DIREKT Design DNA Brief and Flagship Review Contract

Programme: VC1–VC8 world-class modernization
Governing issue: #259
Status: Three differentiated directions defined; final visual direction requires explicit owner approval during VC2
Primary surfaces: Android, customer/provider web/PWA, privileged operations portal

1. Purpose

This brief controls high-fidelity visual exploration and the owner approval checkpoint before broad design propagation.

It does not authorize a replacement application or generated business logic. Existing canonical API, database, authorization, trust, privacy, payment and release controls remain authoritative.

2. Target character

DIREKT must feel:

  • world-class and current;
  • trustworthy but not governmental;
  • local, human and Zambia-relevant without looking informal or low-budget;
  • visually rich enough to feel like a real marketplace;
  • precise around trust, verification and limitations;
  • fast and task-oriented;
  • accessible and low-bandwidth aware;
  • consistent across Android/PWA/operations while allowing platform-native adaptation;
  • AI-assisted without looking like a chatbot product.

Avoid:

  • blanket Verified badges;
  • decorative trust seals;
  • generic fintech/glassmorphism;
  • generic SaaS dashboard treatment for customer/provider journeys;
  • luxury styling that suggests exclusivity;
  • large marketing heroes that delay service discovery;
  • primitive glyph navigation;
  • developer/API/workstream labels;
  • AI sparkle/robot decoration without clear function;
  • AI-generated claims that appear equivalent to canonical trust decisions.

3. Composite global benchmark

Use benchmark capabilities rather than literal visual copying:

  • Urban Company — marketplace maturity, provider professionalism and service-quality presentation;
  • Checkatrade — visible trust/rechecking proposition;
  • Taskrabbit — short customer path from need to provider/action;
  • Thumbtack — AI-guided natural-language/multimodal project discovery;
  • DIREKT — check-specific proof, privacy-by-precision, payment-independent trust, Africa-first resilience and human-accountable operations.

4. Immutable trust DNA

Every direction must preserve:

  • service fit/locality before promotional storytelling;
  • individual check states rather than one global verification badge;
  • what was checked;
  • check scope/category where relevant;
  • date/currentness/expiry;
  • limitations/what the result does not prove;
  • public-safe source class where appropriate;
  • commercial/payment state in a separate visual family;
  • private evidence never used as public imagery.

Provider imagery can create confidence and context, but it must never visually overpower or substitute for proof.

5. Privacy DNA

Public-safe design content may include approved:

  • provider display identity;
  • service/category;
  • public locality/service area;
  • consented premises location;
  • public work/premises imagery;
  • public trust claims;
  • reviews;
  • availability.

Never expose in prototypes/creative tools/public screenshots:

  • real identity/certificate evidence;
  • storage object references;
  • exact private base locations;
  • unconsented contacts;
  • private reviewer notes;
  • secrets/tokens;
  • real complaint/evidence material.

Use synthetic fixtures.

6. Accessibility and resilience DNA

Every direction demonstrates:

  • readable body typography;
  • 48dp-class primary interaction targets;
  • keyboard/focus/TalkBack/screen-reader semantics;
  • status not colour-only;
  • 200% scaling/reflow;
  • reduced motion;
  • map/list equivalence;
  • low-bandwidth imagery fallback;
  • loading/empty/error/offline/retry;
  • AI unavailable/manual fallback;
  • no critical action dependent on image or AI alone.

7. AI experience DNA

AI should appear as bounded assistance embedded in the task.

Required patterns:

  • Describe what you need composer;
  • clarifying-question chips/cards;
  • editable suggested category/query;
  • Why this result explanation;
  • provider comparison summary;
  • provider onboarding assistant;
  • operations case copilot;
  • AI unavailable/manual fallback.

Visual hierarchy must distinguish:

  1. user-entered facts;
  2. canonical DIREKT facts;
  3. AI-generated suggestion/summary;
  4. user-confirmed content;
  5. human-authorized outcome.

Never style AI output as a trust-success state.

8. Direction A — Structured Trust

Intent: calm, precise, confident marketplace with strongest information architecture and proof presentation.

Visual character

  • crisp neutral surfaces;
  • deep ink typography;
  • restrained green;
  • moderate radii/elevation;
  • generous whitespace;
  • compact semantic iconography;
  • supporting provider photography;
  • trust cards that feel understandable and authoritative without becoming government forms.

Customer

  • compact search + area first;
  • clean category tiles;
  • provider cards with one strong work image, service/locality, availability and two trust snippets;
  • provider profile puts What DIREKT checked immediately after identity/service;
  • results use list-first fallback and desktop list/map split.

Provider

  • task-oriented readiness;
  • Next actions, Services & availability, Enquiries, Publication status;
  • vertical evidence requirement/timeline;
  • calm action-required states.

Operations

  • dense three-pane queue/case/evidence-decision workspace;
  • evidence central;
  • neutral surfaces with semantic accents;
  • strong keyboard/focus treatment.

AI

  • minimal assistant chrome;
  • suggestions appear as precise contextual panels rather than conversational bubbles;
  • strong separation between generated synthesis and canonical facts.

Strengths

  • strongest trust transparency;
  • accessibility/low bandwidth;
  • easiest cross-platform consistency.

Risk

Can feel clinical unless imagery and copy add warmth.

9. Direction B — Neighbourhood Marketplace

Intent: warmer, more human and visually rich while keeping proof immediately accessible.

Visual character

  • brighter natural provider/work imagery;
  • Zambia-relevant illustrated/photo category moments;
  • warmer neutral surfaces;
  • softer card shapes;
  • stronger local marketplace personality.

Customer

  • friendly location context;
  • visually rich category grid/mosaic;
  • image-led provider cards with trust snippets directly below identity/service;
  • provider profile uses stronger gallery/story but trust summary remains above reviews/storytelling.

Provider

  • welcoming business identity header;
  • visual task/readiness cards;
  • What needs attention, Your services, New enquiries, How customers see you.

Operations

Consumer warmth is reduced. Operations remains structured and evidence-first with only modest softer surfaces.

AI

  • more conversational wording and approachable prompts;
  • visual suggestions remain compact and confirmable;
  • no cartoon assistant persona.

Strengths

  • strongest marketplace warmth and provider humanity;
  • improves first impression and category discovery.

Risk

Imagery may overpower proof or increase bandwidth unless tightly governed.

10. Direction C — Field Utility

Intent: fast, efficient, service-first product optimized for task completion, field use and operational density.

Visual character

  • flatter UI;
  • fewer decorative containers;
  • grouped rows and compact cards;
  • strong labels and scan patterns;
  • minimal decorative imagery outside customer discovery/profile.

Customer

  • extremely fast search/category/area entry;
  • compact provider result rows/cards;
  • list/map controls prominent;
  • trust snippets highly scannable.

Provider

  • dashboard resembles an efficient business task center;
  • strong pending-action queue;
  • evidence and enquiry workflows optimized for one-hand mobile use.

Operations

  • strongest dense keyboard-driven queue/case review;
  • table/list efficiency;
  • evidence and decision always contextually adjacent.

AI

  • inline suggestions and command-like assistance;
  • minimal conversational UI;
  • ideal for operator summarization/checklist assistance.

Strengths

  • fastest low-bandwidth/task completion;
  • strongest provider/operations efficiency.

Risk

Can feel too enterprise/utility-like for customer marketplace discovery.

A potentially strong combination to test is:

  • A — trust hierarchy, semantic clarity, accessibility;
  • B — category/provider imagery, marketplace warmth and local identity;
  • C — provider workspace and operations density.

This is a hypothesis only. The owner must explicitly approve the final direction after seeing the same flagship scope rendered across candidates.

12. Required VC2 flagship review set

Each candidate direction renders the same seven experiences.

Customer

  1. Discover/Home
  2. search;
  3. area;
  4. category discovery;
  5. relevant provider preview;
  6. concise trust education;

  7. compact and desktop navigation.

  8. AI-assisted service-need entry

  9. natural-language problem description;
  10. optional future photo/voice affordance shown only as concept if not runtime-active;
  11. clarifying suggestion;
  12. confirm/edit/manual category fallback;

  13. AI unavailable state.

  14. Search results + map/list

  15. query/category/area context;
  16. filters;
  17. privacy-safe map/list equivalent;
  18. provider cards with imagery/locality/availability/scoped trust;

  19. explainable relevance treatment.

  20. Provider profile + trust detail

  21. public imagery separated from evidence;
  22. services/coverage/availability;
  23. check-specific trust summary/details;
  24. dates/scope/limitations;
  25. reviews;
  26. save/share/enquiry.

Provider

  1. Provider workspace/overview
  2. readiness;
  3. next actions;
  4. services/availability;
  5. enquiries;
  6. publication state;

  7. commercial state separate from trust.

  8. Verification/evidence status

  9. requirements/timeline;
  10. upload/retry;
  11. action required;
  12. expiry;
  13. private evidence explanation;
  14. optional AI quality assistance concept clearly labelled.

Operations

  1. Verification queue + case/evidence review
  2. desktop queue/case/evidence-decision composition;
  3. priority/SLA/ownership;
  4. secure synthetic evidence state;
  5. checklist/reason/decision;
  6. audit/history;
  7. AI summary/checklist assist clearly secondary;
  8. compact triage/field sample.

13. Required variants

  • compact/mobile customer/provider;
  • expanded desktop customer/provider web;
  • tablet/adaptive where composition materially changes;
  • desktop operations review;
  • compact operations triage/field;
  • representative loading;
  • empty/no results;
  • error/offline;
  • AI unavailable/manual fallback;
  • dark state where required by final theme policy.

14. Owner decision record

VC2 exits only when exactly one is recorded:

  • APPROVE DIRECTION A;
  • APPROVE DIRECTION B;
  • APPROVE DIRECTION C;
  • APPROVE HYBRID with exact elements borrowed;
  • REVISE with concrete required changes.

Silence or implementation convenience is never approval.

15. Design tooling contract

Google Stitch

Use as the primary high-fidelity exploration source for:

  • candidate directions;
  • flagship screens;
  • mobile/desktop/adaptive variants;
  • visual Design DNA extraction.

Generated code/design metadata is advisory until reconciled with repository authority.

Higgsfield

Use only as a secondary creative layer for:

  • category concepts;
  • generic provider/work imagery;
  • illustrations;
  • moodboards/store assets.

Never send private evidence, real participant data, raw contacts, exact private locations, restricted operations screenshots or secrets.

16. Implementation handoff

After owner approval:

  1. record the approved reference/project/screens;
  2. extract/reconcile Design DNA;
  3. map tokens/components to existing Compose/web/operations stacks;
  4. document rejected patterns;
  5. create synthetic-safe visual baselines;
  6. proceed VC3 onward in bounded exact-head tested slices.

No design tool may redefine canonical business logic, trust policy, authorization or integration truth.