Human Control in AI: Rahul Kumar Highlights Relaticle’s Approval-Gated Approach

In a recent LinkedIn post, Rahul Kumar discusses the proliferation of AI features in products, questioning the superficial implementation he often observes. He highlights Relaticle, a product featured on Product Hunt, as a noteworthy example of how AI can be integrated more thoughtfully into Customer Relationship Management (CRM) systems.

Kumar contrasts common AI integrations with Relaticle’s approach, noting that many products simply add a chatbot and label it as AI. He states:

“Here’s a chatbot. We added AI.” 😅

This, according to Kumar, misses the mark. He introduces Relaticle as a platform that rethinks AI’s interaction with CRMs, focusing on a core principle: transparency and human oversight.

Rethinking AI Interaction with CRMs

Rahul Kumar argues that the key differentiator for Relaticle lies in its “approval-gated writes.” This means that instead of an AI agent making changes to customer data autonomously, the system presents proposed changes for human approval. Kumar believes this is more than just a user experience decision; it represents a fundamental shift in how we should approach AI integration.

As Kumar elaborates on this critical distinction:

As AI agents become capable of doing more things on our behalf, the real question won’t be: “Can AI do this?” It’ll be: “Should AI be allowed to do this without asking me?”

This perspective underscores a growing concern about the autonomy of AI agents, particularly when dealing with sensitive business data like customer information. Kumar suggests that Relaticle’s approach directly addresses this concern by prioritizing human control.

The Philosophy of Empowered, Yet Controlled, AI

Kumar outlines several key features of Relaticle that support this philosophy:

• 37 first-party MCP tools
• Connect agents like Claude, ChatGPT, Cursor & Gemini
• Approval-gated writes
• Customizable CRM data models
• Self-hosting with AGPL-3.0
• Local AI inference with Ollama
• No per-seat pricing for the cloud

The underlying philosophy, as highlighted by Kumar, is to “Give AI more power, but don’t remove human control.” This balanced approach, he contends, is particularly crucial for managing customer data, which requires a high degree of trust and accountability.

The Importance of Human Oversight

Kumar emphasizes that the decision to require human approval for AI-driven changes in a CRM is a significant one. He posits that as AI capabilities expand, maintaining human oversight becomes paramount to prevent unintended consequences or data breaches. This focus on human control differentiates Relaticle from what Kumar perceives as more superficial AI implementations.

The discussion then pivots to a question posed by Kumar to his audience:

Would you trust an AI agent with policy-based auto-approval for certain CRM actions, or should every single write require human approval?

This question invites further consideration of the practical application and acceptance of AI in business workflows, particularly concerning the balance between automation efficiency and the need for human judgment and control. Kumar’s post serves as a platform for examining these evolving dynamics in AI adoption.