Ankit Kumar

We recently explored a human-in-the-loop approval flow for AI-powered workflows.

When an AI system wants to perform an important action, it can pause, explain what it plans to do and wait for a user to approve, reject or modify the action.
This helps reduce unchecked AI actions while still saving time across business workflows. Some practical responsible AI controls include:

Human approval for sensitive actions
Clear reasoning and action context
Controlled tool permissions
Audit logs for AI-triggered workflows
Safe interruption before execution

How are you designing human approvals without making AI workflows feel slow or frustrating?

Hey everyone, I ve been thinking about how the CRM space is changing and wanted to hear different views from people building, selling or using these tools.

CRM used to mostly mean contacts, pipelines, and follow-ups. But now teams are expecting more: AI summaries, automated follow-ups, connected emails, client portals, project delivery, invoices, support, and deeper workflow automation.

So I m curious, do you think the future of CRM is still around sales pipelines or will it move more toward complete customer/workflow operating systems?

For the past few months, I have been working on Navigator.

The objective here is not to create another AI chatbot or another code generator.