Major

Meta Ads Automation: API, AI Agents, and Governed Workflows

Connect Meta Ads to governed reporting, creative review, and approved campaign operations on Major. Explore native AI, API permissions, eight endpoints, and cross-system workflows.

Ritvij GautamUpdated
Major x Meta Ads logo lockup

Meta ads automation connects campaign reporting, anomaly review, creative preparation, and approved changes across your marketing stack. This guide covers Meta Ads' native AI, its Marketing API, and how to build governed apps, agents, and workflows on Major without handing an agent unrestricted spending authority.

What Meta Ads is and who relies on it

Meta Ads is the advertising system for reaching audiences across Meta technologies. Its Marketing API overview describes programmatic campaign management and links the advertising objects that developers work with: campaigns, ad sets, ads, and creatives.

Paid media managers organize objectives and delivery settings. Creative operations teams supply the assets and messages. Marketing operations teams reconcile spend and outcomes with analytics and CRM records, while finance partners check pacing against approved plans.

Those jobs need different evidence. A delivery issue belongs with the campaign owner; a creative refresh belongs with the brand reviewer. A revenue discrepancy needs the reporting window, attribution assumptions, and source records before anyone changes a budget.

Major is the enterprise platform where agents build the software they run on. For this team, an agent can build a deterministic reporting app that stores campaign history and calculates agreed metrics. The agent then interprets exceptions, rather than reasoning through the same joins every morning. People retain the controls for approving consequential actions.

Meta Ads' built-in AI and where Major adds value

Meta already provides substantial optimization inside its advertising system. The Advantage+ campaigns guide describes audience, campaign budget, and placement automation for sales, app, and leads campaigns. These features help distribute delivery within the advertiser's chosen campaign setup.

Meta also documents generative AI features for ads: text generation, image expansion, and background generation. Text generation proposes variations from existing copy and business context. Image expansion adapts imagery for additional placements. Advertisers remain responsible for previewing generated creative before publishing.

Use those features for the jobs they perform within Meta. Major adds the surrounding business process: comparing spend with another channel, connecting outcomes to CRM revenue, collecting a brand review, and keeping the evidence behind a decision.

For example, Meta can generate alternative copy while an agent on Major judges whether the proposed message conflicts with a product restriction stored elsewhere. The app holds the draft, source guidance, reviewer, and version. The workflow routes approval and executes only an authorized change. Neither system needs to pretend the other does not exist.

How Meta Ads connects to Major

Major has a native Meta Ads connector, identified in its connector catalog as metamarketing. IT sets up the connector once and grants access to the required advertising assets. Credentials stay centrally stored in the connector vault, not in application code or agent prompts.

Major orchestrates Meta Ads; it never replaces it. Ads Manager and Meta's delivery systems remain the advertising control and execution environment. Major provides the cross-system app layer, judgment steps, orchestration, and review record around that environment.

Start with reporting permissions and an explicit account allowlist. Separate the reporting process from any action that can create or alter live advertising. The presence of a native connector does not mean every Marketing API operation is exposed or authorized in a particular workspace. Confirm the available operations before enabling a write path.

SSO, RBAC, and audit logs cover the apps, agents, and workflows involved. Give analysts access to results without automatically granting campaign mutation rights. Give approvers authority over defined accounts, and keep connector administration with IT rather than embedding that responsibility in a marketing prompt.

Meta Ads API guide for automation builders

The Marketing API uses the Graph API host, https://graph.facebook.com. Reference paths below omit the version prefix; production requests should pin and test a supported version rather than silently changing behavior.

Authentication, permissions, and testing

Meta documents user and system user access tokens. User tokens represent a person; system user tokens support business server-to-server work. Token validity and asset access still require monitoring. An app access token has a separate role when configuring webhook subscriptions.

For reporting, request ads_read. Campaign management requires ads_management. Meta's authorization guide distinguishes permission access levels from Marketing API access tiers. Managing other people's accounts requires appropriate advanced permission access; app review and business verification depend on the intended access.

Do not describe development access as a sandbox. Meta explicitly states that calls at any access level operate against production data. Begin with read-only queries, captured response fixtures, and paused test objects in a controlled account, with approval before creation or activation.

Pagination, limits, and events

Cursor pagination uses before, after, and limit parameters. Follow the returned next-page link until it disappears, even when a page is empty. Do not treat cursors as permanent checkpoints.

Marketing API rate limits vary by tier, operation, and account. The documented account score ceiling is 60 for development access and 9,000 for Full access, with a 300-second decay period. Reads generally cost one point and writes three. Separate restrictions also apply, including four ad set budget changes per hour.

Monitor X-Ad-Account-Usage and X-FB-Ads-Insights-Throttle, plus the documented business-use-case usage headers. Use the returned reset estimates and error subcodes rather than assuming one universal requests-per-minute limit.

Ads webhooks notify on delivery status, creative fatigue, recommendations, and processing issues. Read current API data after notification. Webhook setup requires a verification challenge and HMAC-SHA256 validation of the raw body using X-Hub-Signature-256. Meta's general webhook guide documents immediate retry followed by decreasing-frequency retries over 36 hours, so duplicate handling is essential.

GET /act{adaccount_id}/campaigns

Lists campaigns associated with an ad account; the default read excludes archived and deleted campaigns.

  • App: Maintain the campaign directory and account relationships.
  • Agent: Interpret an unfamiliar campaign name using its objective and reporting context.
  • Workflow: Refresh the directory before joining spend, with explicit status filters when historical coverage matters.

GET /act{adaccount_id}/adsets

Returns ad sets and requested configuration fields for an account.

  • App: Store the ad set inventory beside daily snapshots.
  • Agent: Assess whether a performance change needs an audience or creative investigation.
  • Workflow: Refresh configuration before presenting a proposed intervention, so reviewers see current settings.

GET /act{adaccount_id}/ads

Reads the account's ads and supports filters such as effective status and update time.

  • App: Show the affected ads in an exception queue.
  • Agent: Group related delivery problems into one useful explanation.
  • Workflow: Retrieve affected objects after an event instead of sending an alert from an incomplete notification.

GET /act{adaccount_id}/adcreatives

Lists the account's creative objects, with requested fields defining the returned content.

  • App: Associate creative identifiers with review records.
  • Agent: Compare permitted creative content with the current brief.
  • Workflow: Refresh the creative inventory before a review batch, avoiding repeated requests for unchanged material.

GET /act{adaccount_id}/insights

Returns advertising performance with requested metrics, reporting levels, date ranges, and supported breakdowns.

  • App: Calculate pacing and cost metrics from versioned source rows.
  • Agent: Explain an anomaly while distinguishing platform attribution from CRM outcomes.
  • Workflow: Collect a bounded reporting window and finish pagination before releasing the digest. Some metrics are estimated or in development, so preserve their interpretation.

POST /act{adaccount_id}/campaigns

Creates a campaign with an objective and other required settings.

  • App: Store the approved creation request and returned identifier.
  • Agent: Draft a campaign proposal from a brief, identifying missing requirements.
  • Workflow: Submit only an approved payload, explicitly using paused status for a controlled creation path.

POST /act{adaccount_id}/adcreatives

Submits creative configuration under the account, subject to Meta's format and validation rules.

  • App: Retain the asset references and reviewer-approved version.
  • Agent: Draft creative alternatives and flag unsupported claims.
  • Workflow: Submit the approved creative configuration only if that operation is available and permitted through the connector.

POST /act<ADACCOUNTID>/subscribedapps

Connects an account to an app's configured ads webhook subscriptions using an authorized account admin token.

  • App: Track subscription health by account.
  • Agent: Summarize which reporting or alerting processes a disconnection affects.
  • Workflow: Route reconnection to an administrator after account reactivation. Meta does not automatically restore connections removed when an account closes or becomes disabled.

Meta ads automation use cases on Major

Cross-channel spend reconciliation: Meta Ads and Google Ads

This is grounded in the queued reporting use case for the two advertising platforms. Keep platform conversion definitions visible rather than letting a model silently blend incompatible measures.

  • App: Join campaign inventories and daily spend using an approved channel mapping, currency policy, and reporting window.
  • Agent: Explain unusual cost or return movements and distinguish missing data from genuine deterioration.
  • Workflow: Schedule collection, reject incomplete batches, and deliver the reviewed weekly summary.
  • Meta endpoints: Campaigns, ad sets, and Insights, all GET operations above.

Revenue context: Meta Ads and HubSpot

A high platform-reported return does not explain whether sales-qualified pipeline followed. Keep the CRM outcome alongside the platform metric, with its own definition and date.

  • App: Hold approved campaign-to-CRM mappings and separate outcome columns.
  • Agent: Investigate discrepancies and draft questions for the campaign owner.
  • Workflow: Refresh both sources, wait for complete data, and route unresolved attribution questions.
  • Meta endpoints: GET Insights and campaigns.

Creative review: Meta Ads, Google Drive, and Slack

Bring the brief, the current creative, and the authorized reviewer together before submission. This is a proposed build pattern, not a claim of an existing customer deployment.

  • App: Store creative versions, brief references, and approval status.
  • Agent: Review language against the brief and draft a specific revision request.
  • Workflow: Notify the reviewer in Slack, then submit only the approved version if the required write operation is exposed.
  • Meta endpoints: GET and POST ad creatives, followed by a separate activation review rather than automatic launch.

How to build the architecture on Major

Meta Ads and Google Ads flow through a Major workflow, analysis agent, policy gate, and review app

Figure 1: Cross-channel reporting with judgment and durable review state

Build the deterministic foundation first. The app needs tables for accounts, campaign mappings, reporting snapshots, run status, proposed actions, and approval receipts. Use source identifiers rather than campaign names as keys, since names can change.

The workflow starts on a schedule or verified event, calls the scoped connectors, and runs repeatable validation and calculations. Only complete, labeled evidence goes to the agent. Its job is to interpret anomalies, prioritize investigation, and draft recommendations, not to improvise arithmetic or redefine metrics.

The decision node enforces policy in code. A read-only digest can proceed to its audience; a consequential proposal goes into the app's review queue. The app holds the record while the workflow waits for a person.

This is Major's reason once, run forever pattern in practice. Once an agent builds the ingestion and reconciliation software, later runs execute that code. Judgment remains with the agent where context changes; database state and repeatable processing do not depend on remembering an earlier conversation.

For example: a weekly reporting review with numbers

Consider an illustrative paid media team reviewing 40 campaign groups each week across Meta Ads and Google Ads. Suppose exporting, matching, and checking each group takes six minutes. That is 40 times six, or 240 minutes before the weekly discussion.

With a reporting app, assume a person spends one minute checking each completed group, plus 20 minutes resolving exceptions. The arithmetic becomes 40 times one plus 20, or 60 minutes. The illustrative reduction is 180 minutes per week, not a measured customer result or a promise about advertising performance.

The run begins with account inventories and a fixed reporting window. The app checks currencies, mappings, and missing rows. The agent reads the completed comparison and might flag five groups for investigation, with evidence attached. A person reviews those cases in the same app.

If one source fails, the workflow labels the report incomplete instead of presenting a false improvement. If a recommendation suggests changing delivery, the reporting run ends with a proposal, not an unapproved spend change. Measure data completeness and reviewer acceptance before expanding permissions.

Governance and failure handling for live advertising

Meta Ads recommendations split into read-only reporting and a human-approved write path with freshness checks

Figure 2: Approval binds to an exact proposal before any permitted Meta Ads write

An agent must never independently increase spend, activate advertising, widen targeting, or upload customer audience data in this design. Approval should bind to the exact account, object, proposed values, and expiration time. A changed proposal requires another approval.

Before execution, the workflow rechecks permissions and current values. If the campaign changed during review, return it to the queue. Keep approver identity, evidence snapshot, requested payload, API response, and outcome in the audit record.

For throttling, apply bounded exponential backoff with jitter and account-specific scheduling, following Meta's limit guidance. Do not blindly retry invalid permissions or malformed requests. After an ambiguous creation timeout, reconcile remote state before repeating the request.

For events, validate signatures before accepting data, persist receipt before acknowledging, deduplicate repeated changes, and fetch current state. Combine webhook-driven updates with scheduled reconciliation to catch missed events. Route exhausted retries and revoked credentials to an accountable owner, and keep an account-level pause switch available to operators.

Meta Ads AI Apps, Agents and Workflow Automations

This is the currently available core guide for this hub. The Meta Ads and Google Ads reconciliation article is queued, so there is no invented tutorial link for it here. Use this hub to choose the API objects and controls; use the linked guide for the broader marketing operating model.

For planning, start by choosing one reporting owner, a shared metric definition, and the account set that the owner can review. Decide what evidence an exception must contain and who receives incomplete-run notices. Those decisions make the pilot reviewable before anyone adds mutation permissions. A digest that reliably explains missing data is a better foundation than an unattended campaign launcher with no durable record of why it acted.

Frequently asked questions

What is meta ads automation?

Meta ads automation uses software to collect reporting, monitor delivery, prepare changes, and run approved advertising processes. On Major, the app stores state and runs repeatable calculations, the agent interprets context, and the workflow handles triggers, routing, retries, and approvals.

Does Meta Ads already have AI?

Yes. Meta documents Advantage+ audience, budget, and placement automation, plus generative AI features including text generation, image expansion, and background generation. Major adds cross-system context, review queues, durable state, and governed execution around that native functionality.

Does Major have a native Meta Ads connector?

Yes. Major has a native Meta Ads connector, identified as metamarketing. IT configures the connection centrally and scopes access. Verify the workspace's available operations before designing any campaign write action.

Which Meta permissions does reporting need?

Meta documents adsread for reading advertising reports and adsmanagement for campaign management. Permission access levels, account asset access, and the Marketing API access tier are separate requirements. Start with read-only access and grant additional authority only for a defined process.

Is Marketing API development access a sandbox?

No. Meta states that calls at any access level operate against production data. Test with read-only calls, captured fixtures, and controlled paused objects. Do not assume that a development tier prevents real account changes.

Can an AI agent change campaign budgets without approval?

The recommended Major design does not allow it. The agent drafts a reasoned proposal; an authorized person approves the exact change, and the workflow rechecks permissions and current state before any permitted execution. Reporting can run without granting campaign mutation rights.

Sources

Start building on Major

Build a Meta Ads reporting and review app with scoped connectors, durable campaign history, and human approval before consequential changes. Start with a read-only digest, then expand only after the team trusts the evidence and controls.

About the author

Ritvij Gautam is Head of Growth at Major.

Last reviewed November 20, 2026 by Ritvij Gautam.

Frequently asked questions

What is meta ads automation?
Meta ads automation uses software to collect reporting, monitor delivery, prepare changes, and run approved advertising processes. On Major, the app stores state and runs repeatable calculations, the agent interprets context, and the workflow handles triggers, routing, retries, and approvals.
Does Meta Ads already have AI?
Yes. Meta documents Advantage+ audience, budget, and placement automation, plus generative AI features including text generation, image expansion, and background generation. Major adds cross-system context, review queues, durable state, and governed execution around that native functionality.
Does Major have a native Meta Ads connector?
Yes. Major has a native Meta Ads connector, identified as metamarketing. IT configures the connection centrally and scopes access. Verify the workspace's available operations before designing any campaign write action.
Which Meta permissions does reporting need?
Meta documents ads_read for reading advertising reports and ads_management for campaign management. Permission access levels, account asset access, and the Marketing API access tier are separate requirements. Start with read-only access and grant additional authority only for a defined process.
Is Marketing API development access a sandbox?
No. Meta states that calls at any access level operate against production data. Test with read-only calls, captured fixtures, and controlled paused objects. Do not assume that a development tier prevents real account changes.
Can an AI agent change campaign budgets without approval?
The recommended Major design does not allow it. The agent drafts a reasoned proposal; an authorized person approves the exact change, and the workflow rechecks permissions and current state before any permitted execution. Reporting can run without granting campaign mutation rights.