Product Guide

How NudgeX
works

Understand the concepts, workflow, and design decisions behind the Action Queue — from inbox connection to follow-up execution.

Core concepts

The building blocks

Six concepts that make up the NudgeX system. Understanding these helps your team get the most out of every follow-up.

Action Queue

Your prioritized list of follow-ups. NudgeX analyzes your inbox and surfaces the conversations that need attention — ranked by urgency, recency, and context. One queue, one decision at a time.

Reason Codes

Every recommendation comes with an explanation: why this conversation, why now, and why this priority level. Your team always understands the logic behind each suggestion before acting on it.

Guardrails

Configurable safety rules that apply to every outbound message. Set tone presets, restricted phrases, and follow-up frequency limits. Rules are enforced automatically — they can't be bypassed.

Drafts & Approval

AI generates a suggested reply for each follow-up. Your team reviews, edits if needed, then explicitly approves before anything is sent. No message ever leaves without a human click.

Feedback Loop

After each follow-up, mark the outcome: worked, didn't work, or unsure. This builds an audit trail and helps the system surface better recommendations over time.

Do-Not-Contact List

Maintain a workspace-level list of contacts who should never receive follow-ups. Contacts on this list are automatically excluded from the Action Queue — no manual checking required.

Workflow

From inbox to follow-up in six steps

Every follow-up passes through the same structured workflow. No shortcuts, no surprises.

01
CONNECT

Link your inbox

Connect your email provider through its official authorization system. NudgeX never sees your password — it uses secure, revocable OAuth access with read and send permissions only.

02
SYNC

Inbox analysis

NudgeX reads your recent email threads to understand conversation state: who replied, when, what was discussed. This is read-only analysis — nothing in your inbox is modified.

03
QUEUE

Queue generation

Based on the analysis, NudgeX builds a prioritized Action Queue. Each item includes the thread context, a priority score, reason codes explaining why it's recommended, and a suggested action.

04
REVIEW

Human review

Your team reviews each recommendation individually. Read the context summary, check the reason codes, and decide: follow up now, snooze for later, or dismiss entirely. Every action is explicit.

05
DRAFT

Draft & validate

When you choose to follow up, AI generates a draft reply. Before sending, the draft is validated against your guardrails — tone, restricted phrases, and follow-up limits are all checked automatically.

06
APPROVE

Approve & send

Review the final draft, edit if needed, then click to send immediately or schedule for a specific time. This explicit approval step is by design — it's how NudgeX keeps your outreach safe and intentional.

Guardrails

Safety rules your team can trust

Guardrails are enforced on every outbound message — whether written by AI or edited by your team. They protect your brand and prevent mistakes.

Tone presets

Choose from predefined tone profiles — professional, friendly, or formal. Every AI-generated draft and every manual edit is checked against your selected tone before it can be sent.

Restricted phrases

Define words or phrases that should never appear in outbound messages. The system flags violations before send, whether the draft was written by AI or edited by your team.

Follow-up frequency limits

Set a maximum number of follow-ups per thread. Once the limit is reached, the thread is automatically excluded from the Action Queue — preventing over-contacting.

Server-side enforcement

Guardrails are validated on the server, not in the browser. This means they can't be circumvented by editing the page, using browser tools, or through API calls.

AI & human-in-the-loop

What AI does — and what it never does

NudgeX uses AI as a tool for summarization, drafting, and analysis. Your team makes every decision. Here's the exact boundary.

AI DOES

Summarize conversations

AI reads your email threads and generates concise summaries so your team can quickly understand context without re-reading entire conversations.

AI DOES

Suggest draft replies

Based on conversation context, AI generates a suggested follow-up message. Your team always reviews and can edit before approving.

AI DOES

Extract objections and signals

AI identifies buying signals, objections, and key points from conversations — helping your team craft more informed, relevant follow-ups.

AI NEVER

Sends messages autonomously

No message is ever sent without an explicit human approval click. This is an architectural decision, not a setting. There is no way to enable auto-sending.

AI NEVER

Trains on your data

Your email content and conversation data are never used for model training, behavioral profiling, or any purpose beyond generating your recommendations.

AI NEVER

Makes decisions for you

AI recommends and explains. Your team decides and acts. The system is designed to augment human judgment — never replace it.

Your data

What we access, store, and protect

Transparency about data handling — for security teams, compliance reviewers, and anyone who wants to understand what happens under the hood.

AreaDetails
Inbox access
Read and send only — via your email provider's official OAuth authorization. No passwords stored. Fully revocable from your provider's settings at any time.
Data read
Recent email threads — subject lines, participants, timestamps, and message content. Used solely to generate follow-up recommendations and conversation summaries.
Data stored
Thread metadata, conversation summaries, action queue state, drafts (transient), feedback history, and audit logs. Raw email content is processed but not permanently stored.
Encryption
All sensitive credentials are encrypted using bank-grade encryption at rest. Each credential is individually secured with unique keys that are never stored alongside your data.
Workspace isolation
Every query and API call is scoped to your workspace at the infrastructure level. Your data is structurally separated from every other customer.
Audit trail
Every draft, edit, approval, send, and schedule is recorded with timestamps and context. The log is append-only — entries cannot be modified or removed.
FAQ

Common questions

Answers for teams evaluating NudgeX — from technical reviewers to end users.

NudgeX currently supports Gmail through Google's official OAuth 2.0 authorization. Additional providers are planned for future releases. You can connect and disconnect your account at any time.

No. Every outbound message requires an explicit human approval click. This is an architectural constraint, not a setting — there is no way to enable automatic sending. Your team stays in full control of every message.

The queue uses a combination of deterministic rules (overdue detection, conversation recency, contact frequency) and AI analysis (conversation context, response patterns) to prioritize your follow-ups. Each item includes reason codes that explain the recommendation.

Contacts on the DNC list are automatically excluded from the Action Queue. Any thread involving a DNC contact will not generate follow-up recommendations. The list is managed at the workspace level and applies to all team members.

Yes. Guardrails let you set tone presets, restricted phrases, and follow-up limits. These rules apply to every draft — whether AI-generated or manually edited. You can also edit any draft before approving it.

No. Your email content and conversation data are never used for model training, behavioral profiling, or any purpose beyond generating your follow-up recommendations and conversation summaries.

After each follow-up, you can mark the outcome: worked, didn't work, or unsure — with an optional note. This creates an audit trail and helps NudgeX surface better recommendations over time. Feedback is append-only and cannot be deleted.

Every database query and API call is scoped to your workspace. Your data is structurally separated from every other customer at the infrastructure level — not through software settings, but through architectural boundaries.

Ready to take control of your follow-ups?

Prioritized queue. Human approval on every send. Complete audit trail. Start following up with confidence.