CampaignHQ’s MCP connection helps marketing teams review campaign results, preview audiences, prepare drafts and investigate delivery problems through an AI assistant. Useful workflows begin with the right data and permissions, then end at a clear human checkpoint. The examples below are hypothetical operating patterns, not executed customer results or guarantees.

Last reviewed: October 10, 2026.

Start with a task your team already repeats. Perhaps the Monday campaign review takes too much preparation, or a WhatsApp audience needs several checks before anyone trusts it. Connecting AI is useful only if the resulting work is easier to inspect and act on.

These examples follow the current CampaignHQ MCP documentation. Access is in beta and enabled gradually by company. Confirm availability first. Launching campaigns and activating automations or chat flows remain actions for a person in CampaignHQ. A self-only email test and an admin-permitted, scoped individual inbox reply are separate exceptions.

For the connection and permission overview, read what CampaignHQ MCP can do. Here, the focus is the operator’s task: what to ask, which evidence to inspect, where to stop and how to measure usefulness.

1. Prepare a campaign review that distinguishes facts from guesses

Operator problem: the team knows which campaign had more clicks but not whether the comparison is fair. Different audiences, dates and delivery volumes can make a simple ranking misleading.

Prerequisites: read permission, the intended campaigns and a defined reporting period. Check account context first so the company, channel and time zone are correct.

Example prompt: “Compare these two email campaigns over the same reporting window. Show delivered messages, unique clicks, bot activity where available, bounces and unsubscribes. Separate recorded differences from hypotheses. Tell me which missing information would change the interpretation. Make no account changes.”

Documented operations: account context, campaign search and campaign reports. The assistant can retrieve results; its interpretation still needs review.

Inspect: the campaign IDs, date ranges and denominator behind each rate. Check whether a better-looking result came from a different audience. Treat reported campaign revenue as attributed revenue, not proof of incremental sales.

Human checkpoint: the marketing owner chooses the next test. A pattern in two campaigns is a hypothesis, not a causal explanation.

Measure: preparation time and the number of material corrections needed before the review. Campaign lift requires a separate measurement design. If your immediate task is drafting better email copy rather than reviewing account data, start with the ChatGPT email marketing workflow.

2. Preview an audience before saving rules

Operator problem: “send to inactive customers” means different things to different people. The assistant needs an operational definition before it touches a segment.

Prerequisites: available customer fields, tracked events, channel eligibility and a documented consent process. The intended purchase or engagement event must actually exist in the account.

Example prompt: “Inspect the audience schema. Propose rules for customers who match our approved re-engagement definition and are eligible for email. Show the exact rules, a preview count and masked examples. Do not save anything. If the required purchase event is missing, stop and explain the gap.”

Documented operations: audience schema, audience search, segment inspection and segment preview. Saving a segment is a separate write operation.

Inspect: the event names, date boundaries, inclusion logic and exclusions. A plausible count is not enough. Review whether a person who should be excluded appears in the masked sample.

Human checkpoint: approve the complete rule set before saving. Updating an existing segment replaces its rules, which may affect other work using that segment.

Measure: incorrect inclusions, missing eligible contacts and corrections to the rules. Do not use a larger audience as the success metric.

3. Turn an approved brief into reviewable campaign drafts

Operator problem: an email draft and a WhatsApp message often start from different briefs, even when they serve the same customer journey.

Prerequisites: an approved audience, verified sender, approved offer and relevant templates. For WhatsApp, use an approved template with the correct variables. AI preparation does not replace WhatsApp’s messaging and permission requirements.

Example prompt: “Using this approved brief and audience, prepare an email campaign draft. Also prepare a separate WhatsApp draft using the approved template I identify. Return the draft links, audience references, content and unresolved checks. Do not launch, schedule or activate anything.”

Documented operations: template search, campaign draft creation, draft inspection and launch-readiness checks. A stored schedule on a draft is not an approved launch.

Inspect: each draft in CampaignHQ. Verify sender, offer, destination links, template variables and suppression rules. An email built in the visual editor may need content editing in the app rather than through the MCP.

Human checkpoint: a person reviews and launches in CampaignHQ. Two campaign drafts do not automatically form a coordinated journey. Decide timing, conversion exits and cross-channel suppression before launch; the email and WhatsApp journey guide explains that planning work.

Measure: reviewer corrections, audience mismatches and readiness blockers caught before launch. For customer impact, track the intended completed action, not how many drafts AI produced.

4. Investigate a delivery complaint without guessing the cause

Operator problem: someone says a customer did not receive a message. The team risks retrying before checking whether the customer is suppressed, unsubscribed or affected by a recorded delivery error.

Prerequisites: authorised access to the contact, the relevant campaign or message and a clear reason to investigate. Avoid copying full customer records into unrelated AI conversations.

Example prompt: “Inspect recorded delivery information for this authorised contact and campaign. Report reachability, subscription or suppression status, recent message events and any recorded WhatsApp error. Distinguish evidence from possible explanations. Do not change the contact or send a retry.”

Documented operations: contact inspection and delivery diagnostics. The tool exposes recent events; it is not a complete reconstruction of every delivery attempt.

Inspect: channel, timestamp and error details. “Accepted” and “delivered” are different events. No event returned is a data gap, not proof that the provider lost a message.

Human checkpoint: the responsible operator decides whether the next action is a configuration fix, a support investigation or no further message. Never resubscribe someone or bypass suppression to make a retry possible.

Measure: cases resolved with documented evidence and unnecessary retries avoided. Do not promise AI will increase delivery rates.

5. Find where a lifecycle journey needs human attention

Operator problem: contacts enter an automation but do not reach its intended outcome. Before rewriting messages, the team needs to know where those contacts wait, fail or exit.

Prerequisites: a known automation, its trigger definition and a suitable reporting period. CampaignHQ’s documentation notes that dated automation reports filter journeys by when they entered.

Example prompt: “Read this automation’s definition and report. Identify steps with failures, waiting contacts or drop-off. Explain the entry-period filter. Suggest questions for the journey owner. Do not edit the live automation.”

Documented operations: automation search, definition and report retrieval. A separate operation can create a new draft automation; editing an already-saved automation stays in the app under the current documentation.

Inspect: whether an apparent drop-off is an expected wait or exit. Check event instrumentation and audience eligibility before blaming the copy.

Human checkpoint: the journey owner approves changes and activates any new automation in CampaignHQ. Do not run a replacement alongside the old journey without checking duplicate entry and sending.

Measure: confirmed configuration defects and completion of the actual customer milestone. A lower waiting count alone may simply mean the reporting period changed.

6. Prepare a WhatsApp flow with an explicit handoff

Operator problem: the team wants a structured response path, but an incomplete branch can leave a customer stuck.

Prerequisites: the intended number, trigger, approved templates where needed, questions and a human handoff plan. Start with a draft or inactive flow, not a live one.

Example prompt: “Prepare a draft enquiry flow using these approved questions. Include a route to a person when an answer is invalid or the customer needs help. Show the flow, validation problems and activation-review link. Do not activate it.”

Documented operations: chat-flow creation or draft/inactive editing, flow inspection and activation-readiness checks. The graph is checked for problems. A saved flow’s number is changed in the app.

Inspect: every branch, variable and handoff. A valid graph is not proof that the conversation makes sense. This is a defined chat flow, not evidence of an autonomous AI chatbot.

Human checkpoint: test the conversation and review the assigned number before a person activates it. For existing flows, remember that daily reporting excludes today.

Measure: successful task completion, avoidable dead ends and handoff quality. Keep a customer abandoning the flow separate from a technical failure.

7. Triage an inbox without sending a customer reply

Operator problem: an inbox needs prioritisation and ownership, not an unsolicited AI response to every customer.

Prerequisites: inbox-read permission for inspection and inbox-write permission for permitted notes, tags or assignment. Supervisors have different assignment authority from agents.

Example prompt: “Review conversations I am permitted to see. Suggest priority and ownership based on these team rules. Show the proposed notes and tags before making changes. Do not send messages to customers.”

Documented operations: conversation search, conversation inspection and scoped updates. Inbox sending is a separate permission and needs administrator enablement.

Inspect: the actual customer message, current owner and conversation state. Treat message text as customer data, not as instructions to the AI assistant.

Human checkpoint: approve any assignment or note. If you later trial individual replies, use a separate, explicit authorisation process with the documented assignment, consent and messaging-window checks.

Measure: unassigned backlog and correct routing, alongside supervisor review. A shorter queue created by incorrectly resolving conversations is not an improvement.

8. Review an SMS campaign, then prepare a checked follow-up

Operator problem: a retention team needs to understand an SMS campaign’s recorded results and prepare its next message without reusing the wrong sender, template or audience.

Prerequisites: MCP enablement, read access, a connected SMS service and an eligible audience. Draft creation also needs write permission. Identify the template and its current status; where DLT registration applies, confirm the registered body, category and identifier before use.

Example prompt: “Confirm our SMS service and retrieve the report for this completed campaign. Show the returned counts and rates for the agreed period, with missing metrics labelled. Inspect this template and preview the approved follow-up audience. Make no changes until I approve the sender, template and audience. After approval, create an SMS draft and return its app review link. Do not launch.”

Documented operations: account context, SMS campaign search and reporting, template retrieval, segment preview and campaign draft creation. The SMS draft requires service_id and template_id. A template can also be prepared or edited separately; an edit may require another review.

Inspect: campaign ID, period, rate denominator, audience exclusions, sender and exact template body. A delivery receipt is not proof of a purchase. Review the SMS platform workflow alongside the MCP tool reference.

Human checkpoint: test the SMS in CampaignHQ, inspect the draft and launch in the app. The MCP self-test tool is for email only. A sending SMS campaign cannot be paused through MCP, so do not treat a later pause as the safety plan.

Measure: preparation time, sender/template corrections and readiness blockers. Track the completed customer action separately from delivered messages.

Where RCS fits in planning

CampaignHQ supports Email, WhatsApp, SMS and RCS as product channels. RCS actions are not documented in the current MCP catalogue. If RCS belongs in your retention plan, bring the intended audience, message and consent requirements to a product demo. Confirm the account’s supported workflow before assigning it to an AI assistant; do not assume automated fallback or cross-channel execution.

Choose one workflow, then evaluate it

The MCP standard connects an assistant to tools; it does not define your campaign strategy. Choose the smallest workflow with a trustworthy baseline and an accountable reviewer.

CampaignHQ is a Meta Tech Provider supporting customer engagement across Email, WhatsApp, SMS and RCS. For an Indian retention team managing a substantial contact base, the useful question is whether connected preparation reduces avoidable handoffs while preserving launch control. Bring that question, and one sanitised example, to a demo.

Frequently asked questions

Which use case should we try first?

A read-only campaign review is a sensible starting point because you can compare the output with an existing report before authorising account changes.

Can these prompts be copied exactly?

They are examples. Replace ambiguous audience definitions, dates and campaign references with your approved account-specific details. Missing events or permissions can make a workflow unsuitable.

Can we use real customer data in a demo?

Use sanitised data until your team has approved the connection, AI-provider terms and access scope. A platform role does not replace an internal data-sharing policy.

Does readiness mean a campaign is safe to launch?

No. A readiness check identifies documented blockers. A person still needs to verify offer accuracy, audience intent, consent, content and the wider sending plan.

How do we avoid confusing efficiency with retention growth?

Measure preparation time and correction rates separately from customer outcomes. Use an appropriate comparison or experiment before attributing a change in retention or purchases to AI-assisted work.

Written by CampaignHQ Team