Follow Up Boss automation: what to wire up first

In short: Follow Up Boss handles a lot of follow up on its own. What still leaks sits at the edges, where the CRM meets your phone line, your portals, and your calendar. This is a build guide to six of them: the trigger, the steps, and the human gate for each, grounded in what the API and webhooks actually support.

Key takeaways

• Turn on what you already pay for first. Action Plans, Automations, Smart Lists, and the Team Inbox cover follow up that teams routinely pay someone to rebuild.
• Follow Up Boss does not hide its API behind a higher plan. The docs put an API key on the Admin then API screen for every user, carrying that user's permissions.
• Send leads with `POST /v1/events`, never `POST /v1/people`. Their docs say the people endpoint skips deduplication, Lead Flow assignment, agent notification, and action plans.
• Webhooks announce that something changed, not what changed, so a build reads the record back and tolerates repeats.
• Every workflow below has a named human gate. The system captures, qualifies, books, and hands off. It never negotiates, never advises on price, never discusses commission.

Where Follow Up Boss's own automation stops

Follow Up Boss ships real automation, and for most teams the first honest answer is to use more of it. Their product pages describe Action Plans that send an auto text and drip emails alongside task checklists, Automations that watch existing leads and fire on triggers, Smart Lists for who to follow up with and when, a Team Inbox routing calls, texts, and emails to business numbers, and native calling and texting. Their AI features page adds smart summaries in beta, suggested tasks, smart messages, and predictive lead prioritization, at no extra charge today, though many rely on data produced by FUB Calling.

Plan matters. Follow Up Boss publishes its prices: Grow at $69 per user a month, Pro at $499 a month including ten users, Platform at $1,000 a month including thirty. Calling, texting, team inboxes, shared automations, and call reporting come with Pro and Platform.

What none of it covers is what happens before the record exists, or outside the CRM entirely. The call at 9pm. The portal lead whose source string is a mess by the time anyone reports on it. The showing agreed to in a text thread that never reached a calendar. That is the work below, and the work we do. The vertical picture is on our real estate page; the wider stack is in real estate AI integrations.

The ground rules, from Follow Up Boss's developer docs

• The API is REST at `api.followupboss.com/v1`, authenticated with HTTP Basic using your API key as the username and a blank password, or OAuth 2.0 for apps acting on a user's behalf.
• Serving a Follow Up Boss customer means registering your system and sending `X-System` and `X-System-Key` headers. That is not paperwork: registered systems get a global limit of 250 requests per sliding ten second window instead of 125, and the same key is what signs your webhooks. Going over returns 429 with a `Retry-After` header.
• Webhooks are registered by the account owner, capped at two per event per system. The event list is wide: people created, updated, staged, and tagged, plus notes, emails, texts, calls, tasks, appointments, and deals.
• Payloads are thin: an event id, a timestamp, the event name, and resource ids with a URI to fetch the record. One bulk action can arrive split across several deliveries, so a build has to be idempotent. Every request carries an `FUB-Signature` header, an HMAC SHA256 keyed with your system key. Verify it.
• Failed deliveries retry five times over roughly eight hours, and answering with 406 or 410 disables the webhook outright. That failure pages nobody. It just stops.

One rule of ours, not theirs. An API key carries the same privileges as the login it belongs to, so an owner's key is the whole database. We do not put production integrations on a personal key. Staff leave, keys get rotated, and the day that account changes is the day leads stop arriving.

1. The after-hours call that becomes a lead

Today, by hand. After close, the line rings out or hits voicemail. Someone works the backlog at 9am, by which time the buyer who called at 8pm has spoken to two other brokerages.

The trigger. The phone rings when nobody is on it, or every agent is already on a call.

What the system does. A voice agent answers on the first ring, establishes whether this is a buyer, seller, renter, or cooperating agent, and asks what a good ISA would ask: the property, the timeline, the callback number. It writes the lead into Follow Up Boss through the events endpoint with the source set honestly, the transcript as a note, and the recording against the call. How the agents work is on our voice agents page.

The human gate. It captures, books, and hands off. It does not negotiate, advise on price, or discuss commission, and it does not present itself as a licensed representative. Trading in real estate in Ontario requires registration under the Trust in Real Estate Services Act, and software cannot hold one.

What changes. On a brokerage line we run, 120 of 167 calls in the measured window ended with a callback number and enough context for someone to act on the next morning. That is a count of what came through the line, not a before and after, because nobody measured the equivalent window beforehand. The method is in the impact study.

2. Speed to lead on the first touch

Today, by hand. Portal leads land in an inbox and the CRM at once. Whoever is free replies. On a Saturday, nobody is free.

The trigger. A new lead from a portal, a form, an IDX site, or the phone.

What the system does. It posts to `POST /v1/events` with the event type that matches what actually happened, because only some types carry downstream behaviour: the docs name Registration, Property Inquiry, Seller Inquiry, and General Inquiry as the ones that fire action plans and automations. It then sends the first message inside your window, answers factual questions from the listing data, and asks two or three qualifying questions in the reply rather than in a form.

The human gate. Anything resembling an offer, a condition, or a price conversation stops and pages a person. So does anyone who asks for a human, first time, no retention loop.

What changes. Response time stops being whoever is having their worst day.

3. Lead source hygiene, so the reporting means something

Today, by hand. Sources arrive as free text from a dozen integrations. Three months later nobody can say which spend produced which closing, because the same portal appears under four names.

The trigger. Every inbound lead, continuously.

What the system does. It normalises the source before the record is written, keeping Follow Up Boss's own distinction between `source`, the marketing name a broker recognises, and `system`, the technology that delivered it. Deduplication is theirs: they search for an existing contact by phone or email and update rather than duplicate, but only when you go through the events endpoint.

The human gate. The source taxonomy is a business decision, not a model output. A person owns the list and approves additions.

What changes. Cost per lead becomes a number you can defend in a partners' meeting.

4. Nurture that escalates when intent shows

Today, by hand. Long term nurture runs as a drip. Someone replies with a real question, the drip keeps sending, and the reply sits unread for two days.

The trigger. Webhooks on inbound texts, notes, and email marketing opens and clicks.

What the system does. It reads the reply and separates acknowledgements from buying signals. A question about a specific address, a changed timeline, or a mortgage approval creates a task for the assigned agent, moves the stage, and pauses the sequence so the human is not competing with the automation.

The human gate. Escalation runs one way. The system can hand a conversation to a person. It can never take one back.

What changes. High intent replies stop being buried under the low intent ones.

5. Appointments that land on a calendar

Today, by hand. A showing gets agreed to in a text. It exists in one person's memory until it does not.

The trigger. A caller or a reply accepts a slot.

What the system does. It checks live availability, offers slots, and creates the appointment through `POST /v1/appointments` with a type, a location, and the person attached. The detail that breaks most builds is in the docs: appointments created in Follow Up Boss sync to Gmail and Outlook only if the invitees list contains the Follow Up Boss user whose calendar is synced, alongside the other invitees. Miss it and the appointment is real inside the CRM and invisible on the agent's phone.

The human gate. Access instructions and lockbox codes are set by a person. A system that improvises property access is a security problem, not a convenience.

What changes. No shows drop, because the appointment exists where the agent actually looks.

6. Requalifying the stale ones

Today, by hand. Thousands of old leads sit in a database nobody touches, because working them is a week of calls at a bad hit rate.

The trigger. Scheduled, against a Smart List filtered by last activity and stage.

What the system does. It works the list in small dated batches, sends something short and specific rather than a blast, and updates the record on any reply, including the ones that say no. Live responses route to the assigned agent with the history summarised.

The human gate. Consent. Canada's anti-spam legislation requires consent for commercial electronic messages, express or implied, plus your business identification and a working unsubscribe, and the Government of Canada's guidance is that implied consent is time limited and unsubscribes must be actioned within ten business days. A person approves the segment before the first message goes.

What changes. The dead list becomes a scheduled process with a paper trail.

What we will not build

Anything that negotiates. Price, conditions, offers, commission. Those conversations are the job, and they belong to a registered human who is accountable for them.

Anything that impersonates. The agent says what it is when asked. A team that needs callers to believe otherwise is building something it will regret.

Anything that decides who deserves a callback. Surfacing likely intent is fine. A score that quietly buries an inquiry is not, in a business where human rights obligations apply.

Where to start

If you are not using Action Plans and Smart Lists properly yet, do that first. It is the cheapest improvement available and you have already bought it.

If those are running and the leak is still obvious, build the phone first, then attribution, then appointments. The phone is where the unanswered lead lives, and it is the only one that stops the day you switch it on.

Two bills, as always. Ours: the audit is a fixed fee agreed before it starts, a build is a fixed fee quoted against it, costing more the more systems it touches, and operating it is a monthly fee, with how the pricing works in what AI consulting actually costs. Theirs: seats, calling if you are on Grow, telephony, and per use model costs, which continue every month whether anyone is watching or not.

Frequently asked questions

Do I need a higher plan to use the Follow Up Boss API?

Nothing in their published pricing gates the API behind a tier, and the docs put an API key on the Admin then API screen for every user. What the plan changes is what surrounds it: calling and texting come with Pro and Platform and cost extra on Grow. Check before scoping, because a build assuming native calling on a Grow account carries a cost nobody mentioned.

Why does it matter whether leads go through the events endpoint?

Because Follow Up Boss's documentation says so. `POST /v1/people` creates a record but does not run the automations, so no duplicate search, no Lead Flow assignment, no agent notification, no action plan. `POST /v1/events` does all of it. Most broken integrations we are asked to fix write to the wrong endpoint, and the symptom is always the same: leads that exist and were never worked.

Will this replace our ISA?

No, and we decline engagements whose purpose is cutting headcount. It takes the 9pm calls, the retyping, and the follow up that depends on somebody's memory. What is left is the hard part: the conversations, the negotiation, and the judgment a machine should never have been near.

What happens when Follow Up Boss changes something?

Something will: a field, a webhook behaviour, a rate limit. That is why operating a build is a phase and not an afterthought. The failure mode is not an error on a screen, it is a webhook disabled after a bad status code and a week of leads nobody knew were missing.