Mobile App Case Study
Turning conference handshakes into follow-up that actually gets sent. Scout captures who a rep met and why it mattered, has AI draft the personalized follow-up, and tracks every LinkedIn invite through to “accepted.”
Role
UX Designer
End-to-end iOS
Team
CEO, AI engineer, frontend engineer
Timeline
3 weeks design-to-ship for MVP
Sales teams spend tens of thousands of dollars a year sending reps to conferences, but most of the people they meet end up as LinkedIn connections that never get a follow-up. A Teameight customer asked us for a better way to track and follow up on conference contacts, so I designed Scout, a mobile app where a rep scans a contact, records a quick voice note about the conversation, and gets an AI-drafted LinkedIn message ready to send. Scout also tracks whether the invite was accepted, so reps know which messages worked. I designed and shipped the MVP iOS app in three weeks. Customers used it at real conferences within a month of launch.
The problem
A rep has dozens of conversations a day at a conference, and each one ends with a LinkedIn scan and a “let’s talk soon.” By the hotel that night they have 100+ new connections, a notes app full of fragments, and no way to remember who was who. Follow-up either takes hours or doesn’t happen, and the invites that never get accepted quietly disappear.
Companies spend a lot of money to put reps in that room. Every lead not acted on is pipeline down the drain.
Teameight already covered the prospect meeting and everything after it. Scout filled the gap before, making sure the people reps met actually booked that first meeting.
10–100
conversations per rep, per day
100+
new LinkedIn connections with no context attached
A handful
of follow-ups actually sent, most of them generic
Research and approach
Scout started as a customer ask. An existing Teameight customer couldn’t keep up with follow-up after conferences and wanted a better way to track who their reps met. With a small team and a tight timeline, research was informal but focused: conversations with that customer, a walkthrough of how a rep’s conference day actually goes, and a look at the tools reps were already using.
Mapping the rep’s day
I mapped a rep’s day at a three-day conference to find where the pain actually sits. It isn’t in the conversations, which go well. It’s in everything after, and it compounds: each unlinked scan and half-typed note is small on its own, but by the hotel that night they add up to a pile the rep can’t work through.
What reps were already using
Each tool reps reached for solved one piece of the job, and each one taught us something about what to build.
Native QR connect
Universal. Everyone has it, a scan takes two seconds, and the connection is instant.
The connection is all it captures. No event, no note, no reason the person mattered. Invites that never get accepted sink into the list and are forgotten.
Build around the scan, not against it. Keep QR as the entry point and attach the context in that same moment.
Popl
Digital business cards
Easy way to share your own digital card and collect names at a booth.
Too heavy for a rep with 30 seconds between conversations. Contacts stay in Popl as bare names, with no context and no path into the CRM as usable leads.
Scope narrower. One flow tuned to the handshake, feeding Teameight directly.
Notes app
Plus the occasional spreadsheet
Zero friction. Already on the phone, and reps reach for it without thinking.
Gets messy fast. After a day it’s a wall of fragments tied to no one, and nobody stops to type after every conversation anyway.
Reps will capture, but they won’t type and won’t do it per contact. That pushed us to voice.
What we set out to build
1
Capture a contact in one tap, tagged to the event, so nothing depends on memory.
2
Let reps add context by voice or typing. A quick memo right after the conversation, and for the ones they miss, a single end-of-day recap that AI parses and routes to the right contacts.
3
Have AI draft the follow-up, and keep pending invites visible so reps know who to nudge.
4
Sync everything into Teameight, so contacts land in the team’s pipeline rather than a standalone tool.
The system
Scout had to work as one continuous loop, capture → context → AI-drafted action → tracked outcome, rather than as a set of disconnected utilities. And it had to keep working when the room it was built for, a packed conference floor, actively worked against it.
01
Frictionless capture
Before
Reps relied on LinkedIn’s connect flow or paper business cards, with no link back to the event or conversation. Anything captured outside a formal scan arrived as a name with nothing behind it, or no record at all.
The solve
One-tap capture via LinkedIn QR scan or business card photo, auto-tagged to the current event. When neither happened in the moment, a LinkedIn name search (“That’s Them!”) or manual entry serves as a fallback, so a missed scan doesn’t mean a lost contact.
02
Context that survives the flight home
Before
Notes lived in a generic notes app, disconnected from the contact, and only useful if the rep stopped after every conversation to write them.
The solve
A hands-free voice memo right after each conversation, transcribed and cleaned up automatically and attached to that contact. For the conversations a rep doesn’t get to, a single end-of-day voice recap of the whole event that AI parses and routes to the right contacts. The recap even surfaces people the rep mentioned but never formally scanned, so they don’t disappear.
03
AI drafts the follow-up, and tracks whether it landed
Before
Follow-up meant a blank message box, days after the conversation, with fading memory of what was said. It either took too long or got skipped. And afterward, invites that sat unaccepted fell off the radar, with nothing prompting the rep to follow up again.
The solve
AI drafts a personalized LinkedIn message from the captured context, using a template the rep can customize with variables (contact name, event, company) and edit by hand or by voice. One tap copies it to the clipboard and opens the contact’s profile. Scout then tracks the invite through to “Connection Accepted,” turning a one-way send into a trackable outcome.
04
Resilient by design
Before
Conference wifi dead zones and unreadable business cards are the norm, not the edge case. A tool that only worked in ideal conditions would fail exactly when reps needed it.
The solve
Scans and notes taken offline are visibly marked, queue locally, and get parsed and enriched once the rep is back online, with a persistent “Currently Offline” state so they always know whether a capture is live or queued. When AI can’t read a card (handwriting, glare, odd angle), Scout says so plainly and keeps the photo visible for manual correction instead of guessing.
Rounding out the system, a team contact pool (“My Contacts” and “All Contacts,” filterable by event, date, source, and type, with CSV export) makes sure conference contacts land in the team’s pipeline instead of one rep’s pocket.
Decisions and tradeoffs
Voice memos over a structured form
Options
A typed form (name, company, role, next step) or a voice memo.
Chose
Voice. Reps won’t stop to type between conversations, and leadership saw voice-to-text as where every workflow was heading.
Tradeoff
Gave up clean, filterable data. Voice needs transcription and AI-linking before it’s useful, which added real work for the AI engineer.
Copy and open, not one-tap send
Options
Send the message directly through LinkedIn, or copy it and open the profile.
Chose
Copy and open. LinkedIn has no API for sending messages, so this was the closest achievable one-tap.
Tradeoff
The rep still pastes. One small manual step, but the hard part, the writing, is done.
Show the error, keep the photo
Options
When AI can’t read a business card, retry silently or guess, or show the failure.
Chose
Show it, and keep the photo visible so the rep can fix it. A confidently wrong contact is worse than “we couldn’t read this.”
Tradeoff
A less “magical” happy path, in exchange for records the rep can trust.
Working inside the constraints
The MVP had to exist before a specific customer conference, which set a hard three-week design-to-ship window. That forced scope discipline. QR scan, contact capture, and basic CRM sync made v1. Voice memos, business card scanning, and offline handling followed in P2.
With a four-person team (CEO, me, an AI engineer, and a frontend engineer), there was no room for a layered handoff process. I worked directly with the AI and frontend engineers throughout, negotiating what was actually buildable in real time rather than writing a spec and throwing it over the wall. The wifi dead-zone problem was anticipated from the team’s own prior conference experience, so offline queueing was built in before it became a live failure rather than after.
Outcome
3 wks
design to production, MVP
Oct 22
2025, App Store launch
P2
voice, card scan, offline by end of Nov
3
conferences where the customer used it
Scout shipped its MVP within the three-week window and was in customer hands at real conferences by mid-November 2025, with P2 following by the end of the month. The customer who asked for it used it at three conferences and was very satisfied. It did exactly what they needed: reps captured contacts and sent follow-up from the app in the chaotic conditions it was built for.
Scout was never Teameight’s core product, though. About eight months after launch, the company refocused on its meeting-to-close platform, and Scout, as a separate mobile app, was deprioritized.
TeamEight Scout
Automate event networking
Free · iPhone · Business
5.0 ★
1 rating
16+
age
TeamEight.AI Inc
developer
30.6 MB
size
View on the App Store
What I learned
Designing for AI means designing for its failures on purpose
The more a product hands work to AI, the more the design has to account for when it’s wrong, as a first-class part of the flow rather than an edge case bolted on later. Scout’s answer was to surface failure honestly and never let the underlying information, the photo or the raw capture, quietly disappear just because the AI couldn’t parse it. That protected both the data and the rep’s trust in the tool.
Shoulder-to-shoulder beats handoff
The clipboard workaround only exists because I was in the room when the engineer said there was no send API. A constant back-and-forth about what was possible versus what I wanted to design is what let us hit a three-week MVP at all.
Deadline pressure plus every hat at once forces real ownership
With three weeks and no one else to design, decisions got made fast, and I owned outcomes end-to-end rather than a single piece of a larger pipeline. That meant making decisive tradeoffs sooner than I would have with more time, and living with them.