AI-Powered Semantic Search Reference (Companies)

Reference — lookup information.

AI-Powered Semantic Search Reference

Last Updated: January 26, 2026 Release: Closed Beta (Labs) - January 2026 Object Tags: Search, AI, Semantic Search, Discovery, Global Search, Natural Language, Labs

Overview

Semantic Search uses AI to understand natural language queries, allowing you to search for companies based on business context, attributes, and characteristics - not just names or URLs. This enhancement to Affinity’s global search helps you discover companies using the same language you’d use when talking to a colleague. What it does: Transforms global search from a basic name lookup tool into an intelligent discovery engine that understands concepts like “healthcare companies my firm met with last week” or “companies backed by Sequoia Capital.” Current release: Company search only. People search coming April 2026. Who can access: Affinity customers enrolled in the Closed Beta via Labs How to enable: Settings > Labs > Turn on “Search with AI” Where to use: Global search bar (Cmd+Option+K or Alt+K) with “Search with AI” option

Key Concepts and How It Works

What is Semantic Search?

Traditional keyword search (current default):

Matches exact words you type
For companies: Only searches name and website URL
Requires you to know exactly what you’re looking for
Example: “Stripe” finds Stripe, but “payment processing company” finds nothing Semantic search (AI-powered):
Understands the meaning and context of your query
Searches across company attributes like industry, location, interactions, and relationships
Recognizes concepts, synonyms, and business context
Works with natural language descriptions
Example: “payment processing companies” finds Stripe, Square, and other relevant companies

How Semantic Search Works

When you enter a natural language query:

AI analyzes your query to understand what you’re asking for:

Attributes mentioned (industry, location, investors)
- Context clues (“my firm met with”, “in our pipeline”, “in our network”)
- Time references (“last week”, “recently”, “this quarter”)
- Business concepts (“healthcare”, “AI companies”, “solar energy”)

Searches across available company data including:

Affinity Data: Industry, description, location, investor information
- Smart Attributes (firm-wide): First/Last Contact, First/Last Meeting, First/Last Email, Next Meeting
- Relationship Intelligence (firm-wide): Interaction history (meetings, emails), relationship scores
- CRM Context: Org-relevant status (companies in your firm’s CRM)

Ranks results by relevance:

Prioritizes org-relevant companies (companies your firm knows) over universal dataset
- Considers how well each company matches your query
- Shows most relevant results first

Returns results typically within 10 seconds with:

Matching companies
- Explanation of how results were found
- Ability to refine your query if results don’t match expectations IMPORTANT - Firm-Wide Data Only: When you use “I” in queries (e.g., “companies I met with last week”), the AI interprets this as your FIRM’s interactions, not your personal interaction history. Semantic search uses firm-wide data only:
Interaction dates reflect ANY team member’s interactions
Relationship scores are firm-wide maximum
“I” and “we” in queries mean “my firm”
You cannot search for “companies only I personally contacted”

What Data is Searchable

For Companies (January 2026): ✅ Affinity Data:

Name and domain
Description
Industry
Headquarters location
Investor information (investors who have funded the company) ✅ Smart Attributes (Firm-Wide):
First/Last Contact with your firm
First/Last Meeting with your firm
First/Last Email with your firm
Next Meeting scheduled with your firm ✅ Relationship Intelligence (Firm-Wide):
Interaction history (meetings and emails between company and your firm)
Relationship scores (firm-wide max relationship strength) ✅ CRM Context:
Org-relevant status (companies your firm has interacted with or added to Affinity)
Note: “In our pipeline” searches org-relevant companies, NOT specific list membership

When to Use Semantic Search vs. Keyword Search

Use Semantic Search When:

You know characteristics but not the name:

“Solar energy companies in our pipeline”
“Healthcare companies in Boston”
“Companies in the payments space” You’re searching by firm interactions:
“Healthcare company my firm met with last week”
“Companies we haven’t contacted in 90 days”
“Companies our firm had meetings with this quarter” You’re searching by investors:
“Companies backed by Sequoia Capital”
“Companies backed by both Andreessen Horowitz and Sequoia” You’re searching by relationship strength:
“Early stage companies with low relationship scores”
“Companies where we have strong connections”
“Companies with no recent contact” You’re exploring or discovering:
“AI companies in San Francisco”
“Solar energy companies that use battery storage technology”
“Healthcare companies focused on cancer diagnosis” You don’t remember exact details:
“That biotech company someone at our firm talked to at the conference”
“The solar company mentioned in our partner meeting”

Use Keyword Search When:

You know the exact name:

“Stripe” (instant results)
“OpenAI” (direct match) You have domain:
“stripe.com”
“openai.com” You want instant results:
Keyword search is faster (<1 second vs ~10 seconds)
Better for quick lookups when you know what you’re searching for

Example Queries

Supported Company Search Examples (January 2026)

By Industry and Location:

“Healthcare companies in our network”
“Fintech startups in San Francisco”
“Climate tech companies in Boston”
“Enterprise SaaS companies in New York”
“AI companies focused on healthcare” By Investors:
“Companies backed by Sequoia Capital”
“Companies backed by both Bain Capital and Bessemer”
“Portfolio companies of Andreessen Horowitz” By Firm Interactions:
“Solar energy company my firm met with last week”
“Companies we haven’t contacted in 90 days”
“Healthcare companies our firm had meetings with this quarter”
“Companies we emailed recently” By Relationship Strength (Firm-Wide):
“Early stage companies with low relationship scores”
“Companies where we have strong connections”
“Companies with weak relationships” By Industry Concepts and Technology:
“Payment processing companies”
“AI companies focused on enterprise automation”
“Solar energy companies using battery storage technology”
“Healthcare companies focused on cancer diagnosis using AI”
“Biotech companies working on extracellular vesicles”

Understanding Search Results

Results Display

What you’ll see:

Ranked list of matching companies
Most relevant results appear first
Standard company preview information:
Name, description, location, industry
- Relationship indicators
Companies from your firm’s CRM (org-relevant) prioritized over universal dataset Result explanation:
AI-generated explanation of how results were found
Shows which attributes/criteria were used to match
Helps you understand if query was interpreted correctly
Use this to refine your query if results don’t match expectations

Refining Your Queries

If results don’t match what you wanted:

Edit and re-run your query:

Add more specificity: “healthcare” → “digital health companies”
- Add constraints: “AI companies” → “AI companies in San Francisco”
- Add time context: “companies” → “companies our firm met with last month”
- Add industry detail: “healthcare” → “healthcare companies focused on cancer diagnosis”

Check the explanation:

See how AI interpreted your query
- Identify misunderstandings
- Adjust wording based on how it was parsed

Try different phrasings:

“Payment processing companies” vs “Fintech companies in payments”
- “Last week” vs “In the past 7 days”
- “Solar energy” vs “Renewable energy companies”

Be more specific:

Add location: “Healthcare companies in Boston”
- Add investor: “Companies backed by Sequoia”
- Add technology detail: “AI companies building customer service automation tools”
- Add sub-sector: “Biotech companies focused on gene therapy”

Key Features and Capabilities

Natural Language Understanding

AI recognizes:

Industry terms: “Healthcare”, “fintech”, “climate tech”, “enterprise SaaS”, “biotech”, “AI”
Technology descriptors: “Solar energy”, “battery storage”, “cancer diagnosis”, “payment processing”, “gene therapy”
Time references: “Last week”, “recently”, “this quarter”, “in the past 6 months”, “90 days”
Location: Cities, states, countries, regions
Relationship context: “In our network”, “in our pipeline” (org-relevant), “our firm met with”
Investors: Recognizes investor names and can search by backing firms

Context Awareness

Understands business context:

“In our pipeline” = searches org-relevant companies (companies your firm knows)
“My firm met with” = searches firm-wide interaction history
“In our network” = searches org-relevant entities with connections
“We haven’t contacted” = searches by recency of firm’s last contact date Important clarifications:
“I met with” is interpreted as “my firm met with” (firm-wide interactions)
“We” always refers to your firm, not a specific team or individual
Cannot search for interactions with specific individuals at your firm Combines multiple attributes:
“Healthcare companies in Boston our firm met with recently” = industry + location + firm interactions
“AI companies backed by Sequoia” = industry + investor
“Solar energy companies with low relationship scores” = industry + firm-wide relationship strength

Smart Prioritization

Results prioritize:

Org-relevant companies (in your firm’s CRM) over universal dataset
Recency of firm interactions when time-based queries
Firm-wide relationship strength when relevant to query
Exact matches over fuzzy matches
Multiple criteria matches over single attribute matches

Use Cases and Examples

Quick Company Lookup (Can’t Remember Name)

Scenario: Someone at your firm met with a company last week but you can’t remember their exact name, only that they were in healthcare. Query: “Healthcare company our firm met with last week” What AI searches:

Industry = Healthcare (or related terms)
Your firm’s meeting history in past 7 days
Org-relevant companies Result: Companies matching all criteria, ranked by relevance

Note: Shows companies ANY team member met with, not just your personal meetings.

Contextual Discovery (Finding by Attributes)

Scenario: You’re sourcing fintech companies in New York. Query: “Fintech companies in New York” What AI searches:

Industry = Fintech (financial services, payments, banking tech, etc.)
Location = New York (city, state, metro area) Result: Companies matching criteria from your firm’s CRM and universal dataset

Investor Intelligence (Companies by Backers)

Scenario: You want to find companies backed by Sequoia to understand their portfolio overlap. Query: “Companies backed by Sequoia Capital” What AI searches:

Investor fields in Affinity Data
Looks for “Sequoia Capital” and variations (Sequoia, Sequoia VC, etc.) Result: Companies with Sequoia as an investor

Note: Uses Affinity investor data only.

Pipeline Management (Finding Companies Needing Attention)

Scenario: Quarterly review coming up, need to find companies that haven’t been contacted recently. Query: “Companies in our pipeline we haven’t contacted in 90 days” What AI searches:

Org-relevant companies (companies your firm knows)
Last contact date from your firm (>90 days ago)
Your firm’s interaction history Result: Org-relevant companies needing outreach

Note: “In our pipeline” means org-relevant companies (your firm knows them), not membership in a specific list named “Pipeline.”

Industry-Specific Discovery

Scenario: Looking for companies in a specific technology area. Query: “Solar energy companies that use battery storage technology” What AI searches:

Industry = Solar energy (clean energy, renewable energy, etc.)
Description contains concepts related to battery storage Result: Solar companies with battery technology focus

How to Access and Use

Enabling Search with AI

Navigate to Settings > Labs
Find “Search with AI” feature
Toggle ON
Return to global search

Using Search with AI

Open global search (Cmd+Option+K or Alt+K)
Click “Search with AI” button
Type your natural language query
Press Enter
Review results (appears in ~10 seconds)
Click any company to view profile
Edit query and re-run if needed

Tips

Example queries are shown in the search interface
Click an example to pre-fill your search bar
Query stays visible for easy editing and refinement

Best Practices

Writing Effective Queries

Be specific:

❌ “Companies” (too broad)
✅ “Healthcare companies in Boston” (specific attributes) Use natural language:
✅ “Show me AI companies our firm met with recently”
✅ “Find payment processing companies”
✅ “Which healthcare companies are in our network?” Combine multiple criteria:
“Healthcare companies in Boston our firm met with this quarter”
“AI companies backed by Sequoia that we haven’t contacted” Use time references naturally:
“Last week”, “recently”, “in the past 6 months”
“This quarter”, “this year”
“Haven’t contacted in 90 days” Leverage relationship context (firm-wide):
“In our network” (org-relevant companies)
“In our pipeline” (org-relevant companies)
“Our firm met with”, “We haven’t contacted” Add industry detail for better results:
Instead of just “healthcare”, try “healthcare companies focused on cancer diagnosis”
Instead of “AI”, try “AI companies building enterprise automation tools”
Instead of “energy”, try “solar energy companies using battery storage” Remember “I” = “my firm”:
“Companies I met with” searches your FIRM’s meetings, not just yours personally
“I haven’t contacted” searches your FIRM’s contact history

Interpreting Results

Review result explanations:

Understand which attributes matched
See if AI interpreted your query correctly
Use insights to refine future queries Check result relevance:
Are top results actually what you wanted?
Are org-relevant results appearing first?
Do you need to add more constraints or detail? Iterate if needed:
Results too broad → Add more specific criteria or technology detail
Results too narrow → Remove some constraints
Wrong matches → Rephrase using different terminology
No results → Try broader terms or check if data exists in Affinity

When to Switch to Lists/Filters

If you find yourself:

Running the same semantic search repeatedly → Save as filtered list view
Needing exact field-level control → Use Lists filters
Working with results (editing, exporting) → Move to Lists
Sharing query with team → Create saved view in Lists
Searching within a specific list → Use Find in View or Lists filters Workflow: Semantic search for discovery → Create filtered list view for ongoing use

Technical Notes

Performance:

Results typically return within 10 seconds
May take longer for very complex queries
Keyword search remains faster (<1 second) for simple name lookups Data freshness:
Newly created companies searchable within ~10 seconds
Updates to company data available in search within ~10 seconds
Interaction data (meetings, emails) syncs continuously Privacy and permissions:
Semantic search respects same permissions as keyword search
You only see companies you have permission to view
Interaction history is firm-wide (not individual-specific)
Relationship scores are firm-wide max relationship strength Result limits:
Returns most relevant results (typically 10-20 companies)
Can refine query for different results
For comprehensive analysis, use Lists filters instead

Frequently Asked Questions

How is this different from regular search? Regular keyword search only matches exact company names and URLs. Semantic search understands what you’re describing and searches across industry, location, investors, and firm interaction data. Do I need to use special syntax? No. Write queries like you’re asking a colleague: “Show me healthcare companies in Boston” or “Find companies backed by Sequoia.” The AI understands natural language. Does this replace keyword search? No. Keyword search remains the default and is still best for quick name lookups. Semantic search is an optional enhancement for contextual discovery. How long do results take? Semantic search typically returns results within 10 seconds (vs <1 second for keyword search). The AI needs time to understand your query and search across data. Does semantic search use my personal interaction history or firm-wide data? Semantic search uses FIRM-WIDE interaction data only. When you search for “companies I met with last week,” the AI searches for companies YOUR FIRM met with, not just your personal meetings. This means:

Interaction dates reflect any team member’s interactions
Relationship scores are firm-wide maximum
“I” and “we” in queries are interpreted as “my firm”
You cannot search for “companies only I personally contacted” How do I know which mode I’m using? You’ll see “Search with AI” in the global search interface. You must enable it in Settings > Labs first. Keyword search remains the default.

How are search results ordered, and how can I prioritize companies my firm knows? Semantic search ranks results using multiple factors: Ranking factors:

Semantic relevance score (primary): How well the company matches your query across all attributes
Org-relevance (secondary): Companies your firm knows are prioritized but not guaranteed first
Data completeness (tertiary): Companies with richer descriptions and data rank higher This means a non-org-relevant company with a perfect semantic match may appear before an org-relevant company with a weaker match.

To limit results to only companies your firm knows: Add these phrases to your query:

“in my network”
“in our pipeline”
“companies we know” Examples:
“healthcare companies in my network”
“AI companies in San Francisco in our pipeline” Without these phrases, search returns both org-relevant and non-org-relevant companies from the universal database, ranked by semantic relevance.

What if I get no results or irrelevant results? Try:

Rephrasing your query with different terms
Being more specific (add industry detail, location, investor)
Being less specific (remove constraints)
Checking the result explanation to see how AI interpreted your query
Switching to keyword search if you know the exact name
Rating the results using feedback options to help us improve What does “in our pipeline” actually search? “In our pipeline” searches org-relevant companies (companies your firm has interacted with or added to Affinity). It does NOT search for companies in a specific list named “Pipeline.” To search within a specific list, use Lists filters or Find in View instead.

Is this available to everyone? Currently available as a Labs feature (Closed Beta) to select customers starting January 2026. Must be enabled in Settings > Labs. Broader availability to be announced. How accurate are the results? Accuracy depends on data quality in your CRM. The AI uses available Affinity Data and Smart Attributes to find matches. Companies with more complete descriptions and accurate industry tags are more discoverable.

This section is for internal audiences only

Future Capabilities and Current Limitations

This section describes features not yet available in the January 2026 release and planned future enhancements.

What’s Coming

April 2026 (End of Q1)

People Search:

Search for people by role, company characteristics, and firm interactions
Example queries:
“CTOs at Series A healthcare companies”
- “Founders in our network”
- “People our firm met with last week”
- “CFOs at fintech companies in New York” Data for People Search will include:
Current job title
Current organization
Organization associations (all companies person has been associated with)
Smart Attributes (firm-wide): First/Last meeting, contact, email
Firm-wide relationship scores People Search limitations (April 2026):
Only org-relevant people (people your firm has emailed)
Current role only (not full work history)
Firm-wide relationship data only (cannot search by specific internal team member relationships)
Cannot search non-org-relevant people (universal database)

February/March 2026

Smart Attributes from Chat Messages:

Include chat interaction metadata in searches
Search by recent chat activity

What’s NOT Supported (No Current Plans)

Third-Party Enrichment Data:

❌ Crunchbase, PitchBook, Dealroom data
❌ Cannot search by: Funding amounts, funding stages, employee counts, growth metrics, valuation
Workaround: Use Lists filters for funding/stage-based queries with enrichment data Notes and Files:
❌ Notes content (title, body, metadata)
❌ Uploaded files
❌ Email file attachments
❌ Email body content
Workaround: Use keyword search in Notes tab or search within Lists Custom Fields and Lists:
❌ Custom fields you’ve created
❌ Specific list membership (“companies in my Series A list”)
❌ List names
Workaround: Use Lists filters or Find in View Opportunities:
❌ Opportunity-specific searches
Workaround: Use Opportunity Lists filters Questions About Specific Companies:
❌ “Who is our strongest connection at Anthropic?”
❌ “Who are the C-suite at OpenAI?”
❌ “What is [company]‘s funding amount?”
Why not supported: These are better answered by visiting the company profile directly for faster, deterministic information
Workaround: Use keyword search to find the company, then view their profile Individual-Specific Relationship Queries:
❌ “Companies I personally contacted” (vs firm-wide)
❌ “Companies where [team member] has the strongest relationship”
❌ “Who internally has the best connection at [company]?”
Explanation: Only firm-wide relationship and interaction data is included
Workaround: Use company Connections tab for individual relationship details

Example Queries NOT Supported

Funding/Financial (No Enrichment Data):

❌ “Companies that raised more than $10M”
❌ “Series A companies”
❌ “Companies with highest employee growth”
❌ “Companies with valuation over $100M”
Use instead: Lists filters with enrichment fields People (Coming April 2026):
❌ “CTOs at healthcare companies”
❌ “Founders in my network”
❌ “People our firm met with last week” Notes/Content-Based (No Current Plans):
❌ “Companies mentioned in my notes about AI”
❌ “Companies where we discussed Series B in emails” List-Specific (No Current Plans):
❌ “Companies in my Series A pipeline list”
❌ “Companies with status = Active in my deals list”
Use instead: Lists filters or Find in View

Comparison: Semantic Search vs Lists/Filters

Semantic Search is Better For:

Quick discovery without setup:

No need to create filters or views
Ask questions in natural language
Get results in ~10 seconds
Explore your data ad-hoc Contextual exploration:
“What healthcare companies are in our network?”
“Show me companies our firm met with recently”
Testing hypotheses about your pipeline When you don’t know exact criteria:
Industry concepts vs exact field values
Natural language vs structured filters

Lists/Filters are Better For:

Precise field-level queries:

Funding amount > $10M (uses enrichment)
Status = Active AND Stage = Series A
Custom field values Persistent views you’ll reuse:
Create saved views with specific filter combinations
Share views with team members
Reference the same filtered view repeatedly List-specific searches:
Companies within a specific list
List entry metadata (date added, owner) Bulk actions:
Edit multiple entries
Export filtered datasets
Apply batch updates Performance at scale:
Filtering large lists is faster than repeated semantic searches
Better for working with same dataset multiple times

Known Issues and Feedback

If you encounter issues:

Use the in-app feedback button to report problems
Rate search results (helps AI improve)
Email your feedback to product-feedback@affinity.co Common issues being tracked:
Results don’t match expectations → Rate results and describe what you expected
Query misinterpreted → Share the query and how AI should have interpreted it
Missing data types → Let us know what data would make search more useful Your feedback directly influences:
Which data types we prioritize adding
How we improve query interpretation
Which features become permanent vs experimental

Search

Documentation Index