Can a Small Business Owner Use AI to Write a Monthly Newsletter and Still Approve Every Word?

Summary

The steps

1. 1

   Create the drafting agent and the review agent

   This workflow uses two separate agents: one to draft the newsletter and one to check the draft against the facts the owner supplied. Creating both before building the canvas keeps the workflow configuration clean.

   Navigate to Agents in the sidebar. You need the Admin or Create role to see this section. If the section is not visible, see Why Can't I See Agents, Workflows, or Other Sections in Dave by voolama? to diagnose your role assignment.

   Drafting agent: Click New Agent. Name it "Newsletter Drafter". Click Create Agent, then click Create Version on the detail page. In the Config JSON editor, write a system prompt that tells the agent to write in the business's voice and draw only from the facts in the workflow context. A starting config:

   {"systemPrompt": "You are writing the monthly email newsletter for [Business Name]. Write in a warm, direct tone that sounds like the owner speaking to regular customers. Use only the facts, announcements, and offers provided in the input. Do not add claims, statistics, or details that were not supplied. Structure the newsletter with a short greeting, one paragraph per news item, and a brief closing. Keep the total length under 350 words.", "model": "gpt-4o", "temperature": 0.4, "maxTokens": 800}

   Review agent: Click New Agent again. Name it "Newsletter Fact Checker". Create a version with a system prompt that instructs the agent to compare the draft against the original news input and return a clear pass or fail:

   {"systemPrompt": "You are a fact-checking agent. You will receive two inputs: the original news brief the owner provided, and the newsletter draft an AI wrote from it. Check whether every factual claim in the draft can be traced to something in the original brief. Return PASS if all claims are supported. Return FAIL followed by a one-sentence explanation if any claim in the draft is not supported by the brief.", "model": "gpt-4o", "temperature": 0.1, "maxTokens": 200}

   The Config JSON fields Dave by voolama agents accept are: systemPrompt, model, temperature (0.0 deterministic to 1.0 most varied), maxTokens, and topP. Note the version number assigned to each agent (v1 for the first version of each).

   Do this

   • Go to Agents in the sidebar (Admin or Create role required).
   • Click New Agent, name it "Newsletter Drafter", click Create Agent.
   • On the detail page, click Create Version. Paste the drafting Config JSON with your business name, voice instructions, and model name. Click Create Version. Note the version number (v1).
   • Click Back to Agents. Click New Agent, name it "Newsletter Fact Checker", click Create Agent.
   • Click Create Version. Paste the fact-checking Config JSON with temperature set to 0.1. Click Create Version. Note the version number (v1).

   Example

   A hypothetical garden center might set the drafting agent's system prompt to: "Write in the voice of the owner, who is friendly and knows her customers by name. Use only the facts in the brief. Do not add planting tips, seasonal advice, or product claims that were not in the brief." The fact-checking agent's system prompt stays generic: it only needs to compare the draft to the brief, not understand gardening.

   Best practice

   • Keep the drafting agent's system prompt specific to your business's voice. Name the tone, the audience, and any phrases the business always or never uses.
   • Set the fact-checking agent's temperature to 0.1 or lower. You want a consistent, deterministic PASS or FAIL result, not creative variation.
   • Name agents after their function, not their model. "Newsletter Drafter" is easier to find in a workflow than "GPT-4o Agent 2".
   • If you update your newsletter style, create a new version of the drafting agent rather than editing the existing one. Retiring is a one-way action in Dave by voolama: you cannot un-retire a version, but existing running instances continue using the old version until they complete.
2. 2

   Create the workflow and open the Workflow Editor

   Navigate to Workflows in the sidebar. You need the Admin or Curate role to see this section. Click New Workflow, name it "Monthly Customer Newsletter", and save. The Workflow Editor opens: a three-panel canvas with the Node Palette on the left, the canvas in the center, and the Properties Panel on the right.

   Dave by voolama's Workflow Editor uses a ReactFlow canvas. A node and edge counter in the top-right corner shows the current count as you build. Each click of Save Version creates a new immutable snapshot of the graph. Previous versions are preserved, so you can iterate without losing earlier designs.

   Drag a Start node from the Node Palette onto the canvas. Every workflow requires exactly one Start node. Click Save Version to create your first snapshot.

   The ten node types available in the Node Palette are: Start, End, User Interaction, Human Review, Agent Interaction, Agent Review, Routing, API Call, Information, and Safety. This workflow uses Start, User Interaction, Agent Interaction, Agent Review, Routing, Human Review, and End.

   Do this

   • Go to Workflows in the sidebar (Admin or Curate role required).
   • Click New Workflow, name it "Monthly Customer Newsletter", save.
   • Drag a Start node from the Node Palette onto the canvas.
   • Click Save Version to create your first snapshot.

   Example

   A hypothetical accounting practice might name the workflow "Client Newsletter" and add a description: "Owner types in the month's news, AI drafts the newsletter, a second AI checks the facts, and the owner approves before copying the text into their email tool." That description appears in the workflow list and helps anyone else on the team understand the process without opening the editor.

   Best practice

   • Add a description to the workflow before you start placing nodes. It costs 30 seconds and prevents confusion when the workflow list grows.
   • Save a version after every meaningful change. Each save is non-destructive.
   • Name nodes descriptively in the Properties Panel as you place them. "Collect Month's News" is clearer than the default "User Interaction" label.
3. 3

   Place the User Interaction node and collect the month's news

   The User Interaction node (user_interaction, blue) pauses the workflow and creates a task in the Task Inbox where the owner types in the month's news. That input is the only source of facts the drafting agent is allowed to use.

   Drag a User Interaction node from the Node Palette and drop it below the Start node. Click the node to open the Properties Panel. Change the label to "Collect Month's News". Under the Fields section in the Properties Panel, configure the fields the owner will fill in. Suggested fields:

   • Business announcements (text area): new products, services, events, or changes
   • Special offers (text area): any promotions or discounts to mention
   • Tone notes (text): any specific phrases to include or avoid this month
   • Newsletter subject line suggestion (text): the owner's preferred subject line

   Connect the Start node's source handle (bottom circle) to the User Interaction node's target handle (top circle) by dragging between them. When a workflow instance runs and reaches this node, the owner receives a task of type user_interaction in the Task Inbox. They fill in the fields and click Approve to submit. The workflow resumes with those values in its context.

   For a structured brief format to paste into the text area fields, the Small business newsletter brief template covers exactly what to include.

   Do this

   • Drag a User Interaction node onto the canvas below Start. In the Properties Panel, change the label to "Collect Month's News".
   • Under Fields in the Properties Panel, add: Business announcements (text area), Special offers (text area), Tone notes (text), Subject line suggestion (text).
   • Connect Start to User Interaction: drag from the bottom handle of Start to the top handle of User Interaction.
   • Click Save Version.

   Example

   A hypothetical florist might configure the User Interaction node with two fields: "This month's news" (a single large text area) and "Subject line idea" (a short text field). Keeping it to two fields reduces friction for a busy owner who runs the business alone. The owner pastes the brief they prepared using the newsletter brief template, clicks Approve, and the workflow moves to the drafting step automatically.

   Best practice

   • Keep the User Interaction fields to what the drafting agent actually needs. More fields mean more friction for the owner each month.
   • Use a text area rather than a short text field for the main news input. The owner needs space to paste a full brief.
   • The owner completes this task from the Task Inbox in the sidebar. They need the Admin or Use role to see the Task Inbox. Confirm their role before the first run.
4. 4

   Add the Agent Interaction node to draft the newsletter

   The Agent Interaction node (agent_interaction, purple) calls the Newsletter Drafter agent automatically. No task is created and no human is involved. The agent receives the workflow context, including the fields the owner submitted in the previous step, and produces a draft newsletter that becomes part of the workflow context passed to the next node.

   Drag an Agent Interaction node onto the canvas below the User Interaction node. Label it "Draft Newsletter". Click the node to open the Properties Panel. Under Agent Assignment, select the "Newsletter Drafter" agent and version v1. Connect the User Interaction node's source handle to this node's target handle.

   The agent's system prompt instructs it to draw only from the facts in the workflow context. Because the User Interaction node's fields are part of that context, the agent has access to everything the owner typed in and nothing else. This design choice prevents the agent from adding claims or details that were not supplied. The Agent Review node in the next step verifies that the boundary held.

   For the broader pattern of chaining Agent Interaction and Agent Review nodes, see How Do I Route AI-Generated Drafts Through Approvals and Compliance Checks in Dave by voolama?. For a ready-to-reference example of a similar draft-and-review structure, see Example Workflow: Blog Draft and Review.

   Do this

   • Drag an Agent Interaction node below User Interaction. Label it "Draft Newsletter".
   • In the Properties Panel under Agent Assignment, select "Newsletter Drafter" and version v1.
   • Connect User Interaction to Agent Interaction: drag from the bottom handle of User Interaction to the top handle of Agent Interaction.
   • Click Save Version.

   Example

   A hypothetical pet supply shop might find that the first draft the agent produces uses a slightly more formal tone than the owner intended. Rather than editing the draft manually each month, the owner updates the drafting agent's system prompt to add: "Use casual, friendly language. Contractions are fine. Avoid formal phrases like 'we are pleased to announce'." They create a new agent version (v2) with the updated system prompt and update the Agent Interaction node to use v2. The old v1 is then retired.

   Best practice

   • The Agent Interaction node runs automatically when the workflow reaches it. The owner does not need to do anything at this step.
   • If the drafts are consistently too long or too short, adjust the maxTokens value in the agent's Config JSON and create a new version.
   • The agent does not send the newsletter anywhere. It produces text that becomes part of the workflow context. The owner sees that text in the Human Review task Payload in Step 6.
5. 5

   Add the Agent Review and Routing nodes

   The Agent Review node (agent_review, pink) is the step that distinguishes this newsletter workflow from a general content approval workflow. It runs an automated quality check before the draft reaches the owner. The Newsletter Fact Checker agent reads the draft and the original news input side by side and returns PASS or FAIL. The Routing node (routing, purple) reads that result and sends the workflow to the Human Review node if the result is PASS, or back to the Agent Interaction node for a redraft if the result is FAIL.

   In the agency workflow at How Do I Set Up AI Content Drafting with Client Approval at My Agency?, the account manager does the fact-checking manually before the client sees the draft. In this newsletter workflow, a second AI agent does it automatically, so the owner only sees drafts that have already passed a machine check. The owner's Human Review task is then a genuine editorial decision, not a fact-checking exercise.

   Agent Review node: Drag an Agent Review node below the Agent Interaction node. Label it "Check Facts Against Brief". In the Properties Panel under Agent Assignment, select "Newsletter Fact Checker" and version v1. Connect the Agent Interaction node's source handle to this node's target handle.

   Routing node: Drag a Routing node below the Agent Review node. Label it "Pass or Fail?". The Routing node is logic-only: it evaluates a condition and sends the workflow down one of two output ports. No task is created and no AI is called. Connect the Agent Review node's source handle to this node's target handle.

   In the Properties Panel under Routing Configuration, set up two output ports:

   • Pass port: condition evaluates the Agent Review result field for the string "PASS". This port connects to the Human Review node in Step 6.
   • Fail port: condition evaluates the Agent Review result field for the string "FAIL". This port connects back to the Agent Interaction node ("Draft Newsletter") to trigger a redraft.

   The field name in the Routing condition must match exactly the field name the Agent Review node uses for its result. Check the Properties Panel of the Agent Review node to confirm the output field name before configuring the Routing node.

   Do this

   • Drag an Agent Review node below Agent Interaction. Label it "Check Facts Against Brief".
   • In the Properties Panel under Agent Assignment, select "Newsletter Fact Checker" and version v1.
   • Connect Agent Interaction to Agent Review.
   • Drag a Routing node below Agent Review. Label it "Pass or Fail?".
   • Connect Agent Review to Routing.
   • In the Routing node's Properties Panel under Routing Configuration, configure two output ports: Pass (result contains "PASS", connects forward to Human Review) and Fail (result contains "FAIL", connects back to Agent Interaction for a redraft).
   • Click Save Version.

   Example

   A hypothetical law office using this workflow for a client newsletter might add a second Routing condition on the Fail port: if the draft has failed the fact check three times in a row, route to a Human Review node instead of looping again. This prevents the workflow from cycling indefinitely if the agent cannot produce a passing draft from the brief supplied. The owner then sees the best available draft and decides whether to approve it or go back to the User Interaction step with a clearer brief.

   Best practice

   • Keep the fact-checking agent's pass and fail keywords simple and consistent. "PASS" and "FAIL" in all capitals are easy to match in a Routing condition and unlikely to appear accidentally in other output.
   • The Routing node does not create a task or call an AI. It only evaluates conditions and directs the flow. If the workflow appears to hang at the Routing node, check that the condition field name matches exactly what the Agent Review node outputs.
   • Consider adding a loop counter or a maximum retry limit to the Fail path so the workflow does not cycle indefinitely if the agent cannot produce a passing draft.
6. 6

   Add the Human Review node and End nodes, then activate

   The Human Review node (human_review, amber) gives the owner the final decision. They see the newsletter draft in the task Payload, click Approve or Request Changes, and add a comment if edits are needed. Every decision is logged with the user's identity and a timestamp.

   Human Review node: Drag a Human Review node onto the canvas and connect it to the Pass output port of the Routing node. Label it "Owner Approval". In the Properties Panel under User Assignment, assign it to the owner. The task Payload will contain the newsletter draft produced by the Agent Interaction node and verified by the Agent Review node.

   End nodes: Connect an Approve edge from the Human Review node to an End node labeled "Newsletter Approved". Connect a Request Changes edge back to the User Interaction node ("Collect Month's News") so the owner can update the brief and restart the drafting process with clearer input. Connect a Reject edge to an End node labeled "Newsletter Rejected". Every path must end at an End node: dangling paths cause instances to hang at runtime.

   Task status values in Dave by voolama are: pending (amber), in_progress (blue), completed (green), failed (red), cancelled (gray). The Admin or Use role is required to see and complete tasks in the Task Inbox.

   Activate the workflow: In the Workflows list, change the workflow status from Draft to Active. Run a test instance first: submit a hypothetical brief through the User Interaction step, confirm the agent produces a draft, confirm the Agent Review returns a PASS result, and confirm the Human Review task appears in the owner's Task Inbox with the draft in the Payload section.

   After approval: Dave by voolama does not connect to any named email marketing platform. The approved newsletter copy is the output of the workflow. The owner copies it from the task Payload or the instance output and pastes it into their existing email tool. The FTC CAN-SPAM Act (15 U.S.C. 7701) requires commercial emails to include a physical address and an unsubscribe mechanism: those elements belong in the email tool's template, not in the AI-drafted copy. If the newsletter includes testimonials or endorsements, the FTC's Endorsement Guides (16 CFR Part 255) apply.

   Every Human Review task completion is logged in the audit trail with the acting user's identity, the decision made, any comment they added, and a timestamp. When SOC 2 or GDPR compliance mode is active in Dave by voolama, these records support SOC 2 CC6.1 (logical access controls) and GDPR Article 30 (records of processing activities) as described in the Dave by voolama Tasks help documentation. The audit log is per-tenant and encrypted with AES-256-GCM authenticated encryption.

   Do this

   • Drag a Human Review node and connect it to the Pass output port of the Routing node. Label it "Owner Approval". Assign to the owner in the Properties Panel under User Assignment.
   • Connect the Approve edge from Human Review to an End node labeled "Newsletter Approved".
   • Connect the Request Changes edge from Human Review back to the User Interaction node ("Collect Month's News").
   • Connect the Reject edge from Human Review to an End node labeled "Newsletter Rejected".
   • Confirm every path leads to an End node. Check the node and edge counter in the canvas top-right.
   • Click Save Version.
   • Run a test instance with a hypothetical brief. Confirm the draft appears in the Human Review task Payload.
   • Set the workflow status to Active in the Workflows list.

   Example

   A hypothetical bookshop owner runs the workflow for the first time. She opens the Task Inbox, sees the User Interaction task labeled "Collect Month's News", fills in three announcements and a 10% discount offer, and clicks Approve. The Agent Interaction node drafts the newsletter, the Agent Review node returns PASS, and a Human Review task labeled "Owner Approval" appears in her Task Inbox. She reads the draft in the Payload section, adds a comment asking for the discount to be mentioned in the opening paragraph rather than the closing, and clicks Request Changes. The workflow routes back to the User Interaction step with her comment visible in the context. She updates the brief to specify where the discount should appear, resubmits, and the next draft places it correctly. She clicks Approve, then copies the approved text into her email tool.

   Best practice

   • Assign the Human Review task to a role rather than a specific user where possible. If the owner is unavailable, anyone holding that role can approve.
   • Add a comment when clicking Request Changes. The comment is stored in the audit log and appears in the workflow context, so the drafting agent can use it to improve the next draft.
   • The CAN-SPAM Act (15 U.S.C. 7701) requires commercial emails to include a physical address and an unsubscribe mechanism. Confirm your email tool's template handles these before sending the approved copy.

What makes this newsletter workflow different from a general content approval workflow

Who needs which role before you start

What this guide does not cover