A Journey connects Sources, AI Agents, Filters, branches and Destinations into one flow. You describe the steps in order, Claude previews each one with you, and nothing is saved until you approve the full step graph.
Before you build a Journey, create the Sources, Personas, AI Agents and Destinations it uses. See the related articles at the end of this one.
The base prompt
Build a Tapistro journey named "[your name for this journey]".
Goal: [one or two sentences on what goes in and what comes out].
Schedule: manual / daily at [time] UTC, max [N] records per run / hourly.
Add it to the play "[play name]".
Steps, in order:
1. [Step]
2. [Step]
3. Branch "[name]" ([account / person] level):
- If [condition A], go to step 4
- If [condition B], go to step 6
- Otherwise, go to step 8
4. ...
Preview every step before you add it to the journey. Show me the preview results for each step and wait for my OK before moving to the next one.
Once every step is approved, show me the step graph as a numbered list with every branch, and wait for my OK before saving. Don't run it yet.
Note: Name your steps, especially Filters and branches. "Filter: B2B only" is much easier to read in a Step Report than "Data Filter."
Step wording
Use this wording for each step in your prompt.
Step | Say this |
Source Ingest | "Ingest from source [name]." Give each Source its own ingest step. |
Source Discovery | "Source Discovery," right after each ingest step. It matches records to accounts and people. |
Account enrichment | "Enrich accounts with [Apollo / Explorium]." |
Person enrichment | "Enrich people with [Wiza, ContactOut, Apollo]. Required fields: first name, last name, title." |
Find people | "Find people with persona [name] using [ContactOut, Explorium, Apollo]. Validate work emails with NeverBounce." |
Person count | "Count people per account with titles [list] who started in the last [N] days. Save it as '[field name]' and refresh every 14 days." |
Signals from a Source | "Add signals from source [name] to each account." |
Network overlap | "Check LinkedIn network overlap for each [account / person]." |
Email validation | "Validate work emails with [Hunter / NeverBounce]." |
Phone | "Find phone numbers," "Validate phone numbers," or "Check the US Do Not Call registry." |
REST API | "Call this REST API: [details]." |
AI Agents | "Run agents [names]. Run their dependencies first. Generate the full email sequence." |
Filter | "Filter ([account / person] level) '[step name]': keep records where [condition]." |
Branch | "Branch '[name]': if [A], go to... if [B], go to... otherwise..." |
Top N | "Keep the top [N] by [field]." |
Distribute | "Split across the team: [Rep A 50%, Rep B 50%]." |
Assign field | "Stamp [stamp agent] with '[value]'." |
Wait | "Wait [N] days." |
Merge | "Merge the paths from steps [X] and [Y]." |
Destination | "Send to destination [name]." |
Conditions in plain English
Describe Filter and branch conditions in plain English. Claude turns them into checks.
Say this | What Claude checks |
"Exclude accounts on the HubSpot lists [customers list] and [competitors list]" | The account isn't a member of either list |
"Valid emails only" | Email status is clean, unknown, or not yet checked |
"Only accounts where the agent [name] returned [value]" | An AI Agent's output |
"51 to 5,000 employees" | Employee size range from the best available source |
"HQ in the US or Canada" | HQ country from the best available source |
"Score of 80 or higher" | A Calculate Agent's number compared to a threshold |
Branch rules
When a branch has tiers, list the strictest condition first. For example: 80 and up, then 60 and up, then 40 and up.
Always end with a catch-all path, such as "everyone else," so each record lands in exactly one path.
When a person-level Journey branches on account data, add "(account level)" to the branch.
Preview each step
Previewing step by step catches problems early, such as a Source that returns the wrong companies or an Agent that misclassifies. How Claude previews depends on the step:
Step | Ask Claude to |
Account search Source | "Preview 10 accounts from this source." |
Persona (Find people) | "Preview the persona on [2 or 3 sample domains]." |
AI Agent | "Preview the agent on [2 or 3 sample accounts or people] and show me the output." |
Filter or branch | "Show me which of the sample records would pass each condition, and why." |
Destination | "Show me exactly which fields and values would be sent for 2 sample records." |
Test the Journey on a small batch
Before you turn on a schedule, run the whole Journey on a few records:
Run the journey "[name]" now on up to 10 records, then show me how many records entered and passed each step.
Check results
Show me how many records entered and passed each step of "[name]" in the last run.
You can also open the Journey's Step Reports in Tapistro. See Use Step Reports to Track and Troubleshoot Journey Runs.
Worked example: qualify an account list and sync it to HubSpot
This Journey takes an uploaded account list, enriches and filters it, adds research, and syncs target-industry accounts to HubSpot. Competitors exit early to a HubSpot suppression list, and every other account goes to a nurture list.
Build a Tapistro journey named "[your name for this journey]".
Goal: Take my uploaded account list, enrich and qualify it, research the qualified accounts, and push the results to HubSpot. Send competitors to a HubSpot suppression list.
Schedule: manual.
1. Ingest from source "[your account list source]".
2. Source Discovery.
3. Enrich accounts with Apollo.
4. Filter (account) "Exclude customers": exclude accounts on the HubSpot list [customers list].
5. Run agents: [B2B classifier agent], [company size agent], [industry agent], and [competitor check agent].
6. Branch "Competitor check" (account level):
- If [competitor check agent] returns Competitor, go to step 7
- Otherwise, go to step 8
7. Send to destination "[your HubSpot suppression destination]". Stop here.
8. Filter (account) "B2B, 50+ employees": [B2B classifier agent] returns B2B, and [company size agent] returns SMB, Midmarket, or Enterprise.
9. Run agents: [target ICP research agent], [buying signals research agent], and [account summary research agent].
10. Branch "Industry fit" (account level):
- If [industry agent] returns [your target industries], go to step 11
- Otherwise, go to step 12
11. Send to destination "[your HubSpot account sync with research fields mapped]".
12. Send to destination "[your nurture list destination]".
Preview every step before you add it to the journey. Show me the preview results for each step and wait for my OK before moving to the next one.
Once every step is approved, show me the step graph as a numbered list with every branch, and wait for my OK before saving. Don't run it yet.