ZoomInfo AI: API, Agents and Governed Automations
Understand ZoomInfo's native AI, eight verified API endpoints, and how Major adds stateful apps, agent judgment, and governed workflows for sales operations.

ZoomInfo AI combines B2B company and contact intelligence with features such as Copilot account summaries and GTM Studio's AI-assisted plays. This guide covers those native capabilities, the ZoomInfo API, and how RevOps teams use Major to turn intelligence into governed work across sales systems.
Major agents build deterministic apps that retain enrichment history, approvals, and handoff status. The agent handles judgment; the app holds state; the workflow controls execution.
What ZoomInfo does for revenue teams
ZoomInfo helps sales, marketing, and revenue operations teams discover companies, identify relevant contacts, enrich records, and interpret buying signals. Its API overview separates discovery, enrichment, business context, audiences, and agent-related capabilities, which matters when designing an integration.
An SDR manager needs contacts at accounts that match an approved market, not simply a large list. An account executive needs the right stakeholders and recent account context before a meeting. A RevOps owner needs consistent identifiers, current company attributes, controlled data consumption, and a way to reconcile enrichment against the CRM.
The Companies and Contacts documentation describes company and person data as separate resources. Preserve that separation in your app: one company can have several contacts, and a contact changing employers should not silently rewrite an account relationship.
On Major, the useful starting point is an operational queue rather than another chat window. Store the incoming record, the matched ZoomInfo identity, proposed changes, and final disposition. That gives both the rep and the agent a durable record of what already happened.
ZoomInfo AI: native features and where Major fits
ZoomInfo already provides AI within its revenue products. GTM Studio supports natural-language audience exploration, AI scoring, message creation, and signal-triggered plays. Its product page describes specialized agents for enrichment, scoring, routing, and messaging, plus connections to CRM, engagement, and marketing systems.
Copilot adds account intelligence for sellers. The account summary reference exposes generated summaries for companies configured as Target Accounts, and explicitly requires a Copilot entitlement. ZoomInfo's Signals and Insights guide distinguishes raw events from curated insights that combine signals with account context, ownership, CRM data, and recency.
These capabilities are valuable inside ZoomInfo. Major adds an enterprise-specific execution layer when the next action depends on policies and state across systems: an existing Outreach enrollment, a protected Salesforce field, an unresolved approval, or an account owner's exception.
An agent on Major can assess conflicting context and recommend the next action. The deterministic app preserves evidence and previous decisions so later runs do not repeat completed work. This complements native intelligence without claiming ZoomInfo lacks automation or cross-system features.
How ZoomInfo connects to Major
Major has a native ZoomInfo connector, identified as zoominfo in its connector catalog. IT sets up the connector once, approves the required access, and stores credentials centrally in the connector vault. Credentials stay out of application code and agent prompts, while connector permissions limit what each app, agent, and workflow can access.
Major orchestrates ZoomInfo; it never replaces it. ZoomInfo remains the source of licensed intelligence, while your CRM and engagement platform remain the systems where sellers work.
Before implementation, verify the tenant's API entitlement and the native connector operations available to your organization. A documented vendor endpoint is not a promise that every connector action exposes it automatically. Start with the smallest approved operation set and test the actual account.
Separate service access from personalized access. ZoomInfo's authentication guide supports server-to-server Client Credentials and user-based Authorization Code with PKCE. Choose the identity that matches the job, then apply SSO, RBAC, and audit logging across the Major app, agent, and workflow. A shared integration should not accidentally stand in for a seller's personalized recommendations.
ZoomInfo API guide for builders
Authentication, permissions, and testing
ZoomInfo uses OAuth 2.0 Bearer access tokens. In the Client Credentials flow, the application exchanges its client ID and secret for a token, with actions attributed to a selected integration user. The PKCE flow represents a signed-in user and supports access and refresh tokens.
The scope list names api:data:company, api:data:contact, api:data:intent, and api:account-summary:read. Request only the capabilities the build needs. Scopes do not replace package entitlements, such as Intent or Copilot access.
For testing, the Test API Access guide provides a portal-generated Bearer token for exploration. A separate isolated sandbox or guaranteed non-billable test environment is not documented in that guide. Test a small approved dataset and confirm credit treatment before bulk runs.
Pagination, limits, and event delivery
Search uses page[number] and page[size], described in the pagination guide. Save the page and query definition in app state. Search reference pages allow sizes up to 100, while the rate-limit guide describes direct Search throughput using 25 records per request. Start at 25 and verify accepted behavior rather than designing around the larger figure.
The rate-limit guide documents Builder at 5 requests per second, 10,800 per hour, and 129,600 per day; Standard at 25, 54,000, and 648,000; Scaling at 35, 75,600, and 907,200. All three windows apply. Watch X-RateLimit-Limit-Second, X-RateLimit-Remaining-Second, and the corresponding Hour and Day headers. On 429, inspect X-RateLimit-Rejected-Bucket, Retry-After, and X-RateLimit-Reset.
ZoomInfo's webhook guide describes asynchronous completion, record changes, credit thresholds, and GTM signals through its Agents API. It says retry frequency and throttling are configurable, but does not publicly specify a signature algorithm, verification header, or universal retry schedule. Confirm those details before accepting callbacks. Do not assume a generic contact-enriched subscription exists; scheduled discovery is a practical baseline.
Search contacts
POST /gtm/data/v1/contacts/search discovers matching people, but does not return engagement-ready emails or phone numbers.
- App: display candidate contacts with their ZoomInfo IDs and selection status.
- Agent: assess role fit from permitted search results and account context.
- Workflow: page through an approved cohort, then enrich only selected candidates.
Search companies
POST /gtm/data/v1/companies/search discovers companies and returns basic information rather than the full enrichment record.
- App: maintain a candidate-account queue with domain and identity matches.
- Agent: assess whether an ambiguous company belongs in the target market.
- Workflow: run approved filters and stage company IDs for enrichment.
Search intent
POST /gtm/data/v1/intent/search finds company-level research signals, requires the Intent SKU, and requires between one and 50 topics per request.
- App: retain topic, company, signal date, and disposition together.
- Agent: judge whether the signal matters alongside account history.
- Workflow: retrieve approved topics on a schedule and route relevant changes.
Enrich contacts
POST /gtm/data/v1/contacts/enrich retrieves details for up to 25 contacts, with selected output fields and per-record matching outcomes.
- App: show the enrichment snapshot beside the existing prospect record.
- Agent: resolve uncertain role matches without inventing missing data.
- Workflow: batch selected IDs, inspect every record outcome, and stop credit-exhausted items.
Enrich companies
POST /gtm/data/v1/companies/enrich retrieves details for up to 25 companies, preferably using IDs established during search.
- App: stage proposed firmographic changes before any CRM update.
- Agent: explain whether conflicting company information needs review.
- Workflow: refresh approved fields and preserve protected CRM values.
Enrich org charts
POST /gtm/data/v1/companies/org-chart/enrich returns contacts organized by seniority for a company and requested departments.
- App: track stakeholder coverage and unresolved identity matches.
- Agent: suggest which stakeholder role is missing from a buying group.
- Workflow: retrieve permitted departments and stage suggestions for the account owner.
Read usage
GET /gtm/data/v1/users/usage returns the current user's API usage and limits.
- App: show usage alongside queued enrichment work.
- Agent: summarize which pending research deserves scarce remaining capacity.
- Workflow: check usage before batches and defer lower-priority work under configured policies.
Retrieve an account summary
GET /gtm/copilot/v1/companies/{companyId}/account-summary retrieves an account summary for a Copilot customer's configured Target Account.
- App: retain a dated briefing with its company ID and source.
- Agent: combine the summary with approved CRM context to prepare a meeting brief.
- Workflow: retrieve eligible summaries before meetings and route unavailable accounts for review.
ZoomInfo automation use cases on Major
Move enriched leads into Outreach with context
Use contact search and enrichment to prepare candidates, then hand approved records to Outreach.
- App: hold source IDs, existing prospect matches, recommended sequence, review status, and resulting enrollment IDs.
- Agent: choose among approved sequences and draft a defensible first touch.
- Workflow: check eligibility and duplicates, request approval, then execute the permitted handoff.
This is grounded in the published ZoomInfo-to-Outreach use case. Enrollment can initiate communication, so a missing suppression result must block it.
Reconcile company enrichment with Salesforce
Use company search and enrichment to propose changes without treating external data as an unquestionable overwrite.
- App: display old values, proposed values, source dates, and field-level protection rules.
- Agent: judge ambiguous entity matches and explain material conflicts.
- Workflow: send protected-field changes to the account owner and apply only approved updates.
A subsidiary and its parent may share branding but require different sales ownership. Preserve both candidate matches until a reviewer resolves the identity.
Prioritize intent signals against HubSpot account context
Use intent search and company enrichment alongside CRM ownership and recent activity.
- App: retain a dated signal queue and previous dispositions.
- Agent: rank relevance and explain why a signal merits attention now.
- Workflow: suppress repeated alerts and notify the assigned owner through an approved channel.
Company research activity is evidence for prioritization, not proof that a named person wants a sales call. Keep that distinction in every recommendation.
How to build the ZoomInfo architecture on Major

ZoomInfo intelligence becomes a governed handoff with durable state on Major.
Build around one durable record per proposed business action. Store source IDs, a snapshot timestamp, the approved query definition, the agent's rationale, approval status, execution attempts, and downstream result IDs. Separate discovery from delivery so an interrupted handoff does not force another search and another judgment call.
- Connect: IT authorizes ZoomInfo and the destination system through scoped connectors.
- Create the app: add managed tables for candidates, recommendations, approvals, and outcomes, plus review and history views.
- Configure the agent: supply approved segments and destination choices, and require a recommendation with supporting evidence or an escalation.
- Orchestrate: schedule discovery, enforce deterministic eligibility checks, invoke judgment where needed, and wait for required approval.
- Record: persist the destination result before marking the action complete.
The workflow node represents orchestration, not judgment. The app is the control surface and execution layer; the agent makes the contextual call. Once a recommendation is recorded, a transient delivery retry should reuse it unless its inputs have materially changed. Reason once. Run forever.
A worked example with round numbers
For example, suppose an SDR manager reviews 600 candidate contacts each week. The manual process takes six minutes per contact to inspect enrichment, check duplicates, choose a sequence, and prepare a handoff. That is 600 × 6 = 3,600 minutes, or 60 hours weekly.
In an illustrative Major design, deterministic checks identify 150 duplicates or ineligible contacts before judgment. The remaining 450 records receive agent recommendations, with people reviewing all initial outbound enrollments during the pilot.
Assume reviewing each prepared recommendation takes two minutes instead of six. That is 450 × 2 = 900 minutes, or 15 hours. Add three hours for exception handling and oversight, bringing the modeled total to 18 hours. The arithmetic difference is 42 hours per week, not a measured customer result or a promised saving.
Track the real baseline during rollout: minutes per review, incorrect matches, override frequency, duplicate enrollments, and time waiting for approval. If accuracy declines, narrow the cohort before increasing volume. API charges, licensing, and ongoing maintenance are separate from this labor illustration.
Governance, approvals, and failure handling

Data refresh and outbound communication follow different approval paths.
Allow routine, non-sending state refreshes only within approved field and budget policies. Require a person for initial outbound enrollment, ambiguous identity, protected CRM changes, or a new sequence mapping. A suppression or unresolved consent flag blocks execution rather than becoming an exception the agent can waive.
The agent must never broaden its own permissions, fabricate a missing email, invent a sequence ID, or interpret intent as consent. Treat vendor data and retrieved text as evidence, not instructions that can override policy. Limit personal data exposure and retention according to your legal and contractual obligations.
For operational recovery, persist an action key before downstream writes. After a timeout, inspect the destination before retrying so an uncertain response does not create duplicate prospects or enrollments. Deduplicate repeated callbacks and overlapping scheduled runs against the same action record.
Follow ZoomInfo's 429 guidance: use backoff for short-window contention and respect Retry-After for longer exhaustion. Inspect partial enrichment outcomes, including credit exhaustion inside an otherwise successful response. Audit the input snapshot, policy version, recommendation, reviewer, and final write so recovery remains attributable.
ZoomInfo AI Apps, Agents and Workflow Automations
Every published Major use case that works with ZoomInfo, newest first. Each one names the ZoomInfo API objects involved and shows the app, agent and workflow split.
- Automate ZoomInfo-to-Outreach Lead Handoffs on Major: Move enriched ZoomInfo contacts into the right Outreach sequence without losing context. This brief maps the data flow, judgment points, duplicate checks, approvals, and controls an SDR team needs.
Frequently asked questions
What is ZoomInfo AI?
ZoomInfo AI refers to intelligence and automation features within ZoomInfo products, including Copilot account summaries and GTM Studio's AI-assisted exploration, scoring, and plays. Major can coordinate that intelligence with enterprise-specific state, approvals, and actions across systems.
Does Major have a native ZoomInfo connector?
Yes. Major's connector catalog lists a native ZoomInfo connector. IT configures it once with scoped access and centrally stored credentials. Verify the operations and account entitlements required for your specific workflow before implementation.
Does ZoomInfo have an API for AI agents?
Yes. ZoomInfo documents search, enrichment, context, and agent-related APIs. OAuth permissions and product entitlements determine available access. An agent should receive only the fields and operations needed for its task, with consequential actions controlled by the workflow.
Does searching ZoomInfo return email addresses and phone numbers?
Contact Search does not return engagement-ready emails or phone numbers. Use contact enrichment for permitted details after selecting candidates. Confirm field access and credit treatment, and never treat an enriched address as evidence of consent to outreach.
Does ZoomInfo support webhooks?
Its public guide describes events for asynchronous work, record changes, credit thresholds, and GTM signals through the Agents API. Signature verification details and a universal retry schedule are not specified there. Verify tenant-specific delivery security before relying on callbacks.
Can ZoomInfo AI automatically enroll leads in Outreach?
A governed integration can use ZoomInfo data to prepare an Outreach handoff. On Major, an agent recommends the sequence, an app stores evidence and review state, and a workflow performs only approved enrollment after eligibility, suppression, and duplicate checks.
Sources
- ZoomInfo API overview: API families and the official starting point for integration planning.
- Authentication: supported OAuth models and Bearer token usage.
- Client Credentials flow: service identity and integration-user attribution.
- Authorization Code with PKCE: signed-in user authorization and token handling.
- ZoomInfo OAuth scopes: exact permission names and capability descriptions.
- Pagination, batching, and bulk: discovery and enrichment workload design.
- Rate limits: simultaneous quota windows, response headers, and recovery guidance.
- Webhooks and events: documented event uses and configurable retries.
- Test API Access: portal-generated tokens for development and exploration.
- Companies and Contacts: company and contact resources for revenue teams.
- Signals and Insights: raw signals, contextual insights, and entitlement boundaries.
- GTM Studio: native AI-assisted exploration, scoring, and execution capabilities.
- Search Contacts: discovery output and its engagement-data boundary.
- Search Companies: company discovery and basic returned information.
- Search Intent: topic requirements and Intent access prerequisites.
- Enrich Contacts: contact batches, field selection, and partial outcomes.
- Enrich Companies: company enrichment and matching behavior.
- Enrich Org Charts: department and seniority context for stakeholder coverage.
- Usage Data: current-user usage and limit visibility.
- Get Account Summary: Copilot and Target Account prerequisites.
- OpenAPI Specification: downloadable schema resources for implementation validation.
Start building on Major
Build a ZoomInfo handoff app with durable enrichment history, scoped connectors, and a review queue. Let an agent make contextual recommendations while workflows control approvals and delivery.
About the author
Ritvij Gautam is Head of Growth at Major.
Last reviewed November 20, 2026 by Ritvij Gautam.
Frequently asked questions
- What is ZoomInfo AI?
- ZoomInfo AI refers to intelligence and automation features within ZoomInfo products, including Copilot account summaries and GTM Studio's AI-assisted exploration, scoring, and plays. Major can coordinate that intelligence with enterprise-specific state, approvals, and actions across systems.
- Does Major have a native ZoomInfo connector?
- Yes. Major's connector catalog lists a native ZoomInfo connector. IT configures it once with scoped access and centrally stored credentials. Verify the operations and account entitlements required for your specific workflow before implementation.
- Does ZoomInfo have an API for AI agents?
- Yes. ZoomInfo documents search, enrichment, context, and agent-related APIs. OAuth permissions and product entitlements determine available access. An agent should receive only the fields and operations needed for its task, with consequential actions controlled by the workflow.
- Does searching ZoomInfo return email addresses and phone numbers?
- Contact Search does not return engagement-ready emails or phone numbers. Use contact enrichment for permitted details after selecting candidates. Confirm field access and credit treatment, and never treat an enriched address as evidence of consent to outreach.
- Does ZoomInfo support webhooks?
- Its public guide describes events for asynchronous work, record changes, credit thresholds, and GTM signals through the Agents API. Signature verification details and a universal retry schedule are not specified there. Verify tenant-specific delivery security before relying on callbacks.
- Can ZoomInfo AI automatically enroll leads in Outreach?
- A governed integration can use ZoomInfo data to prepare an Outreach handoff. On Major, an agent recommends the sequence, an app stores evidence and review state, and a workflow performs only approved enrollment after eligibility, suppression, and duplicate checks.