Your BDR has a call in 15 minutes. The only thing in the CRM is a name: "Sarah Chen — VP Engineering." No LinkedIn. No context. No idea what she cares about or how she likes to communicate. Sound familiar?

We built a pipeline inside Alpha Agent that takes a bare contact name and returns a comprehensive dossier — verified social profiles, communication style analysis, personality insights, conversation starters, and full company context — in about 60 seconds. We've enriched 46 contacts across 5+ client accounts so far, with 21 of those receiving deep personality analysis. Here's how it works, and why it's changing the way we approach sales prep and account management.

The Problem: CRM Graveyards

Every professional services firm has the same disease. Contact records start as a name and an email. Maybe a job title if someone was feeling diligent. Over months, those sparse records multiply into a CRM full of ghosts — hundreds of contacts where you know who but have no idea what they care about, how they like to communicate, or what connects them to your other relationships.

The consequences are real:

• Cold outreach stays cold. Generic "Hope you're doing well" emails with no hook. No reference to their recent conference talk, their open-source project, or the fact that they just changed companies.
• Meeting prep is a scramble. Before every call, someone spends 15-20 minutes frantically Googling, checking LinkedIn, and trying to piece together who this person is and what they've been working on.
• Relationship context lives in people's heads. When your account manager goes on PTO, nobody knows that the VP of Engineering prefers Slack DMs over email, or that the CTO is passionate about accessibility and that's a great rapport builder.
• Renewal conversations start flat. You're supposed to demonstrate deep understanding of the client's world, but your CRM tells you nothing about the humans on the other side.

According to Salesforce's 2025 State of Sales report, sales reps spend only 28% of their week actually selling — the rest goes to admin, research, and data entry.^[1] A significant chunk of that research time is manually enriching contact records that should have been populated months ago.

The Pipeline: Three Skills Working in Concert

Alpha Agent's CRM enrichment isn't a single feature — it's three specialized AI skills that chain together, each with its own data sources, verification logic, and output format. Here's the architecture:

Skill 1: People Research

The people-research skill is the foundation. Given a name (and optionally a company, email domain, or Slack profile), it systematically searches and verifies across platforms:

• LinkedIn (highest priority) — verified against 2+ matching signals: name + company + role + location + profile photo comparison
• GitHub — cross-referenced against known handles, commit emails, org membership
• Twitter/X — bio must mention company or matching role
• Other platforms (YouTube, Instagram, etc.) — only searched when context suggests relevance

The key principle: never guess. Every social profile gets a confidence score based on weighted verification signals:

Medium-confidence matches go to a verification queue where a human can approve or reject with side-by-side photo comparison. Rejected matches are permanently stored — the system never re-suggests them. This prevents the annoying false-positive loop that plagues most enrichment tools.

Skill 2: Company Research

In parallel, the company-research skill builds rich profiles of each contact's employer. For our client accounts — including Lively, Diligent, WOW! (WideOpenWest), Integral Ad Science (IAS), and Impossible Foods — we pull:

• Industry, size, funding, and competitive landscape
• Tech stack (critical for a technology consultancy — knowing they use Contentful or Next.js immediately signals fit)
• Key people and organizational structure
• Social presence and content activity
• Relationship history from our own Slack channels, GitHub repos, and meeting notes

This means when you pull up a contact, you don't just see who they are — you see the full context of where they work, what technology decisions they're making, and how their company is positioned in the market.

Skill 3: Personality Insights

This is where it gets interesting. The personality-insights skill analyzes every data source we have for a person — Slack messages, Zoom meeting transcripts, social media posts, Jira activity — and synthesizes a communication profile that tells you how to talk to them.

The output is structured across four dimensions:

Communication Style Profile

Personality Signals

Actionable Conversation Starters

Not generic icebreakers — specific, researched hooks:

• "They recently posted about distributed systems on LinkedIn — ask about their migration."
• "Shared interest in TypeScript performance with Adam — mention the recent Deno benchmarks."
• "They mentioned frustration with their current CMS in last week's standup — good opening for the Contentful conversation."

Approach Notes

The system generates a plain-language guide: "Keep messages concise, lead with action items. They respond faster to Slack than email. Prefer morning PT. When they go quiet, they may be frustrated — check in proactively."

Data Sources: Deeper Than You'd Expect

What makes this pipeline powerful isn't any single data source — it's the combination. Here's what feeds into the analysis:

Zoom transcripts are the secret weapon. Extended natural conversation reveals far more about someone's personality and communication style than short Slack messages. Through our Zoom server-to-server OAuth integration, Alpha Agent can pull and analyze meeting transcripts automatically — no manual upload required.

Real Results: 46 Contacts, 21 Deep Profiles

Here's where we stand after running this pipeline across our CRM:

• 46 contacts in the Supabase users table with enriched profiles — names, companies, titles, handles, and social links verified
• 21 contacts with full personality insights in the research_insights table — communication style, interests, personality signals, conversation starters, and approach notes
• All 21 deep profiles rated "high confidence" — meaning 50+ Slack messages analyzed, social data verified, and multiple data sources corroborated
• 5+ client accounts with company research — Lively, Diligent, WOW!, IAS, Impossible Foods — each with tech stack, key people, relationship history, and competitive context

The pipeline runs incrementally. It tracks lastResearchedAt, lastSlackPulledAt, and lastZoomPulledAt timestamps per contact. On subsequent runs, it only pulls new data since the last analysis, merges it with existing insights, and updates the profile. This means the first enrichment takes ~60 seconds per contact, and refreshes take ~15 seconds.

What This Changes: Before and After

Before: The 15-Minute Scramble

1. Check CRM — see name and email. Nothing else.
2. Open LinkedIn in new tab. Search. Find 12 people with that name. Try to figure out which one.
3. Open their company's website. Skim the About page. Try to remember what they do.
4. Search Slack for recent messages. Find a few, but no pattern emerges in 3 minutes of scrolling.
5. Give up and wing it on the call.

After: The 10-Second Lookup

1. Open CRM contact card.
2. See verified LinkedIn, GitHub, and social profiles. Read the 2-paragraph personality summary. Note the conversation starters. Check the approach notes.
3. Walk into the call knowing they prefer concise communication, care deeply about developer experience, recently posted about their company's migration to a headless CMS (your specialty), and that they respond fastest on Slack in the mornings.

The difference isn't just time saved — it's confidence. You're not faking familiarity; you actually understand who you're talking to.

For BDRs: Cold Outreach That Isn't Cold

The enrichment pipeline is especially powerful for business development. When you're reaching out to a prospect for the first time, the personality insights give you an unfair advantage:

• Personalized opening lines based on their actual interests and recent activity, not templated "I noticed you work at {Company}" filler
• Channel selection — some people respond to email, others to LinkedIn InMail, others to Twitter DMs. The insights tell you which.
• Tone matching — if they're casual and emoji-heavy, your stiff formal email will feel off. If they're detail-oriented, your two-sentence pitch will feel lazy.
• Timing optimization — the system knows their timezone and when they're most responsive

Research from Gong.io indicates that personalized outreach generates 2.5x higher response rates than generic messaging.^[2] When your personalization is informed by actual communication pattern analysis rather than just "I see you're based in San Francisco," the gap widens further.

For Account Managers: Institutional Memory

This is where the long-term value compounds. The personality insights become institutional memory that survives personnel changes:

• New account manager onboarding — instead of weeks of "getting to know" each contact, they can read the profiles and be conversationally competent on day one
• Relationship continuity — when your main contact at a client leaves and you need to build rapport with their replacement, the enrichment pipeline runs on the new person immediately
• Proactive account health — the warning signals in the personality profile ("when they go quiet, check in proactively") prevent small frustrations from becoming churn events
• Renewal prep — pull up every stakeholder's profile before the renewal conversation, know exactly how to frame the value conversation for each person's priorities and communication style

The Ethics Guardrails

We built this with clear ethical boundaries:

• Only data we already have access to. Slack messages in shared channels, public social posts, meeting transcripts from calls we were on. No scraping private profiles.
• Internal use only. Personality insights are never shared externally. They're tools for building better working relationships, not dossiers to be used manipulatively.
• The discomfort test. If the person being analyzed would be uncomfortable knowing we'd done this analysis, we reconsider. The goal is to be a better partner, not a surveillance apparatus.
• Verification over assumption. The scoring system's refusal to guess means we'd rather have gaps than inaccurate data. A CRM with 5 verified fields beats one with 20 hallucinated ones.

Technical Architecture

For teams considering a similar approach, here's the stack:

• Data storage: Supabase (PostgreSQL) — users table for contact profiles, research_insights for personality analysis, zoom_transcripts for meeting data, clients for company profiles
• Orchestration: Alpha Agent AI agent with three specialized skills — people-research, personality-insights, company-research
• Search & verification: Brave Search API for web research, GitHub CLI for developer profiles, Slack API for message analysis
• Meeting transcripts: Zoom server-to-server OAuth for automatic transcript pull and processing
• Incremental processing: Timestamp tracking per contact per data source to avoid redundant work
• Human-in-the-loop: Verification queue for medium-confidence matches, rejection memory to prevent re-suggestion

The entire pipeline runs as conversation-driven AI — tell Alpha Agent "research Sarah Chen" and it chains the three skills automatically, stores everything in Supabase, and returns a formatted summary. No manual data entry. No switching between tabs and copy-pasting URLs.

What's Next: Relationship Intelligence

We're extending this in three directions:

1. Relationship mapping — automatically identifying connections between contacts ("Jane from Acme used to work at Lively with your primary contact there") by analyzing company history overlaps
2. Refresh cadence — key contacts get monthly insight refreshes, others quarterly. Before important meetings, a targeted refresh pulls the latest Slack messages and social activity.
3. Meeting briefings — before every calendar event, Alpha Agent generates a one-page briefing with every attendee's profile, conversation starters, and approach notes. The 15-minute scramble becomes a 30-second skim.

The Bottom Line

CRM enrichment tools aren't new. Apollo, ZoomInfo, Clearbit — they all sell contact data. But they give you facts: job title, company size, email address. Useful, but not differentiated.

What they don't give you is understanding: how this person communicates, what motivates them, what frustrates them, what they care about right now, and how to build genuine rapport. That's the gap this pipeline fills.

For a 15-person professional services firm like Last Rev, where every client relationship is high-touch and every sales conversation matters, the difference between "I know their name" and "I understand how they think" is the difference between winning and losing work.

Sixty seconds per contact. Twenty-one deep personality profiles and counting. Zero guesswork.

Want to build your own AI-powered CRM enrichment pipeline? Let's talk.

Footnotes & Sources

1. Salesforce, State of Sales, 6th Edition (2025). Sales reps spend 72% of their time on non-selling activities. salesforce.com/resources/research-reports/state-of-sales
2. Gong.io, The State of Personalization in Sales Outreach (2025). Personalized emails see 2-3x higher response rates vs. generic templates. gong.io/resources
3. Contact enrichment data from Last Rev's internal Supabase CRM — users table (46 records), research_insights table (21 records), clients table (5+ accounts). Data as of February 2026.
4. Alpha Agent skill definitions: people-research v1.0.0, personality-insights v1.0.0, company-research v1.0.0.