligent workflows to data-driven systems and scalable growth operations.

Book a Strategy Call