Follow-ups & Cadences
By Conversa Labs
By Conversa Labs
Sequences, steps, channels, composition, enrollment (manual/automation/event/inactivity/AI), funnel, exit and human-in-the-loop AI.
Follow-ups and cadences overview
Overview ConversaLabs Follow-ups is a re-engagement cadence engine: sequences of automated messages, sent at the right moment, to recover stalled conversations, remind contacts of open payments, confirm appointments, revive deals in the CRM and nurture contacts over time — without anyone having to remember to message them manually. Unlike a one-shot campaign, a cadence has several steps with waits between them and smart exit rules: if the contact replies, reaches a goal (pays, wins, shows up) or the conversation is resolved, the cadence stops on its own. You design the sequence once and it works for each contact at their own pace. The module talks to the rest of the platform: contacts, conversations, CRM, Calendar, Payments and Automation. Steps can use the Maestro AI to compose messages, with optional human approval before each send. Prerequisites - The Follow-ups module enabled for your account (opt-in feature, controlled by an administrator). - A user with permission to create and manage cadences. - At least one outbound channel configured (for example, a WhatsApp inbox) so steps can send messages. - For AI steps: the Maestro enabled on the account (without it, AI steps fall back to static text). Step by step 1. Open the Follow-ups module from the sidebar. 2. Create a sequence from scratch or start from a ready-made template. 3. Add steps with the wait, channel and content of each message. 4. Define how contacts enter (enrollment) and when the cadence should stop (exit). 5. Publish the sequence to make it active. 6. Track the funnel and reports to measure replies and conversions. Settings & options - Sequences: the cadence blueprint (steps, trigger, target and exit rules), with versions. - Steps: each message, with its wait, channel, content type and compose mode. - Templates: reusable text to speed up step creation. - Enrollments: the cadence run for each contact, with its state and history. - Settings: time zone, quiet hours and the module's sending limits. Use cases - Recover unanswered conversations with 2 or 3 spaced touches. - Remind contacts of open charges until they pay (the cadence stops when they do). - Re-engage stalled CRM deals and confirm attendance before a meeting. - Nurture new leads with a welcome sequence spread over a few days. Tips, limits & best practices - Keep it simple: 2 to 4 steps usually outperform long cadences. - Always keep an exit on reply rule on so you never message someone who already answered. - Respect WhatsApp's 24h window: outside it, use an approved template (HSM) on the step. - Use quiet hours and sending limits for a natural, safe pace (anti-ban). Troubleshooting - I don't see the module: it may not be enabled for your account or your role — talk to an administrator. - The cadence doesn't send: check that the sequence is published and that a valid outbound channel exists. - The contact got a message even after replying: review the sequence's exit rules. See also - Sequences, steps, channels and composition - Enrollment: manual, automation, event, inactivity and AI - Cadence funnel and exit rules - AI with human approval
Sequences, steps, channels and message composition
Overview A sequence is the blueprint of your cadence. It is made of steps in order, each with a wait, a channel and content. When you publish the sequence, the platform takes a snapshot (version) of the steps: enrollments already running keep the version they started on, so editing a sequence never disrupts anyone already mid-cadence. This article covers how to structure the sequence, configure each step, choose the sending channel, and compose the message with variables and WhatsApp templates. Prerequisites - The Follow-ups module enabled and permission to create sequences. - A valid outbound channel (for example, a WhatsApp inbox, email or SMS). - To use custom variables, the contact's custom attributes configured. - To send outside WhatsApp's 24h window: an approved template (HSM) on the inbox. Step by step 1. In the Follow-ups module, create a new sequence and give it a clear name. 2. Set the cadence target (contact, conversation, CRM deal, appointment or charge). 3. Add the first step: choose the wait, the channel and write the message. 4. Add the following steps, adjusting each one's wait relative to the previous send. 5. Use the variable picker to insert contact/conversation data into the message. 6. Review everything and click Publish to activate the sequence. Settings & options Steps and waits Each step has a wait before it is sent, counted from the previous send. The wait can be in minutes, hours, days or business days. Business days and quiet hours are adjusted by the time zone set in the module's settings. Channels The channel defines how the step sends: | Channel | When to use | |---------|-------------| | Automatic | The platform picks based on the conversation/contact (recommended in most cases). | | Same conversation | Continues in the contact's existing conversation. | | Email | Sends by email. | | SMS | Sends by SMS (mind the message length). | | WhatsApp | Sends by WhatsApp, respecting the 24h window. | Content type and composition - Content type: text, template or rich content. - Compose mode: static (you write the text), AI assisted (the AI drafts and a human approves) or AI full (the AI composes and sends). See the AI with human approval article. Variables (personalization) Write personalized messages with variables (for example, the contact's first name, the agent's name, conversation data and custom attributes). Variables are resolved on the server, at send time — you only insert the placeholder via the variable picker. WhatsApp templates (24h window) If the step is WhatsApp and the contact is outside the 24h window (WhatsApp Cloud), only an approved template (HSM) can be sent. On the step, pick the template and map its parameters with variables. Inside the window (or on WhatsApp Web, which has no window), free text is sent normally. Message templates and sequence templates - Message templates: reusable text that speeds up step creation. - Sequence templates: ready-made cadences by niche; applying one also creates the custom attributes that cadence uses. Use cases - Billing cadence: step 1 immediate, step 2 after 1 day, step 3 after 3 days, all on WhatsApp. - Welcome: 3 emails over a week, using the contact's first name. - Reactivation: 2 touches in the same conversation with 2 and 5 days of wait. Tips, limits & best practices - Space steps out well: waits that are too short feel like spam. - For WhatsApp, always keep an approved fallback template on steps that may fall outside the window. - Test variables with a real contact before publishing so they don't render blank. - Publish again after any edit — only the published version applies to new enrollments. Troubleshooting - The message went out with a blank variable: the contact didn't have that data — pick a variable with a value or provide default text. - The WhatsApp step failed outside the window: the step was missing an approved template (HSM). - I edited the sequence and nothing changed: old enrollments keep the old version; publish and new enrollments will use the new one. See also - Follow-ups and cadences overview - Enrollment: manual, automation, event, inactivity and AI - Cadence funnel and exit rules - AI with human approval
Enrollment: manual, automation, event, inactivity and AI
Overview Enrolling means placing a contact inside a cadence. Each enrollment is the run of the sequence for that contact: it pins the published version, advances step by step on schedule, and stops on its own when an exit rule is met. ConversaLabs offers several ways to enroll, from a manual click to an automatic trigger. The platform avoids messaging the same person twice: there is at most one active (or paused) enrollment per contact in each sequence (the conversation is taken into account too). Repeated attempts are safely ignored. Prerequisites - A published and active sequence. - For event/inactivity/AI enrollment: the matching trigger configured on the sequence. - For automation, macro or flow enrollment: the Automation/Flow module available. - For AI enrollment: the Maestro enabled on the account. Step by step 1. Choose the enrollment method that best fits the case (see the options below). 2. Confirm the sequence target (contact, conversation, deal, appointment or charge). 3. For automatic triggers, configure the condition (event or inactivity time) on the sequence. 4. Enroll (manual) or let the trigger act; the enrollment starts as active. 5. Track the enrollment in the enrollments list, with its state and history. Settings & options Enrollment methods | Method | How it happens | |--------|----------------| | Manual | You enroll one contact (or many, in bulk) directly from the interface. | | Automation | An automation rule runs the enroll action when its conditions match. | | Macro | An agent runs a macro that includes the enroll-into-cadence action. | | Flow | A Flow Builder node enrolls the contact in the middle of a flow. | | Event | A system event (stalled deal, overdue charge, appointment no-show, conversation resolved, etc.) triggers enrollment automatically. | | Inactivity | Conversations idle for more than X minutes are enrolled automatically. | | AI | The Maestro evaluates the conversation and decides whether to start the follow-up now. | | API | An external system enrolls contacts programmatically. | Event-based enrollment Configure the sequence with an event trigger and choose the event (for example, overdue charge, inactive deal, appointment no-show or conversation resolved). When the event happens, the platform enrolls the contact — provided the sequence target matches what the event carries. Inactivity-based enrollment Configure the sequence with an inactivity trigger and set the time limit in minutes and which conversation statuses count (by default, open and pending). Conversations idle beyond the limit are enrolled — each conversation is processed only once per sequence. AI-based enrollment (proposed) With an AI trigger, the Maestro analyzes the conversation and answers whether a follow-up makes sense now. If the Maestro is unavailable, enrollment is not blocked: the cadence proceeds by the normal rules. Use cases - Manual in bulk: select several contacts from a list and enroll them in a reactivation campaign. - Event: every overdue charge automatically enters the billing cadence. - Inactivity: conversations unanswered for 60 minutes enter a recovery cadence. - Automation: new leads from a specific channel enter the welcome sequence. Tips, limits & best practices - Match the sequence target to the enrollment method (a billing cadence has a charge target). - Don't try to enroll the same contact twice in the same sequence — the duplicate is ignored. - For inactivity triggers, tune the time limit to your business so it doesn't fire too early. - Use AI enrollment when you want a relevance filter before the cadence begins. Troubleshooting - The contact wasn't enrolled by an event: check the sequence is active, published and has the right trigger/target. - Inactivity didn't fire: review the time limit and the conversation statuses the sequence counts. - I enrolled and nothing showed up: there was probably already an active/paused enrollment for that contact in that sequence (one per target limit). See also - Follow-ups and cadences overview - Sequences, steps, channels and composition - Cadence funnel and exit rules - AI with human approval
Cadence funnel and exit rules
Overview As an enrollment advances, each step produces a send record with a status, forming the cadence funnel: from how many contacts entered to how many replied or converted. In parallel, exit rules decide when a cadence should stop before reaching the end — for example, as soon as the contact replies. Understanding the funnel and configuring exits well is what makes a cadence polite and effective: it speaks enough to re-engage, but stops the moment it has met its goal. Prerequisites - Enrollments in progress (see the enrollment article). - Permission to view reports and manage sequences. - For exit on goal: the linked record (charge, deal or appointment) tied to the enrollment. Step by step 1. Open the sequence and view the enrollments list and the per-step funnel. 2. On each enrollment, open the timeline to see the history of every step and event. 3. Set the sequence's exit rules (on reply, on resolution and/or on goal). 4. Publish the sequence to apply the exit rules to new enrollments. 5. Track the reports to measure replies, conversions and per-step performance. Settings & options Step statuses (funnel) Each sent step moves forward through statuses, always forward (a later status never goes back): | Status | Meaning | |--------|---------| | Pending | Waiting (includes AI drafts awaiting approval). | | Scheduled | Queued to send at the step's time. | | Sent | Message sent. | | Delivered | Delivery confirmed. | | Read | The contact read it. | | Replied | The contact replied (usually triggers the exit). | | Skipped | The step was skipped by a "skip" rule. | | Failed | The send failed. | Exit rules Exits decide when the cadence stops before the last step: - On reply (on by default): if the contact replies, the cadence stops. This is the most important rule. - On resolution: if the linked conversation is marked resolved, the cadence stops. - On goal: the cadence stops when the goal is reached — for example, a charge paid, a CRM deal won or an appointment completed. The goal is checked even if it was reached outside the cadence. - At the end of the sequence: completing the last step ends the enrollment naturally. Enrollment states The enrollment itself can be active, paused, completed (reached the end), exited (an exit rule stopped it) or failed. You can pause, resume, skip a step or cancel enrollments manually. Reports Reports show performance by sequence, by step and by compose mode, plus the distribution of enrollments by state — useful for tuning waits, copy and the number of steps. Use cases - Measure at which step most replies arrive and shorten the cadence from there. - Ensure the billing cadence stops as soon as the payment lands (exit on goal). - Spot steps with many failures and review the channel or message template. Tips, limits & best practices - Keep exit on reply on by default — it's what avoids nagging someone who already answered. - Use exit on goal for cadences tied to billing, CRM and Calendar. - Use reports to cut low-return steps instead of only adding more. Troubleshooting - The cadence continued after a reply: confirm exit on reply is enabled on the sequence. - Exit on goal didn't work: check the enrollment is linked to the right record (charge, deal or appointment) and that its state changed to the goal. - A step shows as failed: check the step's channel and whether an approved template is needed. See also - Follow-ups and cadences overview - Sequences, steps, channels and composition - Enrollment: manual, automation, event, inactivity and AI - AI with human approval
AI with human approval (human-in-the-loop)
Overview A cadence step can be composed by you, by the Maestro AI, or by a mix of both. The highlight is AI assisted (human-in-the-loop) mode: the AI drafts the step's message, but nothing is sent without a person approving it — and that person can still edit the draft before sending. This gives you the speed and personalization of AI without giving up control and your brand's tone. And if the Maestro is unavailable, the cadence doesn't stall: it uses the step's static text as a fallback. Prerequisites - The Follow-ups module enabled and a sequence with steps. - The Maestro enabled on the account (without it, AI modes fall back to static text). - A user responsible for reviewing and approving drafts (the notification recipient). - For best quality, a step with a clear instruction/intent for the AI. Step by step 1. In the step editor, choose the compose mode: static, AI assisted or AI full. 2. For AI, describe the step's intent (what the message should achieve) and leave a static fallback text. 3. Publish the sequence. When an AI assisted step is due, the AI creates a draft and the step stays pending, awaiting approval. 4. The responsible user receives a notification of a pending draft. 5. Review the draft, edit it if you want, and approve to send — or discard it. Settings & options The three compose modes | Mode | Behavior | |------|----------| | Static | You write the text (with variables). It's sent as is. | | AI assisted | The AI generates a draft and the step waits for human approval before sending. The draft can be edited. | | AI full | The AI composes and sends automatically, with no approval wait. | Human approval (HITL) In AI assisted mode, the step creates a draft and pauses at that point. The reviewer can: - Approve the draft as is; - Edit the text and then approve (the send uses your version); - Discard it so nothing is sent. Notifications When a draft is waiting, the platform notifies the responsible users — usually the enrollment assignee and the conversation assignee, falling back to administrators when no assignee is set. There are also reply received and enrollment failed notifications. Safe degradation If the Maestro is unavailable when the step runs, both AI assisted and AI full send the step's static fallback text. The cadence never stops because of an AI outage. WhatsApp window respect AI composition still respects WhatsApp's 24h window: outside it (WhatsApp Cloud), only an approved template (HSM) can be sent. Keep a template configured on the step. Use cases - AI assisted for sensitive messages (billing, negotiation), where a human checks the tone before sending. - AI full for simple, low-risk touches, scaling without manual review. - Static with variables when the text is standardized and doesn't need AI. Tips, limits & best practices - Start with AI assisted: review a few drafts to calibrate the intent before moving to AI full. - Always fill in the static fallback text — it's your safety net. - Write specific intents (goal, tone, language) for better drafts. - Make sure the responsible users receive notifications so drafts don't sit idle. Troubleshooting - The draft stayed pending and no one saw it: check who the enrollment/conversation assignee is and the account's notifications. - The AI sent the static text: the Maestro was likely unavailable — degradation is the expected behavior. - The AI generated nothing: check that the Maestro is enabled and that the step has a defined intent. See also - Follow-ups and cadences overview - Sequences, steps, channels and composition - Enrollment: manual, automation, event, inactivity and AI - Cadence funnel and exit rules