User Experience Flows
Clay’s user experience is a loop: capture direction, clarify fit, recommend opportunity, help the user act, then learn from the outcome. The product should feel personal and proactive, but every flow must still protect consent and move toward a concrete opportunity action. This spec connects the product primitives in Intentions, Personalities, Opportunities, and Companion experience into end-to-end user stories. See Curious pet memory for the persistent learning layer that organizes user patterns, perspectives, topics, and personality into personal MDX knowledges.Experience Promise
I can say where I am trying to go
The user can name an intention in plain language without turning it into a resume, profile, or
generic prompt.
Clay understands how fit should feel
The user can express working rhythm, communication style, motivation, and social energy without
exposing raw private reflections.
Opportunity becomes actionable
Clay turns fit into a recommendation, intro, brief, draft, reminder, or next step instead of
leaving the user in chat.
Control stays visible
The user can decide what Clay remembers, shares, snoozes, corrects, or keeps private.
Primary Experience Loop
Capture
The user gives Clay one intention, one opportunity lane, readiness, a personality signal, and a
consent boundary. The current iOS slice is documented in Mobile capture
flow.
Clarify
Clay asks one high-leverage follow-up when the intention is too vague, stale, contradictory, or
missing a constraint that would change opportunity fit.
Match
Clay surfaces a person, team, project, role, event, community, collaborator, or partner
experience with a readable fit explanation.
Act
The user chooses what Clay should do next: draft a fit brief, prepare an intro, route the user,
remind them later, reject the option, or ask for a different lane.
First-Run Journey
The first-run journey should borrow the useful parts of mobile questionnaire onboarding without becoming a long personality quiz. Clay needs enough signal to deliver a real first output.| Stage | User question | Clay behavior | Product output |
|---|---|---|---|
| Welcome | ”What is Clay for me?” | Shows the end state: better opportunity fit. | A clear reason to continue. |
| Intention | ”What am I trying to become or do?” | Captures one concise direction. | First intention signal. |
| Friction | ”What blocks or constrains this?” | Asks only if needed to explain relevance. | Constraint signal. |
| Personality | ”What kind of fit actually works?” | Captures rhythm, context, or feedback preference. | First personality signal. |
| Consent | ”What can Clay share outside itself?” | Defaults to translated fit signals only. | Explicit sharing boundary. |
| Demo | ”Can Clay help now?” | Produces a draft fit brief or example opportunity. | Tangible value before account. |
| Save | ”Do I want this remembered?” | Saves only after value is visible. | Account, memory, or local draft. |
The current app starts at the intention stage. It should grow toward a tangible demo output before
introducing heavier account, notification, or recommendation surfaces.
Core Flows
Flow 1: Capture An Intention
| Field | Spec |
|---|---|
| Entry | New user opens mobile, returns from landing, or taps New intention. |
| Main action | User writes one concrete outcome and selects lane, readiness, personality, consent. |
| Success state | Clay can draft a fit brief for the selected lane. |
| Failure state | Intention is empty, too broad, conflicting, or missing a decisive constraint. |
| Recovery | Ask one specific clarification, not a full interrogation. |
| Consent boundary | Save translated fit signals by default; raw reflections stay private. |
Flow 2: Review A Fit Brief
| Field | Spec |
|---|---|
| Entry | User saves an intention or chooses Draft a fit brief. |
| Main action | Clay turns intention, readiness, personality, and constraints into a short brief. |
| Success state | User understands what Clay would share and why it is enough for opportunity fit. |
| Failure state | Brief exposes raw reflection, invents facts, or hides which signals it used. |
| Recovery | Let the user edit, remove, or approve each shareable signal. |
| Consent boundary | Nothing leaves Clay until the user approves the translated version. |
Flow 3: Evaluate An Opportunity
| Field | Spec |
|---|---|
| Entry | Clay finds or receives a candidate opportunity. |
| Main action | User reviews relevance reason, tradeoffs, readiness, and one recommended next step. |
| Success state | User can accept, reject, snooze, ask why, or change the lane in one decision. |
| Failure state | Recommendation feels like a generic feed item, resume filter, or dating-style swipe. |
| Recovery | Ask whether the bad fit was goal, timing, rhythm, trust, geography, or energy. |
| Consent boundary | Show whether action requires sharing outside Clay before the user commits. |
Flow 4: Prepare An Introduction
| Field | Spec |
|---|---|
| Entry | User accepts an opportunity that involves another person, team, community, or partner. |
| Main action | Clay drafts an intro or outreach note using approved fit signals. |
| Success state | User can send, edit, save, or ask Clay to make the note warmer, shorter, or clearer. |
| Failure state | Draft overclaims fit, reveals private context, or pretends Clay has human authority. |
| Recovery | Highlight the shareable claims and require explicit confirmation before sending. |
| Consent boundary | External messages contain approved translated signals only. |
Flow 5: Debrief And Learn
| Field | Spec |
|---|---|
| Entry | User acted, ignored a recommendation, met someone, joined something, or rejected fit. |
| Main action | Clay asks one short debrief question tied to the previous signal. |
| Success state | User can mark fit as good, wrong, too soon, too much, or not relevant. |
| Failure state | Clay nags after dismissal or treats silence as positive signal. |
| Recovery | Snooze, quiet mode, archive intention, or correct the underlying assumption. |
| Consent boundary | Debrief is private unless the user explicitly converts it into a shareable signal. |
User Stories
Story 1: Save First Intention
As a new Clay user, I want to save one concrete intention so that Clay can start helping me find a fit without forcing me to build a full profile. Acceptance criteria- Given the intention field is empty, when the screen renders, then the save action is disabled.
- Given the user enters an intention, when the text is non-empty, then the save action is enabled.
- Given the user saves, when the save succeeds, then the next action changes to a fit brief prompt.
- Given the user edits a saved intention, when the text changes, then the saved state resets.
Story 2: Control Shareable Context
As a privacy-conscious user, I want Clay to show what it will share so that I can benefit from opportunity discovery without exposing raw private reflections. Acceptance criteria- Given Clay creates a fit brief, when the brief is displayed, then it separates shareable signals from private source context.
- Given a user removes a signal, when the brief is regenerated, then that signal is not included.
- Given the user has not approved the brief, when Clay prepares external action, then sending is blocked.
- Given the user approves the brief, when Clay drafts outreach, then it uses only approved translated signals.
Story 3: Explain Opportunity Fit
As a user considering an opportunity, I want to know why Clay thinks it fits so that I can trust the recommendation or correct it quickly. Acceptance criteria- Given an opportunity appears, when the user opens it, then Clay shows the related intention, personality signal, readiness, and constraint tradeoffs.
- Given the recommendation has weak evidence, when it is displayed, then Clay labels uncertainty instead of overstating confidence.
- Given the user rejects the opportunity, when Clay asks why, then the user can choose one bad-fit category or dismiss the question.
- Given the user corrects the reason, when future recommendations are explained, then the corrected signal is reflected.
Story 4: Draft A Consent-Aware Intro
As a user who accepts an opportunity, I want Clay to draft an intro using only approved signals so that I can act quickly without leaking private context. Acceptance criteria- Given an opportunity requires outreach, when the user taps draft, then Clay generates a short editable note.
- Given the note includes a claim, when the user reviews it, then the source signal is visible.
- Given the note contains raw reflection, when validation runs, then the draft is blocked until rewritten.
- Given the user sends or copies the note, when the action completes, then Clay offers a lightweight reminder or debrief.
Story 5: Repair Wrong Fit
As a user receiving proactive suggestions, I want to tell Clay when it is wrong or too much so that the product becomes sharper without making me feel trapped. Acceptance criteria- Given Clay sends or shows a proactive opportunity, when the user marks
wrong fit, then Clay asks one optional correction question. - Given the user marks
too much, when the feedback is saved, then the related nudge category is reduced or paused. - Given the user archives an intention, when Clay evaluates proactive messages, then that intention no longer triggers nudges.
- Given Clay misread a signal, when it acknowledges the correction, then it does not repeat the same assumption.
Story Split Backlog
These slices keep the product vertical: each one should change what the user can experience, not only add a technical layer.| Slice | Story | First useful version | Later sophistication |
|---|---|---|---|
| 1 | Capture first intention | Current iOS form saves local state and unlocks next-action UI. | Persist to backend, multi-intention management. |
| 2 | Draft local fit brief | Generate a deterministic preview from captured fields. | LLM-assisted wording with validation. |
| 3 | Approve shareable signals | User reviews editable brief sections before action. | Per-recipient consent policies. |
| 4 | Opportunity review card | Static opportunity candidate with explainable fit reasons. | Backend-backed opportunity suggestions with reviewed rationale. |
| 5 | Intro draft | Editable outreach text using approved signals. | Send through connected apps after consent. |
| 6 | After-action debrief | One-tap feedback on relevance quality. | Learning loop that adjusts future explanations and cadence. |
| 7 | Quiet and correction controls | Pause a nudge category and correct one signal. | Full memory and preference editor. |
| 8 | Personal MDX knowledges | Generate and edit a profile.mdx preview from approved signals. | Topic pages, correction history, export, and retrieval. |
Interaction Rules
- Ask one question at a time, and only when the answer changes opportunity fit.
- Show the user’s current intention before asking for more context.
- Explain recommendations through readable fit signals, not opaque mechanics.
- Treat silence as unknown, not consent or approval.
- Put the next concrete action above generic chat affordances.
- Offer dismiss, snooze, mute, correct, and archive paths wherever Clay initiates contact.
- Do not present Clay as a therapist, recruiter, romantic partner, or human agent.
Out Of Scope
- Public marketplace browsing with no intention context.
- Resume parsing as the primary onboarding route.
- Dating-style swipe mechanics for people recommendations.
- Auto-contacting people, teams, communities, or partners without explicit user approval.
- Raw personality labels or private reflections shared outside Clay.
- Notification permission prompts before Clay demonstrates a useful proactive moment.
Done Criteria
Clay’s user experience satisfies this spec when:- A new user can reach a concrete fit brief from one clear intention.
- Every opportunity recommendation explains the intention, personality, readiness, and constraint signals behind it.
- Every external action has an explicit consent review.
- Every proactive message has a reason, one ask, and an escape hatch.
- Corrections change future behavior instead of disappearing into chat history.
Related
Mobile capture flow
The first iOS slice that implements the Capture step of the primary loop.
Companion experience
The proactive voice and cadence that makes the loop feel like a friend checking in.
Curious pet memory
The persistent learning layer that backs the Learn step of the loop.
Consent
The boundary that keeps every external action in the loop consent-aware.