Jason AI objects
Specification Reviewed · not yet fact-checked
In short
What is a playbook in Reply and how does it control the messages Jason AI writes?
A playbook is a prompt document of instructions — Do and Don't lists, subject-line rules, and step-specific sections — that controls how Jason AI writes sequence messages. It shapes tone, style, length, and structure but cannot build sequences, pick audiences, or manage replies. Jason can also generate a playbook from your offer, tagged AI Generated.
Playbook (Jason AI object) vs Playbooks (site section). This page specifies the Jason AI configuration object. For human-readable outreach recipes, see Playbooks.
At a glance
| Aspect | Detail |
|---|---|
| What it is | Your guide for how AI SDR generates sequence messages — custom prompt instructions so the content matches your style, sales goals, and strategy |
| Belongs to / Contains | Attached to AI SDR sequences; organized under the Library (ready-made examples) and Team folder tabs; can carry step attachments |
| Surfaces | UI · REST API · MCP |
| Key limit | Step attachments: 3 MB per file, 4 MB total per email step type, 1 file per LinkedIn step |
| Lifecycle | Written manually or generated by Jason from your offer (tagged AI Generated); your own playbooks are editable any time, Library defaults are not; both can be duplicated |
| Does not | Create sequences, control LinkedIn actions, research contacts, or filter who to reach out to — and once a prospect replies, playbook instructions no longer apply |
Definition
A playbook is your own guide for how AI SDR generates sequence messages. Its custom prompts — what to highlight, what to leave out, do's and don'ts, and how to organize the messaging — give you control over how the AI creates outreach messages, so the content matches your style, sales goals, and strategy. A playbook does not have to be long: it can be as simple as one sentence, such as "Don't mention the competitors' names — be polite and avoid hard selling."
Purpose
Playbooks are for message creation, not sequence setup. They let you control tone, style, and length, set subject-line rules, apply variables and personalization points, guide the structure of specific steps, specify when to include links, and keep shaping follow-ups in the right tone until a prospect responds. The best way to control Jason's tone of voice is via a playbook.
Relationships
| Relationship | Object |
|---|---|
| Attached to | AI SDR sequences — click New AI SDR Sequence above the playbook, or Add playbook while setting up a sequence |
| Generated from | Offers — the meta playbook turns the offer's pain points, value propositions, and ICP, plus your uploaded examples, into a tailored playbook |
| Complemented by | Knowledge bases — once a prospect replies, playbook instructions no longer apply; reply handling is configured separately |
| Organized in | The Playbooks list folder tabs: Library (ready-made examples) and Team (your team's playbooks, including AI-generated ones) |
Creation
Writing a playbook manually
- Go to the AI SDR page and open the Playbooks tab.
- Click New Playbook in the top-right corner.
- Structure it with the section buttons for key outreach steps: Do, Don't, First email, First email follow-up, Second email follow-up, Subject Lines, LinkedIn Connect, First LinkedIn Message, Voice Message, LinkedIn InMail, Call script, Variables. The default template includes pre-filled text showing how to structure a playbook.
The Library tab holds ready-made playbooks to use or refer to as examples — review and adjust them to fit your audience, goals, and sequence before using. Library defaults cannot be modified; your own playbooks can be edited at any time.
AI-generated playbooks (meta playbook)
Jason can generate a playbook for you. A system-level meta prompt — the meta playbook — takes your offer's data (pain points, value propositions, ICP, and the wizard's step structure) plus your own outreach examples and generates a custom playbook tailored to that product and selling context:
- Generation runs in the background while you complete the earlier wizard steps — when you reach the Playbook step, the playbook is already present (or loading). If generation is not ready or fails, you can generate one or browse the library instead; the wizard offers two view modes, AI Generated and Browse Library.
- Clicking Continue saves the result: the playbook is persisted to your library under the Team tab, tagged AI Generated, named after the company, and attached to the sequence.
- On a return visit, the existing AI-generated playbook is pre-selected; you can accept it, pick another from the library, or generate a new one — the old version is not deleted.
- Feedback chips and a free-text field drive Regenerate; manual edits to the Do / Don't columns are preserved, and regeneration builds on top of them. Like and dislike reactions are stored to improve future generations.
- To train style, upload past outreach examples: up to 20 files of 100 MB each, in
.txt,.html,.csv,.md, or.pdfformat. The AI-generated playbook's title is capped at 120 characters.
Duplicating
Both custom and default (Community) playbooks can be duplicated from the Playbooks list; the copy opens in a saved state.
Configuration
What a playbook controls
- Tone, style, and length — friendly, formal, short, or detailed.
- Subject line rules — length, style, and personalization across the sequence.
- Personalization — standard or custom variables like
{{FirstName}}or{{Company}}, plus personalization points relevant to your audience. - Step templates — a preferred structure for specific steps (for example, Email #1 or a follow-up email).
- Links — when and how to include hyperlinks to calendars, websites, or resources.
- Follow-ups — the AI keeps shaping follow-ups in the right tone until a prospect responds.
What a playbook cannot do
- Create entire outreach sequences — it controls message content, not sequence structure.
- Control LinkedIn actions — the AI cannot open profiles, send connection requests, or automate other LinkedIn actions; LinkedIn steps belong to the sequence.
- Research contacts — include variables and personalization points instead.
- Filter who to reach out to — audience selection happens beforehand; a playbook cannot exclude prospects by job title or location.
- Manage replies — once a prospect replies, playbook instructions no longer apply; set up reply handling separately in the knowledge base.
Structure of a well-built playbook
A playbook has two main parts:
- General instructions for all steps — overall guidelines (tone, style, length, approach), Do / Don't lists, and subject-line rules. Put anything that is not a template into the Do's and Don'ts sections, keep lists under 10 items, and start Don'ts with "Don't".
- Step-specific instructions — AI-friendly step names ("first email", "email follow-up #1" — not "2nd email"), step-specific Do / Don't lists, frameworks and examples with variables, and per-step subject-line instructions where needed.
Variables and hyperlinks
Use {{double braces}} for standard Reply contact variables and [brackets] for longer instructions, like [Your Value Proposition]. Add hyperlinks with the Insert link button in the editor's tools panel, by pasting a URL directly into the editor, or with a manual HTML anchor tag — the AI shapes it as a clickable link and keeps hyperlinks intact in generated emails.
Multi-language
Playbooks support multiple languages: write the instructions in English or another language, then tell the AI which language the messages should be generated in. For best results, include message examples in the target language.
Step attachments
Files — lead magnets, case studies, documents — can be attached to playbook steps and sent as part of the AI SDR sequence. One file can be assigned to multiple steps at once, and attachments are included in AI-generated personalized messages automatically.
| Step | Attachments |
|---|---|
| First Email | Supported |
| First Email Follow-Up | Supported |
| Second Email Follow-Up | Supported |
| First LinkedIn Message | Supported |
| LinkedIn Connect | Not available |
| Voice Message | Not available |
| LinkedIn InMail | Not available |
| LinkedIn Voice Message | Not available |
Supported file types: .pdf, .doc, .docx, .txt, .ppt, .pptx, .xls, .xlsx, .csv, .mp3, .wav, .m4a, .ogg, .aac, .webm, .jpg, .jpeg, .png, .gif, .bmp, .webp.
Processing behavior
Jason applies the playbook's instructions each time it generates a message, and keeps generating follow-ups in the configured tone until the prospect replies. In an already-running sequence, changes to the playbook and other Jason AI settings apply only to contacts who haven't yet received any steps. When a sequence has multiple playbooks assigned, only the first playbook's attachments are used.
Editing a saved playbook
Editing a playbook that is attached to one or more sequences first asks how to apply the changes:
- Update for all sequences — updates the playbook in place; every sequence using it gets the new version.
- Save as a new playbook for this sequence only — creates a copy, attaches it to the current sequence, and leaves the original (and the other sequences using it) unchanged.
A playbook not attached to any sequence opens straight in the editor. Your own playbooks are editable at any time; Library defaults are not.
Limits
| Limit | Value |
|---|---|
| Attachment upload | 3 MB per file |
| Attachments per email step type | 4 MB total across all files on that step |
| Attachments per LinkedIn step | 1 file |
| Example files for AI generation | Up to 20 files, 100 MB each (.txt, .html, .csv, .md, .pdf) |
| AI-generated playbook title | 120 characters |
Attachment limits are enforced per step type, not per email provider. The documentation does not state a cap on the number of playbooks.
Edge cases
- Some playbook features — step attachments and editor buttons such as Insert link — are rolled out per workspace and may not appear in every account.
- When a sequence has several playbooks assigned, only the first playbook's attachments are used.
- Only the first LinkedIn Message step receives attachments — a second LinkedIn Message step in the same sequence does not.
- LinkedIn Connect, InMail, and voice-message steps cannot carry attachments — LinkedIn supports attachments only on regular LinkedIn Message steps.
- Use AI-friendly step names in instructions: "email follow-up #1" applies correctly, "2nd email" may not.
- After several regenerations of an AI-generated playbook, navigating back in the wizard shows the last generated version, not the original.
FAQ
Can Jason AI generate a playbook for me?
Yes. During sequence creation, a playbook is generated in the background from your offer's pain points, value propositions, and ICP, plus any outreach examples you upload. It is saved to your library under the Team tab, tagged AI Generated, and named after the company. You can regenerate it with feedback or swap it for a library playbook.
Does editing a playbook change sequences already using it?
You choose: Update for all sequences applies the edit everywhere the playbook is used, while Save as a new playbook for this sequence only isolates the change to the current sequence. In a running sequence, changes apply only to contacts who haven't yet received any steps.
Can a playbook control who gets contacted?
No. Audience selection happens before generation — a playbook cannot filter prospects by job title or location, and it cannot control LinkedIn actions or research contacts. See Build an AI SDR for where audience selection happens.
Can I attach files for Jason to send?
Yes — attach lead magnets, case studies, or documents to email steps and the first LinkedIn Message step: 3 MB per file, 4 MB total per email step type, one file per LinkedIn step. Connection requests, InMail, and voice messages cannot carry attachments.
Do playbooks work in languages other than English?
Yes. Write the instructions in English or another language and tell the AI which language the messages should be generated in; include example messages in the target language for best results.
Related
Build with Reply
- REST API: docs.reply.io — playbook operations
- MCP: agents.reply.io/mcp — create, duplicate, and attach playbooks from an agent