Build and launch

Workflow Reviewed · not yet fact-checked

At a glance

Goal

A working AI SDR loop on Reply: Jason AI derives your business context from a website, files, or pasted text, generates offers and a multichannel sequence, sources its own contacts on a schedule, personalizes every message, answers replies from a knowledge base, and books meetings — with a human approving each message, or fully automatic.

When to use this workflow

• You want the whole loop — find, write, send, respond, book — generated and run inside one system, with approval gates where you choose.
• List building is the bottleneck: Autopilot and intent signals keep the sequence fed daily.
• For hand-written sequences with full manual control, start from Launch an outbound campaign instead.

What Reply provides

• Business-context generation from a website URL, uploaded files, or pasted text — company name, description, ICP, pain points, value propositions, proof points, case studies, and CTAs, all editable
• 5 auto-generated offers per sequence creation — different angles by ICP and pain points — plus playbooks for tone, structure, and do's and don'ts
• A generated multichannel sequence with a swappable structure, personalized per contact at send time
• Contact sources: Sales Navigator Autopilot filters, buying-intent signals, LinkedIn post engagers, competitor followers, AI web search, website visitors, and manual methods — every contact tagged with its source
• Per-sequence ICP and LinkedIn Activity scores with hover explanations, usable as filters
• Approval mode with a Pending approvals queue, and reply handling grounded in a knowledge base
• Meeting scheduling against your connected calendar

What you must provide

• The inputs: your website, files, or text — generation amplifies inputs, it does not invent credibility
• Connected sending accounts for the generated channels (email, LinkedIn)
• The qualification bar: approval-mode decisions, ICP-score thresholds, which reply types stay human
• Knowledge base substance for reply handling, and a calendar if Reply schedules meetings

Preconditions

A workspace on an AI SDR plan; a website URL, files, or pasted text to generate from; connected email accounts and a LinkedIn account for the channels you will use; for automatic meeting booking, a calendar linked under Settings → Meetings.

Entities involved

Business inputs → offer (+ playbook, knowledge base) → sequence → contact sources → scored contacts → conversations → meetings and reports.

Flow

Step 1  Business information   → offer, playbook, knowledge base, message settings
Step 2  Contact sources        → Autopilot filters, intent signals, manual methods
Step 3  Personalization        → personalization points for per-contact research
Step 4  Review outreach steps  → generated sequence; edit steps or swap structure
Step 5  Configure settings     → sending mode, reply handling, meeting scheduling
Launch  → approval queue (or automatic sending) → replies handled → meetings booked

Step-by-step

1. Create the sequence. Go to AI SDR → Jason AI SDR, or click New sequence and select Jason AI SDR.
2. Add business information (Step 1). Paste your website URL, upload files such as pitch decks or product documentation (several PDF or text files at once, up to 100 MB), or paste plain text — Jason converts it into a formal offer. The system generates your company name, description, ICP, pain points, value propositions, proof points, case studies, and CTAs; edit any of them or add more with the + button. You can also choose which AI model analyzes your website — for non-English output the system recommends a different model.
3. Attach guidance objects. A playbook shapes how messages are written — tone, style, do's and don'ts, step-specific instructions, subject-line rules; it cannot create the sequence structure, run LinkedIn actions, filter the audience, or manage replies. A knowledge base holds the product content Jason answers replies from. An offer targets a segment: Jason generates 5 offers per sequence creation (one shown in the wizard, four more appear on the Offers page) and pre-selects the first. If you edit offer inputs mid-wizard, choose Update original (applies to every sequence using the offer) or Keep separate (a new offer for this sequence only).
4. Choose message settings. Language, tone of voice, message length (default Short), and — for languages whose grammar depends on it — the sender's gender.
5. Define contact sources (Step 2). Add one or more sources; each both feeds the pipeline and tags its contacts, visible in the contact's activity log and the sequence Logs. Options: Sales Navigator Autopilot (ICP filters pre-filled from Step 1, or a pasted Sales Navigator search URL — unsupported filters are skipped; max 500 contacts per day, default 10, 3 per company), Hiring Intent or Technology Used (one at a time; one-time searches of up to 500 contacts, re-runnable after launch), LinkedIn Post Engagers (up to 50 likers and 50 commenters per post; ICP fit defaults to 60%), Competitor Followers (maximum 500 contacts from all competitor pages combined, scanned daily for new followers), AI Web Search (maximum 500 contacts per search), Website Visitors, and Manual Methods (CSV import, add from existing, create manually, integration, LinkedIn search with your connected account).
6. Select personalization options (Step 3). Personalization points define what Jason researches before writing — company details, the prospect's role or achievements, relevant industry context — so messages go beyond basic contact fields.
7. Review outreach steps (Step 4). Jason generates the sequence and picks a structure suited to your inputs. Steps contain placeholder text that is personalized per contact when the step is scheduled. Edit, remove, or add steps; browse alternative structures in the picker (Default, My, Workspace, Organization scopes) or save your own layout as a structure template (name up to 50 characters).
8. Configure settings (Step 5). Sending mode: Approval mode (every message waits on the Pending approvals page) or Automatic. Enable Handle replies if Jason should respond to incoming messages — automatically or as drafts for review — and decide whether follow-ups to silent leads send automatically or as drafts. Choose meeting scheduling: manual, Reply (link a calendar under Settings → Meetings; Jason checks availability, suggests times, and answers meeting-related replies), or another tool.
9. Launch and work the approval queue. On AI SDR → Pending approvals, sort by newest or oldest and filter by channel (Email or LinkedIn), sequence, or owner. Edit a message inline, click Regenerate with an optional short prompt, approve one by one or with Approve selected in bulk — or delete it, which removes the prospect from the sequence. Approved messages send according to the sequence's schedule and limits.
10. Iterate. Return to Jason AI → Settings to adjust sending mode, AI model, tone of voice, playbook, personalization points, language, or business information — changes apply only to contacts who have not yet received any steps. In the sequence's People tab, filter by ICP score (for example, remove contacts scoring under 60) and lead with LinkedIn for contacts whose LinkedIn Activity score is Frequent.

Important rules

Contact enrollment · Sequence execution · Reply detection · Sequence stopping · When Reply sends

Side effects

Contact sources keep fetching on their schedule until paused — Autopilot daily, competitor pages scanned once a day. Intent searches start when you finish the sequence setup, not when you save. Automatic mode sends without review. Deleting a pending approval removes that prospect from the sequence permanently.

Failure conditions

• Hiring Intent and Technology Used unavailable — both require the Accounts feature enabled, since they build an account list first
• LinkedIn search grayed out — it runs through your own LinkedIn account, which must be connected
• Nothing sends in approval mode — an unworked queue holds all outreach; review it daily or switch modes
• Sequence toggle red or orange — see Why is a campaign not running?

Verification

After launch: contacts appear in the People tab with source tags in the Logs; ICP and LinkedIn Activity columns fill in; generated messages land on Pending approvals (approval mode) or send inside the schedule window (automatic); replies receive drafts or answers according to each handler's delivery mode.

Safe stopping

Pause Autopilot from the People tab toggle without touching the sequence; switch automatic mode to approval mode to put a human back on the send button; or pause the whole sequence — sending stops, state is kept. See Sequence stopping.

Limits

• Autopilot: max 500 contacts per day per sequence; default 10 per day, 3 contacts per company
• Hiring Intent / Technology Used: one active signal at a time; up to 500 contacts per search; re-runnable after launch
• LinkedIn post engagers: up to 50 likers plus 50 commenters per post; ICP fit threshold defaults to 60%
• Competitor followers: maximum 500 contacts from all competitor pages combined
• AI Web Search: maximum 500 contacts per search
• Business-info uploads: several PDF or text files at once, up to 100 MB; knowledge base documents: PDF, DOC, DOCX up to 100 MB per file, other types rejected; links added one at a time
• Structure template name: up to 50 characters
• Email address checks aim for up to 98% accuracy
• Settings changes apply only to contacts who have not yet received a step

Related

• AI SDR capability — what Jason AI covers, end to end
• The AI SDR model — how the pieces fit conceptually
• Offers · Playbooks · Knowledge bases
• Find prospects from an ICP — the sourcing slice in detail
• Process positive replies — the reply slice in detail

Build with Reply

• REST API: docs.reply.io — ai-sdr endpoint group (check operation status)
• MCP: agents.reply.io/mcp — offers, playbooks, knowledge bases, approvals, and sequence generation
• CLI: agents.reply.io/cli — scripting and CI-friendly control
• Agent skills: agents.reply.io/skills — packaged outreach workflows
• Agent workflow guides: agents.reply.io/workflows