Enter the password to continue. If you've been given access, this should take a moment.
A six-week sprint redesigning the citizen support feature on the TAMM mobile app — Abu Dhabi's one-stop government platform. Eighty-six percent of cases were routing around the app, through helpline and email, because the support feature itself was buried, broken, and built around a rigid form. We replaced the form with a chat — and watched successful submissions climb 313% in the first quarter.
TAMM was meant to be the single front door to Abu Dhabi government services. The support feature inside it was meant to be how residents reported problems, requested information, and gave feedback to the right entity. But six months of data told a different story. Of every 100 cases coming in, only 8 were submitted successfully through the app. Six failed mid-submission. The remaining 86 came through the helpline or email — channels that are expensive to staff, hard to track, and slow to resolve.
Three problems sat underneath the number. The support landing was buried in the third layer of the IA with weak information scent. The taxonomy was vague — "Talk to Us", "Check the status of your Messages", "Can't find what you are looking for?" — none of which signalled what the feature actually did. And the submit-a-case form itself was crashing, didn't distinguish required from optional fields, and gave back error states like "Something Went Wrong, select Retry button."
The fix wasn't a redesign of the form — the fix was upstream. Move support out of the third IA layer and into a first-class destination, then rewrite the labels so a citizen knew what they were tapping before they tapped it. "Talk to Us" became six concrete actions: File a Complaint, Report an Incident, Request for Information, Request for Service, Give a Suggestion, Give a Compliment.
Six weeks is short. It works only if every phase feeds the next without rework. We started in the data — six months of support center records told us where the leak was — then benchmarked seventeen apps to see what worked elsewhere, ran a time-boxed ideation session that surfaced 59 ideas, prototyped four candidates, and ran formative usability testing through four iterative rounds. Each round closed gaps the previous round opened.
The brief was explicit — use the existing TAMM Design Language System, with necessary enhancements. So we extended rather than replaced. The same color tokens, the same chrome, the same footer. What we added were three new patterns the existing DLS didn't have: a Support Assistant chat persona, a case timeline component, and a unified attachments + location picker. All three are now part of the system.
With the IA promoted and the system extended, four principles guided every screen we drew across the support flow. The screens themselves are how we evidence them.
Support promoted from a buried 3rd-layer page to a first-class destination. Submit a Case sits at the top with six explicit options. Follow up on cases is one tap away with an active count. Contact and emergency numbers occupy the lower half — there if needed, never in the way.
→ Support landing · six entry points, one tap deep
The form is gone. In its place — a step-by-step chat with a named Support Assistant. The user replies in their own words. The assistant prompts for the next thing it needs. Description, then attachments, then location, then terms — each a turn in the conversation, not a field on a form. Errors and required fields disappear because there are none.
→ File a Complaint · chat-led case submission
Every label rewritten from feature-language to citizen-language. Talk to Us became six distinct verbs. Check the status of your Messages became Follow up on cases. Can't find what you are looking for? — gone, replaced by self-evident structure. The information scent runs from the entry point straight to the case form.
→ Before and after · vague feature labels vs. action-led citizen labels
A submitted case used to disappear. Now every case lives on a follow-up surface — searchable, filterable, with a four-stage timeline at the top of every case detail. Updates from the government entity arrive as messages on the same thread. Citizens can add comments and attachments at any stage, right up until Case Resolved.
→ Follow-up surface · cases tracked from received to resolved
Hand-off to engineering happened at end of week six. The redesign launched into the TAMM iOS and Android apps with the new Support Assistant pattern, the case timeline, and the unified attachment-and-location picker — all built on top of the existing DLS. Within a quarter, the channel split flipped: more cases came in through the app than through helpline and email combined.
The before-and-after that explains the channel shift. The old "My Support" sat behind two taps and a confusing entry; the new TAMM Support is a destination — Submit a Case with six labelled actions, Follow up on cases with a live count, contact and emergency numbers as a calm secondary block.
↳ proves Findability
The signature move. A four-step chat replaces the broken form. The Support Assistant introduces itself, asks for a description, prompts for attachments and location, then asks the citizen to read and agree to terms — each a single turn in the thread. No required-field errors. No mid-flow crashes. No retry loops.
↳ proves Conversational
The iteration arc, in screens. V1 tried a chatbot persona ("TAMI"). V2 expanded to a list-based menu. V3 simplified the menu but introduced new ambiguity in the timeline. V4 — shipped — closed every gap, with explicit action labels, a search-first FAQ, and a follow-up surface front and centre. Task success climbed 60 → 80 → 76 → 92.
↳ proves Clarity
The case lifecycle, end to end. Citizens land on a list of their ongoing cases — filterable by type, searchable by reference number. They open a case to a four-stage timeline. They add comments and attachments at any stage. The government entity replies on the same thread. The case closes with a clear Case Resolved confirmation. Nothing disappears.
↳ proves Continuity
Increase in successful case submissions per quarter — from 300 to 1,240 in the first quarter post-launch.
Drop in helpline + email volume — from 86% of cases to 15%. The channel split flipped.
Six weeks isn't a lot of time, but it's enough when the discovery does its work. The data was unambiguous — eighty-six percent of cases were going around the app — and that single number framed every decision that followed. We didn't redesign the form. We replaced the pattern. The form became a conversation. The buried entry became a first-class destination. The vague labels became six clear verbs. And the case that disappeared at submission became a tracked thread from received to resolved.
The headline numbers — 313% increase in successful submissions, 71% drop in helpline-and-email leak, 92% task success in the final prototype — were what the work bought back: a citizen support channel that citizens actually used, government entities that could track and resolve cases inside one system, and a TAMM DLS three patterns richer than when we started.