Agent day
Check in, open Sales or Service, see attendance, return to recent work.
← Back to selected work
CrystalBio case study
A field app for agents who need to report work without fighting the software.
CrystalBio is now used by the team for daily sales, service, attendance, leave, and admin visibility. The work was about making that live field workflow simple enough to hold up in real use.
System map
The product had to connect field work, admin review, reports, and live-use reliability.
Instead of explaining every screen in paragraphs, this is the simple shape of the product.
Visit capture
Save the real visit first. Add customer, equipment, quote, proof, and notes later.
Admin view
See attendance, leave, agent profiles, entries, reports, and follow-ups.
Live reliability
Real usage, check-in state, reports, backups, and visible failure points.
SignalThe important work was not just making screens. It was deciding what each user should see first, what could wait, and what would break trust once people were using it daily.
Workflow
The core loop is short. The details sit behind it.
This is the product logic I kept coming back to while shaping the experience.
1
Start day
Login, check in, location captured.
2
Save visit
Sales or Service Step 1 records the real field action.
3
Add details
Customer, equipment, quote, proof, and office notes.
4
Admin acts
Attendance, leave, follow-ups, reports, and person-wise activity.
5
Use holds
Check-in state, reports, backups, access, and reliability checks.
Decisions
Six product decisions, shown as operating choices.
These are the choices that made the product calmer, safer, and easier to hand over.
Design around the phone.
Daily actions stay visible: Check in, Sales, Service, Leave, My entries, Attendance, Reports.
Split reports into steps.
Quick capture first. Details later, without losing the visit.
Keep admin focused.
Overview, approvals, agents, profiles, and reports each get one clear purpose.
Treat live use as quality.
Once people use it daily, state persistence, reports, access, backups, and clear failures affect trust.
Use AI with judgment.
Agents helped move fast, but logic, UX, hallucinations, and edge cases still needed review.
Ask production questions.
Storage, monitoring, backups, safer data paths, and hidden debug views came up early.
App screens
Four actual screens in one steady carousel.
The carousel keeps one consistent mockup height and removes the repetitive second visit form.
Agent homeQuick actions make the field loop visible immediately.
AttendanceWork type, location capture, check-in, history, and leave request stay in one place.
Visit updateOne representative visit form is enough; service follows the same interaction pattern.
Admin overviewAdmin sees visits, active agents, leave, follow-ups, and review status.
LiveThe CrystalBio team is using it now.
1 issueCheck-in state disappeared after returning to the app.
No majorNo other major errors were noted.
Live-use view
What had to hold up once people started using it.
A working internal product is not just a good-looking UI. It has to survive real check-ins, return visits to the app, admin review, and daily trust.
Why this project matters
I can bridge product, operations, UX, and AI-agent execution.
CrystalBio is strong because it shows product sense under real-world mess: field users, admin needs, live usage, one known check-in persistence issue, and AI-assisted building.
Product senseMoved scattered reporting needs into a clearer agent/admin structure.
Operating judgmentBalanced client needs, field constraints, daily usage, and reliability concerns.
AI coordinationUsed agents for speed while checking logic, UX, hallucinations, and failure paths.
CommunicationTurned technical and operational issues into plain product decisions.