From signup to activated customer: the split between self-serve and assisted, and an activation check that actually defines what "activated" means.
Customer Onboarding — 9 nodes · 10 connections. Labels on the arrows are the conditions that take each branch.
Most onboarding "processes" are a checklist in someone's head: send the welcome email, maybe book a call, hope the customer sticks. The reason they are hard to automate is not the tooling — it is that nobody has written down where the path splits and what counts as done. "Activated" means something different in every company, and as long as it lives in prose, an AI will substitute its own definition.
This graph forces both decisions into the open. A labelled decision splits customers into self-serve and assisted paths based on plan and seat count, and an evaluator defines activation with checkable criteria — not a feeling, a list. The nurture loop has a stop condition, so no customer receives reminder number seventeen. Open the graph in nodalo, replace the criteria with your own definition of activated, and hand it to your AI to build the sequences or draft the playbook.
Each node carries a contract — what comes in, what happens, what goes out. This is the text your AI receives, node for node, when you hand the graph over.
Entry point. A new account is created — through the website, a sales-led contract, or an invite. Capture: company, plan, seat count, signup source, and the primary contact. This record is the single input every later step reads from.
Within one hour of signup, send the welcome message: one concrete first step the customer can finish in five minutes — not a feature tour, not a documentation link list. One action, one outcome. Everything else waits until the customer has done one real thing with the product.
Gather what the team needs to serve this customer: intended use case, team size, existing tools they migrate from, and the outcome they bought the product for. Collect it inside the product or in the kickoff call — never as a ten-field form before first value. Missing answers are recorded as unknown, not invented.
Split the path. Assisted onboarding for accounts on a paid team plan or with 10+ seats: they get a human. Self-serve for everyone else: they get the guided product path. The branches are exclusive and exhaustive — if neither rule matches, default to self-serve and flag the account for review of the rule itself.
The self-serve route: an in-product sequence that walks the customer to their first complete result — setup, first real artefact, first share or export. Each step unlocks the next; no step is longer than a few minutes. Progress is tracked per account so the activation check can read it.
The assisted route: a 30-minute call within the first week. Agenda is fixed: confirm the outcome they bought for, set up their first real use case live during the call, and agree the next milestone with a date. The call ends with the customer having produced something, not having seen something.
The definition of done for onboarding, checked 14 days after signup. Criteria: the account has created its first real artefact, returned on at least two separate days, and the primary contact has completed the working-context profile. All three pass → the customer is onboarded. Any criterion fails → one pass through the nurture sequence, then check again. After two failed checks, stop looping and hand the account to a human with the history.
Triggered only by a failed activation check, and targeted at the specific criterion that failed: no artefact yet → send the one-step guide for their stated use case; no return visit → send one concrete reason to come back; profile incomplete → ask the one missing question. One message per failure, never a generic drip campaign.
The result: an account marked activated, with its working context, chosen path, milestone dates and activation evidence on record. Ownership moves from onboarding to regular customer success. The record is the input for renewal and expansion conversations later.
Edit the decision node's contract: the example splits on plan tier and seat count, but the branch labels are yours to rewrite — deal size, industry, region, whatever your team actually uses. The only rule the graph enforces is that the branches stay labelled and exhaustive.
Whatever behaviour predicts that a customer stays. The template uses "created their first graph within 14 days and returned at least twice" as a placeholder pattern; replace it with your own — first project created, first invite sent, first report exported. The evaluator only works if the criteria are checkable.
Both. Handed to an AI with the right assignment, this graph becomes a working sequence (emails, checks, task creation in your CRM), a step-by-step playbook for your CS team, or a critique of where your current onboarding leaks — depending on what you ask for at handoff.
Open the template, adjust the contracts, and hand the graph to your AI — as a specification it can execute, not prose it has to decode.