Design

How Outbounder makes dense research clear, usable, inclusive, and consistent.

English is the source language for the Premise AI | Outbounder website. Pages are localized into other languages with OpenAI models. Check important information.

One clear path through a dense task

Choice: The information architecture separates public marketing and application details from the working product. Each work page has one main task and stable navigation.
Reason: Account research has many inputs and outputs. A clear route reduces the need to remember where work lives.
Customer and business value: A seller can start, resume, review, and share work with less search. Stable paths also reduce support and training work.
Tradeoff: Some advanced choices stay behind a control. A new user may need to open that control once to learn what is available.
Basis: Visible state, familiar language, and recognition over recall support this choice. Nielsen Norman Group usability heuristics.
Governance: Map every new page or action to an existing task. Reject duplicate routes, labels, or navigation levels unless user evidence shows a distinct need.

Blueprint structure and cobalt action

Choice: Thin blueprint lines show structure. Cobalt marks actions, selection, and progress. Neutral surfaces carry most content. The interface does not use gradients.
Reason: Generated sales material is dense. A restrained palette leaves attention for evidence, state, and the next action.
Customer and business value: Readers can scan faster and learn what blue means across pages. A small palette also makes defects easier to find.
Tradeoff: The product has less room for decorative variation. Color alone can never carry meaning.
Basis: Current Light and Dark tokens, contrast checks, and rendered page tests define the product's real palette in the Outbounder Specs.
Governance: Use a design token, which is one shared value for color, type, or spacing. A new color needs a named purpose plus contrast and state proof.

Type and words that explain themselves

Choice: A compact type scale separates display, section, body, interface, and caption text. Ordinary prose targets a United States Grade 10 reading level. Necessary terms are defined on first use.
Reason: Clear hierarchy helps readers scan long research packages. Plain language makes technical limits easier to judge.
Customer and business value: Sellers find useful material faster. Evaluators can understand a claim without translating internal jargon.
Tradeoff: Readability scores do not prove comprehension. Exact product names and standards must stay when accuracy requires them.
Basis: The writing gate checks grade level, sentence length, jargon, headings, and defined terms. Rendered tests are summarized in the Outbounder Specs.
Governance: Copy changes need a frozen source, evidence map, term register, deterministic lint, and separate writer and subject review.

Content-driven space and responsive reflow

Choice: Shared spacing values create rhythm. Responsive reflow rearranges content at narrow widths instead of clipping it. Content sets its own height.
Reason: Labels and generated text vary by language, viewport, zoom, and user input. Fixed content boxes cannot safely predict those sizes.
Customer and business value: Actions and evidence stay visible on phones, desktops, and zoomed views. Teams avoid device-specific repair work.
Tradeoff: A page can grow when content wraps. A wide layout may collapse sooner in a language with long labels.
Basis: Product tests cover width, zoom, language, theme, and overflow against WCAG guidance on reflow.
Governance: Test affected pages at 320 and 390 browser pixels, representative desktop widths, both themes, and every affected language. Do not hide a defect with a larger fixed frame.

Shared controls with clear state

Choice: Buttons, links, fields, tabs, segmented controls, dialogs, and navigation reuse shared geometry and state rules. Simple and Advanced behave as one segmented choice.
Reason: Consistent controls let people transfer what they learn from one page to another.
Customer and business value: Tasks take less interpretation and fewer corrections. Shared behavior also lowers implementation and review cost.
Tradeoff: A shared component may need a narrow variant for a real context. A variant must not change the control's meaning or keyboard behavior.
Basis: Product browser tests verify names, roles, states, focus, and target size using WAI interaction patterns.
Governance: Reuse an existing component first. A new variant needs a distinct user need plus normal, hover, focus, active, disabled, loading, error, and responsive proof where those states apply.

Honest state and recoverable work

Choice: Progress reflects stored generation phases. Validation appears near what it affects. Errors preserve valid work, explain what happened, and give one next action. Recovery resumes from durable state stored beyond the current screen.
Reason: Research can outlast a page view or meet a temporary service failure. False progress and vague errors break trust.
Customer and business value: A seller knows whether to wait, correct an input, retry, or reopen saved work. Preserving valid work reduces delay and repeated provider cost.
Tradeoff: Honest states can expose pauses that a decorative animation would hide. Recovery is limited to work the product actually retained.
Basis: Current phase, refresh, persistence, and no-partial-save tests apply the system-status and recovery guidance in the Nielsen Norman Group usability heuristics.
Governance: Define waiting, success, empty, error, retry, refresh, and handoff behavior before changing asynchronous work. Tests must follow the durable record, not a timer or mock sequence.

Access is part of the design

Choice: Outbounder targets Web Content Accessibility Guidelines (WCAG) 2.2 Level AA. It uses keyboard access, visible focus, semantic web elements, sufficient contrast, and names that identify controls to assistive technology.
Reason: Access needs affect structure and interaction, not a final polish pass.
Customer and business value: More people can complete the task. Clear focus, labels, and error links also help keyboard and power users.
Tradeoff: Automated checks cannot cover every browser, device, or assistive technology. The product does not claim universal certification.
Basis: The product gate checks reflow, contrast, keyboard behavior, names, roles, states, focus, and target size against the WCAG 2.2 Quick Reference.
Governance: Accessibility proof ships with the change that affects it. A waiver must name the untested combination and cannot be reported as a pass.

Nine complete language experiences

Choice: The interface and generated package support English, Arabic, Simplified Chinese, French, Hindi, Russian, Spanish, Swahili, and Brazilian Portuguese. Arabic uses right-to-left layout, localized numerals, logical spacing, and isolation for retained Latin names.
Reason: Translation changes length, script, direction, punctuation, and font needs. String replacement alone does not prove a usable result.
Customer and business value: Teams can work in the chosen language without English controls or unstable mixed-script text breaking the task.
Tradeoff: Proper nouns, product names, URLs, and fixed method names may remain in their official form. Every translation still needs review before use.
Basis: The nine-language registry, localized fixtures, Arabic direction rules, and rendered browser matrix define coverage summarized in the Outbounder Specs.
Governance: Test complete surfaces, not dictionary keys alone. Check every affected language, Arabic RTL, numerals, dates, protected names, wrapping, clipping, mobile, desktop, and both themes.

One system, tested in context

Choice: Long packages use named sections, stable tabs, source links, review cues, and editable text. Shared tokens and components are verified on the real page in every affected state.
Reason: A polished paragraph can look more certain than its evidence allows. An isolated mock can also hide clipping, direction, focus, or saved-work defects.
Customer and business value: Review is faster and unsupported claims are easier to catch. Shared evidence makes release decisions easier to audit.
Tradeoff: Evidence and contextual proof add work and page length. The test matrix stays bounded to affected risks instead of expanding without cause.
Basis: Product gates apply user-centered, accessible, and consistent-system principles from the U.S. Web Design System.
Governance: Each proposal states the user problem, simplest path, alternatives, tradeoffs, changed states, and proof. Approval requires the exact candidate, a hostile overengineering review, and a reversible release plan.