SIMON: an AI support assistant new members can trust

Problem

New members didn't know what Hanger could do, and every question waited on a person

Hanger was about to onboard its waitlist, and the platform does a lot at once. Until then, the only way to get an answer was to contact the team and wait.

How might we give new members instant, trustworthy answers, and a person when the AI can't help?

Competitive analysis

I studied Drift, HubSpot, Chatfuel, and Intercom to learn what users love and hate in support bots

I compared four popular chatbots and read their reviews on G2, Capterra, and TrustRadius to see where each wins and where users get frustrated.

Key takeaways for SIMON

The decision: guided FAQs over open chat

The AI wasn't ready to answer open questions reliably, and a wrong answer on day one would cost more trust than a limited one.

Iteration

Three iterations: the same screens in V1 and V3

Welcome

From a blank greeting to suggested questions

V3 introduces SIMON by name and puts suggested questions front and center.

Answers

From a plain reply to one you can check

Source links, a "Was this helpful?" check, and next-step chips.

Can't help

From "email us" to three ways forward

Rephrase, browse the help center, or talk to a human.

Handoff & edge cases

When SIMON can't help, it says so and gets you to a person

The riskiest moment in an AI chat is the dead end, so every failure gets a clear next step.

Human handoff

Email or ticket, with the chat attached

The confirmation shows a ticket number and when to expect a reply.

Failure states

Offline, failed, and slow, each with a next step

Messages queue while offline, failed sends get a retry, and slow replies offer a way to reach the team.

Final design

SIMON V3: guided, honest, and one tap away from a person

AI disclosure up front

Before the chat starts, users see that SIMON is an AI, what it's good at, and that a person is always available.

Answers you can check

Each answer links to its source and ends with a thumbs up or down, so the team sees what isn't landing.

Resolution and rating

Chats end with a confirmation and a quick rating, a quality signal from launch day.

Feedback feature

A way for members to tell us what's broken, without leaving the page

Research I ran with Claude pointed to persistent, user-initiated feedback widgets getting 2 to 5 times the response rate of targeted surveys. I explored three directions for Hanger.

1 · Send us feedback

Rate the experience, then pick a path

Report a bug, share an idea, or leave general feedback.

2 · Report a bug

Members describe it, we attach the rest

Page, browser, device, and time attach automatically, and a screenshot is one tap away.

3 · Sent or failed

Clear outcomes, nothing lost

A confirmation with a reference number, or an error that saves the draft.

4 · SIMON shortcut

Or just tell SIMON

Members can report a bug in chat, and SIMON files the same report.

My AI workflow

AI helped me explore 3 concepts in 1 day instead of 2

I used Claude, ChatGPT, Gemini, and Figma AI to research chatbot patterns, draft information architecture, and compare early layouts, like an image carousel vs. a compact list for event recommendations. The decisions stayed mine.

Impact

Live as Hanger's first line of support, with a smarter V2 on the roadmap

“There are options for the user to easily pick and choose any questions that are already pre-laid out for them. So I think it's very user friendly.” — Design Lead, Hanger

Support tickets now come in through SIMON's handoff flow, and its UI patterns went into Hanger's Figma design system.

Reflection

What I learned designing for AI