How LiftPilot’s adaptive AI chooses the perfect CTA

Be honest: If someone returns to your website 20 minutes after a sales call where they expressed budget concerns, what CTA do they see?

Would it be the exact same CTA they previously saw?

If it is, you’re not alone. And this is exactly the problem marketing teams are facing.

That visitor isn’t starting their journey—they’re continuing it. They just spent 45 minutes discussing pricing with your AE. They heard the numbers. They have questions. And when they return to your site for answers it acts like you’ve never met.

The CTA that says “Book a Demo”? They just had one. The hero that highlights features? They’re past that stage.

What they actually need is ROI documentation, customer proof points, or a conversation with someone who can work within their budget. But your website doesn’t know that. So it shows them the same generic message it shows everyone else.

The result? They leave to find what they need somewhere else.

Your CTA isn’t wrong. It’s just the same for everyone. And in B2B, where buying journeys span weeks and involve multiple stakeholders, one-size-fits-all CTAs leave conversion opportunities on the table.

Adaptive CTAs solve this by responding to who’s visiting, where they are in their journey, and what they need right now.

Let me show you how this works in practice.

What happened when a salesforce VP returned after a demo

Last week, a VP of Product from Salesforce visited a B2B website for the third time. She’d completed a product demo four days earlier, so this wasn’t a typical casual browse—her session time averaged over four minutes per visit, well above the site’s typical engagement.

Here’s what LiftPilot’s system detected the moment she arrived on the home page:

The identity layer immediately recognized her company: Salesforce and her role: VP of Product. More importantly, the system knew she was tied to an active opportunity worth $120,000 in ARR, with an assigned account executive.

Her previous behavior told a clear story: she’d completed a demo, viewed the pricing page twice, and spent significant time on implementation documentation.

The behavioral context painted an even clearer picture. This was her third visit to the site, and it had been four days since her demo call.

Identity signals:

• Company: Salesforce
• Role: VP of Product
• Deal stage: Active opportunity ($120K ARR)
• Previous actions: Demo completed, pricing page viewed twice
• Assigned AE: Marcus Chen

Behavioral context:

• Visit number: 3
• Time since demo: 4 days
• Last page viewed: Implementation FAQ
• Average session time: 4+ minutes (above average)

Sales intelligence:

• Gong note from demo: “Concerned about implementation timeline and team bandwidth”
• Email engagement: Opened implementation guide twice

But the most valuable signal came from the sales intelligence layer. A note from the Gong recording of her demo flagged a specific concern: she was worried about implementation timeline and whether her team had the bandwidth to roll out a new platform.

Most websites would show her the same homepage CTA everyone sees: “Request a Demo” or “Get Started.”

But she’s not in discovery mode. She’s in decision mode. She’s already demoed. She has specific concerns. She’s evaluating seriously.

What LiftPilot showed instead:

The CTA changed to:

“Continue your conversation with Marcus →”

Clicking took her directly to Marcus’s calendar, pre-populated with context: “Follow-up: Implementation timeline discussion”

The outcome:

She booked a call within 2 minutes of arriving on site. The call happened the next morning. Deal closed at Enterprise tier two weeks later.

This wasn’t a manually configured segment. It was an AI decision made in real-time based on multiple signals about who she was, where she was in the journey, and what she needed next.

How LiftPilot’s AI Actually Makes CTA Decisions

Traditional personalization might swap CTAs based on simple rules: “If enterprise company, show ‘Contact Sales’ instead of ‘Get Started.'”

That’s better than nothing. But it’s still one-dimensional.

LiftPilot’s AI considers multiple factors simultaneously and weights them based on what matters most for this specific visitor right now.

The Decision Framework

When a visitor arrives, the AI analyzes five key dimensions:

1. Journey Stage

This is the foundation stage. Where is this person in the actual buying process? An unknown visitor hitting the site for the first time is in a completely different place than someone who’s been there twelve times over the past two weeks.

• Unknown visitor (first touch)
• Active researcher (multiple visits, content engagement)
• Opportunity stage (demo completed, in CRM)
• Customer (existing relationship)
• Champion (high engagement, forwarding content)

The system distinguishes between a variety of different stages.

2. Intent Signals

The intent signals reveal what they’re trying to accomplish right now, not what they might want in general, such as:

• Casual research (blog entry, thought leadership)
• Active evaluation (pricing page, feature comparisons)
• Technical validation (documentation, integrations)
• Purchase readiness (ROI calculator, case studies)

These behaviors tell very different stories about what CTA would be most relevant.

3. Friction Points

Friction Points identify the obstacles that might stop them from moving forward:

• Implementation concerns (from sales notes)
• Pricing objections (from email responses)
• Security questions (page behavior)
• Buy-in challenges (multiple stakeholders visiting)

4. Relationship Context

Relationship Context captures the existing connection between this visitor and your company:

• New prospect (no prior engagement)
• Marketing qualified lead (form submitted)
• Sales qualified opportunity (demo scheduled)
• Existing customer (plan tier, usage level)

5. Behavioral Patterns

Behavioral Patterns round out the picture by showing how this person engages with your content:

• Visit frequency (returning daily vs. monthly)
• Content depth (skimming vs. deep reading)
• High-intent pages (careers page = competitor research)
• Time on site (1 minute vs. 8 minutes)

Each factor gets weighted dynamically. The AI doesn’t just check whether each condition is true or false and add up a score. It evaluates which signals matter most for this particular moment. A visitor’s journey stage might be the dominant factor in one case, while friction points from sales conversations might override everything else in another case.

How the our AI engine scores each option

Once LiftPilot identifies the relevant signals, it needs to decide which CTA to show. This is not a simple if-then rule. The system scores every available CTA option based on three factors and selects the highest-scoring match.

Factor 1: Relevance

How well does this CTA align with the visitor’s current context? The system evaluates whether the CTA’s tags (stage, intent, friction, relationship) match the signals detected.

For example, if the visitor is in “opportunity stage” with “integration concerns” flagged from their demo transcript, a CTA tagged for “decision stage” and “addresses implementation friction” scores higher on relevance than a generic “Learn More” CTA tagged for “discovery stage.”

Factor 2: Predicted outcome

What is the likelihood this CTA leads to the next meaningful step in the journey? The system maintains conversion probability data for each CTA based on historical performance in similar contexts.

For example, if the visitor shows strong implementation concerns and one CTA directly addresses implementation (“Review implementation plan →”) while another doesn’t (“Get Started →”), the relevant CTA scores higher on predicted outcome for this individual.

Factor 3: Confidence score

How certain is the model about the current context? Strong signals (CRM data showing demo completion, Gong transcript with clear objections, multiple behavioral signals aligned) produce high confidence. Weak or ambiguous signals produce lower confidence.

The confidence score helps weigh the decision—high confidence signals carry more weight in the final CTA selection.

The scoring calculation

Each CTA receives a composite score:

Score = (Relevance × Predicted Outcome) × Confidence

The CTA with the highest expected value wins. If two CTAs score similarly, the system may run a micro-experiment, showing different options to similar visitors and learning which performs better in that specific context.

This is why LiftPilot improves over time. Every CTA shown, every click or non-click, every conversion or drop-off feeds back into the predicted outcome model. The system continuously refines which CTAs work best in which contexts.

The CTA Library

LiftPilot doesn’t just swap between two or three options. It maintains a library of context-appropriate CTAs and selects from dozens of possibilities based on what’s most relevant.

This is how it looks:

Discovery Stage:

“See how it works →”
“Watch 2-minute overview →”
“Explore use cases →”

Evaluation Stage:

“See pricing & plans →”
“View customer stories →”
“Schedule product walkthrough →”

Decision Stage:

“Talk to [AE Name] →”
“See ROI calculator →”
“Review implementation plan →”

Customer Stage:

“View your account →”
“Explore [Next Tier] features →”
“Contact your CSM →”

High-Intent Signals:

“Get custom proposal →”
“Book technical deep-dive →”
“Start free trial →”

Each CTA in the library is tagged with the conditions that make it relevant: journey stage, intent level, relationship status, friction points it addresses.

The AI evaluates these tags against the visitor’s current context and selects the best match. As the system sees which CTAs drive conversions in which scenarios, it continuously refines those selection criteria.