Tap to Trust Designing the Approval Moment for Agentic AI

Problem Statement

AI agents are starting to take real actions on behalf of businesses — upgrading plans, processing refunds, booking appointments, updating customer records. For high-stakes actions, the owner needs to stay in the loop. But „Approve / Deny" is nowhere near enough. Design the mobile-first approval moment: the 10 seconds between the agent asking for the go-ahead and the owner tapping back. Sometimes it's a quick yes. Sometimes it's nuance — a cap, a condition, a whisper in the agent's ear. Sometimes it's „let me handle this myself." Design that full spectrum.


Context & Background 

Superchat is the communication OS for 9,000+ businesses globally — the place where WhatsApp, Email, Voice, and every other channel a customer reaches out on become one. We're now building the next layer on top: AI agents that don't just chat, but act — triggering payments, changing subscriptions, booking slots, updating records across connected tools. That shift — from agents that talk to agents that do — is one of the biggest design frontiers in SaaS right now. And nobody has figured out what it should feel like. This challenge is your shot at defining it.


Target User 

The always-on, never-at-the-desk SMB owner. A dental practice owner between patients. A restaurant manager mid-shift. A personal trainer between sessions. They trust their AI agent to handle most things — but for the moments that matter, they want to stay in control without opening a laptop.


Design Focus Area

Mobile-first interaction design · Notification & alert design · Micro-interactions & motion · Copy and tone of voice · Progressive disclosure of context · Optional: voice UI for true hands-free control.


What does a great solution look like to you? 

A great submission goes far beyond Yes/No. Strong solutions think about:

Context, not clutter. The owner sees exactly what they need — the action, the customer, the history, the risk — and nothing more. Sometimes that's a lot. Sometimes it's almost nothing. Knowing the difference is the craft.

The real spectrum of responses. Not just approve/deny, but approve-with-a-constraint („go, but cap at €50"), modify-and-approve (edit the action itself), whisper („do it, but ask them about X first"), take-over (I'll handle this myself), deny-with-reason (so the agent can explain it gracefully), or defer („ask me again in 10 min").

The customer on the other side. What does the customer see while the owner decides? Silence is a design choice. So is „let me double-check for you."

The handoff back to the agent. Once the owner decides, how does the agent continue — confidently, in-character, without breaking the conversation?

The emotional tone. It should feel light, fast, trusted. The owner keeps sipping their coffee.

Bonus points for thinking about failure modes (what if the owner doesn't respond?), learning loops (does the agent get smarter from each decision?), and voice-first interaction.

Scope tip: focus deeply on one realistic scenario (e.g. a plan upgrade, a refund, a booking change) and design it end-to-end. Depth beats breadth.

All copyrights @XDesignClub

Designed by XDesign Club

All copyrights @XDesignClub

Designed by XDesign Club

All copyrights @XDesignClub

Designed by XDesign Club