How Contact AI Insights Work

This article explains how the AI-generated scores and insights on a contact record are calculated.

Each section covers what signals feed the insight and how the output is produced.

Fit Score

What it measures: How well a contact matches the ideal customer profile.

How it's filled:

Until an ICP is configured, Fit Score is calculated using default parameters:

• Industry — SaaS, Healthcare, Finance, Manufacturing
• Company size — 500+ employees
• Revenue — greater than $1M
• Geography — same country or region as the user
• Job title — C-level, VP, Director

The first time a contact is scored, a tooltip prompts teams to customize their ICP in Settings for more accurate scoring.

Once configured, teams can define their own attributes (any field on the Contact object), apply operators and rules, and assign High, Mid, or Low priority to each attribute. Priority determines how much each field impacts the overall score.

If a field is missing on a contact, the score is normalized based on the available fields.

Engagement Score

What it measures: How well the contact engages relative to the team's outreach effort — not just activity volume.

How it's filled:

The score combines five components:

• Base Score — points from each interaction (email reply, meeting attended, call connected, page visit, trial signup)
• Response Ratio — responses divided by outreach attempts, mapped to a multiplier (strong, moderate, weak)
• Sentiment Weight — adjustment based on whether replies and conversations are positive, neutral, or negative
• Recency Multiplier — recent interactions count more; older interactions decay over time
• Overrides — hardcoded boosts or penalties for strong signals like a positive reply to a first email (+50), an explicit buying statement (+70), an unsubscribe (−60), or an email bounce (−50)

The final score is capped between 0 and 100.

Two edge cases are handled explicitly:

• No engagement initiated → score is 0
• Outreach sent but no replies → score reflects outreach effort rather than defaulting to 0

Response Rate

What it measures: A quantitative view of interactions across channels — distinct from Engagement Score, which factors in sentiment and recency.

How it's filled:

Each channel is normalized to a 0–100 scale:

• Email Response Rate = (Replies ÷ Emails Sent) × 100
• Call Response Rate = (Answered ÷ Calls Attempted) × 100
• Meeting Response Rate = (Attended ÷ Meetings Invited) × 100

The three rates are equally weighted (33% each) to produce the final Response Rate.

If a channel has no data, its weight is redistributed proportionally across the remaining channels. If all three are missing, the score displays as NA.

Buying Intent

What it measures: How strongly the contact is showing intent to evaluate, buy, or move forward.

How it's filled:

Every new email, call, meeting, or deal activity is analyzed and classified as a High, Mid, or Low intent signal. The insight displays the total count of signals detected, with each signal tagged by intent level.

High-intent signals include positive replies with forward momentum, commitments to meetings or reviews, meaningful call duration with clear next steps, new stakeholders added to meetings, and fast stage progression (each stage under 7 days).

Medium-intent signals include non-committal replies, exploratory questions, listening-heavy calls without next steps, and accepted meetings with no follow-up.

Low-intent signals include no replies after multiple emails, explicit deprioritization ("not now," "no budget"), missed or declined calls, no-shows, repeated rescheduling, and stages stuck beyond 21 days.

Risk Factors

What it measures: Specific risks that could affect the relationship or deal.

How it's filled:

Signals from emails, calls, meetings, and contact attributes are continuously monitored. When a signal matches a risk pattern, the risk factor is surfaced with a count and an expandable detail view.

Detected risks may include:

• Not a Decision Maker — role or seniority is low or mid-level
• Single-Threaded Risk — only one contact engaged from the company
• Low Engagement — infrequent replies, skipped meetings, long delays
• Negative Sentiment — skepticism or dismissive tone in communications
• Objections Raised — repeated concerns about pricing, ROI, or features
• Budget Concern — mentions of budget freezes or financial pushback
• Discounts Discussed — mentions of pricing flexibility or percentage off
• Champion Risk — role change or job transition detected
• Procurement or Legal Delays — dependencies flagged but not yet engaged
• Declining Responsiveness — initially engaged, now replying less often
• Misfit Role Engagement — role irrelevant to buying process
• Unclear Next Steps — no committed follow-ups or timelines

Buyer Profile

What it measures: The contact's role in the buying process.

How it's filled:

The AI classifies each contact into one of eight profiles based on job title, seniority, interaction patterns across meetings, and deal-stage behavior.

• Decision Maker — senior title (VP, Director, CXO); joins late-stage meetings; others defer decisions to them
• Influencer — asks detailed feature or workflow questions; shapes opinion without final authority
• Champion — advocates internally, introduces new stakeholders, drives momentum
• Economic Buyer — discusses pricing, contracts, ROI; engages in negotiation
• Technical Evaluator — asks about integrations, security, architecture; joins technical calls
• Blocker or Skeptic — raises repeated objections, delays decisions, questions value
• End User — focuses on usability and daily workflows; joins demos and trials
• Observer — CC'd on emails, silent in meetings, no follow-up actions

Competitor Mentions

What it measures: References to competitors detected in communications.

How it's filled:

The AI scans emails, call transcripts, and meeting transcripts for competitor names. Each mention is logged with:

• Which competitor was mentioned
• When it was mentioned
• The source (email, call, or meeting)
• The relevant excerpt or quote

The insight displays a total mention count, expandable to see each instance.

Best Contact Time

What it measures: The optimal day and time range to reach the contact.

How it's filled:

The recommendation combines several signals:

• Historical engagement patterns — when the contact typically opens emails, answers calls, attends meetings, or replies
• Time zone and work schedule — adjusted to the contact's local time and typical work hours
• Channel preferences — the channel that has historically produced the fastest response
• Recency and frequency — recent engagement patterns are weighted more heavily
• External context — public holidays, end-of-quarter periods, and industry-specific patterns

The AI combines these signals into a probability score for successful contact across different time windows, and surfaces the highest-ranked window as the recommendation.

When composing an email, the recommendation appears with a Schedule Now CTA. If a scheduled send falls on a regional holiday, an alert suggests a better alternative date and time.