While the Salesforce AI conversation revolves around Agentforce and shiny new agents, Marketing Cloud Account Engagement (Pardot) has quietly carried a full set of AI features for years — and if you're on the Advanced or Premium edition, you already own all of them. Most instances we audit have exactly none turned on.
Here's the complete 2026 lineup: five predictive features and one generative assistant, what each actually does, what it requires, and the honest read on which ones earn their keep. (For the big-picture map of AI across the whole Salesforce ecosystem, start with our executive guide — this post is the MCAE deep dive.)
Traditional Account Engagement scoring is arithmetic you wrote: +10 for a form fill, +3 for a page view. Einstein Behavior Scoring is different — it watches your prospects' actual engagement patterns, learns what buying behavior looked like before past conversions, and assigns a 0–100 score reflecting how much this prospect's recent activity resembles a buyer's. It catches the prospect your rules missed because nobody thought to write a rule for their pattern.
The right move is to run both. Your rules-based score encodes strategy; the Einstein score catches what you didn't encode. Route on the combination — "rules score above X or behavior score above Y" — and review disagreements monthly, because that's where you learn. (If your scoring foundation itself is shaky, fix the framework and the scoring-vs-grading split first; AI on top of a broken model just automates the brokenness.)
Where Behavior Scoring reads marketing engagement, Einstein Lead Scoring analyzes your historical Lead records in Salesforce — which ones converted, and what they had in common — and scores new leads on resemblance to past winners. It's a fit signal from CRM history, complementing the engagement signal from Behavior Scoring. Together they answer the two questions every SDR triages by: does this person look like our buyer, and are they acting like one right now?
Campaign Insights watches your connected campaigns and surfaces the factors driving results — which personas, industries, or regions are engaging unusually well or badly. It's the difference between "the webinar campaign underperformed" and "the webinar underperformed with exactly the segment we built it for, but pulled a persona we weren't targeting." The monthly report gets less archaeological.
Rules-based attribution models (first touch, last touch, even splits) are all wrong in different ways, and everyone knows it. Einstein Attribution uses machine learning on your actual campaign-to-revenue history to distribute credit across touchpoints based on what demonstrably influenced closed deals. It requires Connected Campaigns and Customizable Campaign Influence to be set up properly — which is exactly the plumbing most orgs haven't finished, and the reason marketing and sales reports never agree. Do the plumbing first; the AI model is only as credible as the campaign membership data feeding it.
This one builds a scoring model from your historical opportunity data and tiers accounts by how much they resemble the accounts you've actually won. Instead of the ABM target list being whoever sales gut-picked in January, it's data-ranked — and the tiers are usable inside automation rules, dynamic lists, and engagement programs, so "tier A accounts get the high-touch nurture" becomes an actual automation instead of a slide.
Since early 2024, Advanced and Premium editions also include generative AI in the asset builders: Einstein Assistant drafts subject lines, email headlines and body copy, and landing page and form content, with accept/redraft/edit controls. It's a blank-page killer, not a marketer replacement — the drafts are competent and generic, and the value is editing a starting point instead of staring at an empty canvas. Setup note: it requires Einstein Generative AI enabled in Salesforce Setup, and you must connect each Business Unit explicitly — even if you only have one.
Two clarifications that save disappointment. First, features like Send Time Optimization and Engagement Frequency belong to Marketing Cloud Engagement — the separate B2C platform that shares the brand name — not to MCAE. If a listicle promised them to you, it mixed up the products. Second, there are no native autonomous agents inside MCAE: Agentforce lives on the core platform, where it can work the leads and contacts your MCAE instance creates and nurtures, but campaign-building agents are part of Salesforce's next-generation marketing product, not today's Account Engagement. (What that roadmap means for MCAE's future is its own conversation.)
The requirements are consistent across the predictive features: Account Engagement Advanced or Premium (Growth and Plus editions don't include Einstein — it's the headline reason to upgrade), a Salesforce edition of Enterprise or above, a verified Salesforce connector with prospects syncing to leads and contacts, Connected Campaigns enabled (mandatory for Campaign Insights and Attribution), and — the one nobody can shortcut — history. These models learn from your data: months of engagement activity, a meaningful base of converted leads, closed opportunities to learn attribution from. A young or messy instance produces weak models, which is one more reason clean sync plumbing and consistent campaign hygiene come first.
Enablement itself is mostly toggles in Setup plus surfacing the scores where people work — list views, Lightning record pages, routing rules, dynamic lists. That last part is the difference-maker: an Einstein score nobody routes on is trivia.
If you're on Advanced and paying for these features anyway, leaving them off is leaving money on the table — Behavior Scoring and Campaign Insights in particular are low-effort, real-signal wins once the connector and campaigns are clean. Attribution is powerful but earns trust only after the campaign-membership plumbing is genuinely done. And the generative assistant is a pleasant accelerant, not a strategy.
Turning this stack on — prerequisites, enablement, and wiring scores into routing that sales actually follows — is standard work inside our Account Engagement retainer. If you're on Advanced and none of this is live, that's the fastest ROI conversation we know how to have. Let's talk.
Founder of Thompson Technology. Salesforce and Account Engagement consultant for B2B companies.
Connect on LinkedIn →Want an outside read on where your revenue engine is leaking? That’s a 30-minute conversation — no deck, no obligation.
