How do I set up a pipeline for how my customers move through my business?

A pipeline is a named path of steps a contact moves through. You build one in the journey editor: add its steps in order, then give each step the jobs your team (or your customer) does when a contact reaches it. The journey editor is the one place a pipeline's steps and jobs change, for every pipeline you have.

Create a pipeline

1. Open TMS, then Setup, then the Pipelines tab.
2. Click New pipeline, give it a name, and click Create pipeline. The journey editor opens on it straight away, ready for its first steps.
3. If you want this to be the pipeline new contacts go into by default, click Set as default on its tab. Your default pipeline is your customer journey, so every new contact with a phone number or email gets its tasks. While customers are on your current default you can't choose a different one: finish or move them first.

Find your way around the Pipelines tab

Each pipeline has its own tab along the top. A pipeline for one of your services is named after that service, and your default pipeline is called Main journey.

Inside a tab, every step starts folded to one line: its name, how many jobs it has, and how many people are in it right now. Click a step to see its jobs and settings, or use Expand all and Collapse all to open or fold every step in that tab at once.

The tab only shows your steps. To change them, click Edit journey. Test run shows what a customer would get from the steps as they are published today. The pencil beside the pipeline's name renames it or changes which service it is for, and the bin deletes the whole pipeline.

Add a step

1. On the pipeline's tab, click Edit journey. On a pipeline with no steps yet, click "Add the first steps" instead.
2. Click "Add a step" at the bottom, or "Add a step below" under any step, and give it a name.
3. Give it at least one job: see "Add jobs to a step" below.
4. Open "More options for this step" to set its category (a short label for what kind of step it is, like intake or closing), badge colour, icon, default due time and priority, and two numbers that control how a contact is treated over time: the aging weight (how heavily this step counts when we work out how "aged" a contact looks; leave it at 1 unless you want this step weighted more or less than the others) and the dormancy threshold (how many attempts in a row to reach a contact can fail before they're flagged as gone quiet, up to 50). You can also set a default snooze and how many times the step repeats.
5. Click Publish. Your edits are saved as a draft as you go, and nothing changes for anyone until you publish.

A pipeline needs at least 3 steps, with at least one lead step and one customer step, before it can be published. The editor tells you what is missing above the Publish button.

Which step means they have become a customer

The "Start customers here" line in the editor decides this. Every step above the line is a lead step, while you are still winning the work. The step the line sits on, and every step below it, is a customer step. Move the line by clicking "Start customers here" on the first step that means the job is won, then publish.

When a contact is moved into a customer step, they are marked as a customer and the date is recorded, once. It only ever goes forward: moving them back to an earlier step does not undo it. The very first step in a pipeline never marks anyone as a customer, whichever side it is on, because that is simply where contacts start. Because it is one line, a lead step can never sit below a customer step.

Repeat a step

Some work happens in rounds: 12 coaching sessions, 6 monthly check-ins, 4 fittings. Instead of building the same step 12 times, build it once and put 12 in "Repeat this step (times)" under the step's More options. The most is 52.

A customer then works through that step's jobs 12 times, one session after another. A session's jobs are created when the customer reaches it: session 2 starts once every job in session 1 is done or skipped, and the next step only starts once the last session is finished. Every job in every session is a real job, so a step of 5 jobs repeated 12 times is 60 jobs, and each one counts in your team's numbers when it's done.

On a customer's Journey tab a repeating step shows as one row with a counter such as "Session 3 of 12", and opening it shows that session's jobs rather than all 60. The Tasks tab labels each job with its session the same way. If you show the step's jobs to the customer, they see each one with its session number, like "Weekly call (3 of 12)".

A change to the repeat count reaches every customer who has not got that far yet, as soon as you publish, because sessions still to come are read from your pipeline at the moment they start. The session a customer is working on now is never changed. To stop a step repeating, clear the box and publish.

Skip a job for one customer

Staff can switch a job off for one customer without changing your pipeline for anyone else. On the customer's Journey tab, press Edit journey, then Skip next to the job. This works on a job they are working on now and on one they have not reached yet, which shows as Coming up: dimmed, with only a Skip button, because there is no task to open yet for a stage the customer has not reached. The "N of M done" count at the top of the tab counts every step still to come the same way, not only the ones already created, and a stage where every step is still Coming up shows that label in place of a done count.

• A skipped job is never created for that customer. When its stage starts, the stage's other jobs are created without it, and a stage where every job is skipped is passed over.
• Skipping the last unfinished job of the stage a customer is in moves them on to the next stage straight away, the same as finishing it would.
• On a repeating job the button reads "Skip every session": every session of that job still to come is skipped for that customer, and sessions they have already done stay done.
• A job that is already done cannot be skipped, and neither can a step the customer has already moved past.
• Each skip shows on the customer's Timeline with who did it.

Add jobs to a step

In the editor, click "Add a job to this step" to add the actual job a staff member does when a contact reaches it. Click a job to open it: "The job" is its name, "What to do" is the instruction your team (and the AI) works from, and you pick its priority and how long it's given under "Due in". A step can have up to 5 jobs. This is what turns "a contact is at this step" into "someone on your team has something concrete to do."

A step's jobs are always created together, the moment a contact reaches that step, so your team works them side by side rather than one unlocking another. If you want one job to genuinely wait for another, put it in a later step instead: the next step's jobs are never created until every job of this one is done or skipped. If you add a new job to a step a customer is already working through, they get it too, but only once they finish every job the step already had; put the new job earlier in the list than jobs they were already given and they never get it at all, since they've already moved past that point.

One switch on each job decides what your customer sees: "Customer sees this". Switch it off and the job is for your team only: the customer never sees it on their status link or in the portal, and never gets a reminder about it. Switch it on for the steps you are happy for them to follow. Each job's one-line summary says "hidden from customer" when it is off, so you can check at a glance. If nothing is switched on anywhere, anyone who opens their status link gets an empty page, and the Journey tab says so.

The same "Add a job to this step" menu also adds jobs that ask the customer for something: documents, a payment or a signature. See How do I set up my customer journey? for those.

Who does this step

Every job also has a "Who does this" choice, with three options: Person, AI drafts, person checks, or AI does it. Person is the default for every job. With "AI does it", the AI finishes the job on its own, with nobody reviewing it first. See How do I set up my customer journey? for what the second and third options actually do once a job is due.

Once you pick either AI option, the job's More options also shows "Let the AI look things up on the web for this job". It is off unless you turn it on. Turn it on for work that needs facts from outside your business, like a school's current entry requirements or a supplier's published prices. The AI then searches the web while it writes and lists the pages it used under Sources at the end of what it writes, so whoever reviews it can check each one. If the search cannot run, it writes the step without searching instead of leaving it undone.

To have the customer confirm a step, add an "Ask to sign" job. With an agreement attached the customer reads and signs it in the client portal. Without one, the job puts a one-tap Confirm button next to that step on their status link. When they press it we record who confirmed it and when, and you can see it on the step. It does not tick the job off and it does not move the customer to the next step, so your team still finishes it.

An "Ask for documents" job lists the documents this job needs, picked from your own list. Your team sees them on the task itself, with a mark for whatever the customer has already sent, a one-click way to ask for what is still missing, and a way to upload a file straight onto the task if it arrives another way. Pick at least one document: leave every box unticked and the job asks the customer for nothing.

Finish a task by itself when the documents arrive

An "Ask for documents" job has one more switch: "Move on when all are in". It is off unless you turn it on.

1. Tick the documents the job needs, then switch on "Move on when all are in", and publish.
2. From then on, every time a customer sends a file, or your team uploads one or approves one, we check that customer's open jobs of this kind.
3. When every ticked document has at least one file that was not rejected, the job is marked done on its own. The customer's Timeline shows "All requested documents received", and if it was the last job of the step the customer moves on to the next step, the same as if your team had finished it. A timer anyone left running on the job stops too.

A few things never count or never finish a job this way:

• a file you rejected, and a file our virus check stopped;
• a job that asks the customer to confirm it, since that confirmation is its own step;
• a job with an AI draft still waiting for someone to review it.

In a step that repeats, each session's job only counts files sent since that session started, so last month's receipts never finish this month's job. The first session counts anything the customer already sent.

Nobody looks at the files before the job is marked done, so leave this off for any job where a person needs to check them first. Rejecting a file later does not reopen the job: open it again by hand if you need to. An automation on "Task finished" can react to these jobs: pick "All its documents arriving" under Finished by, see How do I send an automatic text, email or task when something happens?

Telling the customer when the AI's work is done

When a job is both shown to the customer and done by the AI (a draft a person approves, or a job the AI does on its own), finishing it sends the customer a short message: one line saying the step is done, and a link to their own page. The work itself is never put in the message. A job your team does by hand, or one the customer cannot see, sends nothing, exactly as before.

The document itself stays with your team. A finished step the customer can see shows them its title, "Done" and the date it finished, the same on their status link and in the portal, nothing more. Your team can still open what the AI prepared on the finished task, the same place a person's own completed work shows.

A job's More options has "Tell the customer by", with three choices:

• Email, the default. A short note with the link.
• Text message. One line with the link. It needs your texting number and the customer's permission to be texted. A text never goes out late at night where they are; it waits until the morning.
• WhatsApp. It needs WhatsApp connected and an approved message template, which you pick right under the choice. Only templates with at least one blank are offered: the last blank gets the link, any earlier ones get the job's name.

If the way you picked cannot reach a customer (no email on file, no permission to text them, no template), we try the other two on our own. A customer who has opted out is never messaged. Only when none of the three can reach them does the job come back to your team, marked "Customer not told", so someone can tell them another way. We also send at most a few of these updates to one customer in an hour, so a journey set up by mistake cannot flood anybody.

Have the jobs suggested for you

If you'd rather not start from a blank box, click "Suggest jobs" on a published step and we'll suggest jobs that step is probably missing. We use what you've told us about your business, what you've written in your knowledge base, the jobs that step already has, and the steps either side of it, so the suggestions fit where they sit rather than repeating the work before or after.

1. Nothing is saved yet. You get a list, each job with one line saying why it belongs at this step.
2. Tick the ones you want and click "Add these jobs". They join the step as jobs you can edit like any other, added at the end of the step's list, so nothing already there is moved or replaced. Nothing is saved until you publish.

A step you added in this draft can't ask for suggestions until it is published. If you haven't told us what your business does yet, we'll say so rather than inventing generic jobs. You can always add a job by hand instead.

Reordering, editing and deleting

In the editor, move a step or a job with its arrows or by dragging it, and remove one with its bin. Changing a step's settings changes what's expected going forward; it doesn't rewrite anything already logged against a contact who passed through it earlier. Nothing changes until you publish, and "Review and publish" lists what you renamed, added and removed first.

A few things are kept when you publish, and the page says so. A step with people standing in it is kept, with how many are in it, and everything else still goes ahead. A job that customers are currently on can't be removed either, and the message tells you how many; wait for them to finish or skip it first, or remove a job nobody has reached yet, which is always allowed. An automation that runs when a customer reaches a step you removed is switched off rather than deleted, and named for you. Deleting a whole pipeline is done with the bin on its tab, and is permanent: a pipeline with contacts still in it can't be deleted until they are moved. It also can't be deleted while an automation is still switched on for one of its steps. Switch that automation off, or point it at another step, under Automations first. An automation that is already switched off never blocks the delete.

Every step in one pipeline needs its own name, so a name another step already has is refused, and upper or lower case doesn't make it a different name.

Who can do this

Setting up pipelines needs an organization admin and an account with Team management turned on. See Who can set up pipelines, journeys and automations for my team?

Give one of your services its own pipeline

When you create a pipeline, or edit one you already have, you can pick which of your own services it's for from a dropdown, or leave it set to "General (no specific service)" the way every pipeline works today. Only one pipeline per service, if you already gave a service a pipeline, that dropdown won't offer it again for a second one. Once a pipeline is tied to a service, drafting its journey and enrolling customers onto it both work the way How do I set up my customer journey? describes. You don't have to create one here first: picking a service on the Journey screen and installing its journey creates its pipeline for you. On the Starter plan you get one pipeline, so there's nothing to tie it to but the one journey you already have, tying a pipeline to a specific service needs Pro and up. A service pipeline you already have keeps working on Starter, and Edit journey still changes its steps.

Each pipeline's tab shows which service it is for. If you remove that service from your services list (Settings, Services), the tab says Service removed; you can still rename the pipeline, then pick another service or General.