Plans

​

1. document purpose

• The user problems being solved
• The business outcomes expected
• The required feature behavior
• The AI interaction model
• The acceptance criteria for release readiness

​

2. vision and product promise

• Automatically allocate income across essentials, variable spending, savings, and goals
• Continuously adapt when real-world behavior changes
• Proactively guide users with timely interventions
• Minimize manual input through agentic AI and interactive UI
• Build durable habits through reinforcement, visibility, and gamification

​

Product promise

1. What must be protected first?
2. What is safe to spend?
3. What should go toward goals right now?
4. What should change if the month does not go as planned?

​

3. business problem and opportunity

​

Current user problems

​

Product opportunity

• Higher retention because the product becomes useful throughout the month, not only during review moments
• Higher perceived intelligence because the product adapts and acts with context
• Higher long-term user value because planning, recovery, and goal completion are emotionally sticky behaviors

​

4. target users and jobs to be done

​

Primary user segments

​

Core jobs to be done

• Help me decide what to do with my income before I spend it.
• Help me stay on track without forcing me to constantly manage a budget.
• Warn me early when my current behavior will hurt my goals.
• Show me that small daily choices affect real outcomes.
• Keep me motivated long enough to build a repeatable financial habit.

​

5. product principles

​

AI-native principles

​

UX principles

• Keep cognitive load low
• Use plain-language recommendations
• Show trade-offs clearly
• Never overwhelm the user with too many prompts at once
• Balance automation with user control
• Reinforce progress more often than failure
• Preserve dignity and avoid judgmental tone

​

6. scope

​

In scope

• Income-based planning
• Allocation across core spending and goal buckets
• Adaptive rebalancing recommendations
• Goal pacing and contribution tracking
• Predictive warnings before goal failure
• Transaction-level behavioral intelligence
• AI-generated missions and rewards
• User controls to accept, adjust, or reject AI actions

​

Out of scope for this BRD

• Tax planning
• Investment portfolio advice
• Household collaboration workflows
• Human advisor escalation
• Credit underwriting or lending decisions

​

7. user experience summary

Income detected or entered
        |
        v
AI proposes an allocation plan
        |
        v
User accepts or adjusts in 1-2 interactions
        |
        v
System tracks spending, contribution pace, and risk
        |
        v
AI suggests rebalancing, warns, explains, and celebrates
        |
        v
User learns, adapts, and builds stronger planning habits

​

Primary user-visible surfaces

• Plan setup and plan refresh cards
• Allocation review screen with sliders
• Goal progress summaries
• Warning and rebalance prompts
• Transaction detail insights
• Daily, weekly, and monthly mission cards
• Celebration moments for streaks, recoveries, and goal milestones

​

8. feature requirements

​

8.1 smart allocation engine

​

Purpose

• Fixed expenses
• Variable expenses
• Savings goals
• Discretionary spending

​

Business inputs

• Net income or expected monthly income
• Income timing and frequency
• Fixed obligations
• Core living costs
• Existing savings or debt-reduction goals
• Goal urgency and target timing
• User-reported lifestyle flexibility
• Income stability level
• Coaching preference

​

Default allocation logic

1. Protect minimum living cost and committed fixed expenses.
2. Reserve an appropriate variable-spend allowance for essential day-to-day categories.
3. Protect a baseline safety margin, especially for irregular-income users.
4. Allocate remaining funds across goals and discretionary spending based on user priorities.
5. Adjust the recommendation to fit one of three planning modes.

​

Planning modes

​

Default allocation guidance

​

Guardrails

• The recommendation must never underfund core living needs.
• The recommendation must account for recurring obligations before optional goals.
• The recommendation must reduce discretionary allocation before reducing critical goals.
• The recommendation must recommend a larger safety margin for variable-income users.
• The recommendation must switch to a stability-first plan when cash pressure is materially high.
• The recommendation must identify when the user has insufficient income to fund all stated priorities and force prioritization rather than masking the trade-off.

​

AI-generated recommendation content

• What was protected first
• Why the proposed percentages fit the user situation
• What trade-off the plan makes
• What the user can safely change without breaking the plan

​

User override capability

• Accept the recommended allocation in one tap
• Adjust allocation percentages with sliders before activation
• Re-rank goals when available funds cannot cover all goals
• Save a custom version of the plan

• Showing the impact on goal pace and discretionary room
• Flagging when the override weakens plan safety
• Allowing the user to proceed after acknowledging the trade-off

​

Acceptance criteria

• User receives a recommended allocation immediately after completing the required planning inputs.
• The recommendation shows percentages and estimated amounts for each bucket.
• The user can accept the recommendation without editing.
• The user can adjust the plan with sliders and see the impact preview before confirming.
• The system highlights when an adjustment drops the plan below the minimum living threshold or jeopardizes a priority goal.
• The system supports at least one active goal at plan setup and can handle multiple goals through prioritization.

​

Definition of done

• Users can understand the recommended plan without external explanation.
• Users can activate or adjust the plan within two primary interactions.
• Users can see the trade-off of any override before confirming it.
• The plan protects essentials in all normal and constrained financial scenarios.
• Variable-income users receive visibly more conservative recommendations.
• Edge cases are handled:
  • Income is too low to satisfy all stated priorities
  • User has no goals yet
  • User changes priorities during setup
  • User has irregular or multi-source income

​

8.2 adaptive rebalancing engine

​

Purpose

​

Business requirement

​

Trigger thresholds

​

Frequency rules

• Standard rebalancing suggestions should appear no more than once per day.
• Non-urgent rebalancing should be consolidated into a weekly summary if multiple small deviations occur.
• High-risk events may trigger immediate intervention.
• The system must pause repetitive prompts when the user repeatedly ignores the same recommendation type.

​

AI explanation of why

• What changed
• Why it matters now
• What the recommended adjustment would protect or improve
• What happens if the user takes no action

​

User interaction model

• Accept: apply the recommendation as proposed
• Tweak: adjust the suggested movement with sliders
• Reject: dismiss the recommendation and keep the current plan

​

Rebalancing principles

• Protect fixed obligations first
• Protect high-priority goals before low-priority goals
• Favor small recoveries before drastic changes
• Preserve user trust by limiting prompt fatigue
• Learn from prior rejections and narrow future suggestions

​

Acceptance criteria

• User receives a rebalance prompt when overspending or income deviation crosses a defined threshold.
• The prompt clearly states the current issue, the recommended change, and the likely outcome.
• User can accept, tweak, or reject the recommendation from the prompt.
• The system updates the active plan summary immediately after the user action.
• Repetitive prompts of the same type are reduced after repeated rejection or non-response.

​

Definition of done

• Users are warned early enough to take corrective action before month end.
• Rebalancing recommendations preserve essentials and priority goals by default.
• Users can act on a rebalance within one to two interactions.
• The system avoids excessive prompt volume in noisy spending periods.
• Rejected recommendations influence future prompt frequency or aggressiveness.
• Edge cases are handled:
  • Multiple overspend categories at once
  • Mid-cycle income loss
  • Windfall income
  • User has already manually overridden the plan several times

​

8.3 goal contribution tracking

​

Purpose

​

Business requirements

• Each goal must show a target amount and target timing where applicable.
• Each goal must have a monthly contribution target derived from the active plan.
• Progress must update in near real time from user-visible activity.
• Users must see both lifetime progress and current-cycle progress.
• Goals must be ranked so users understand which ones are protected first.

​

Required user-facing elements

​

Contribution consistency scoring

​

Monthly target handling

• Monthly target must adjust when the user changes the goal, plan, or income expectations.
• If a contribution is missed, the system must show revised pace options.
• If a goal is completed early, the user must be prompted to reassign future contributions.

​

Acceptance criteria

• User can view monthly target, contributed amount, and remaining amount for each active goal.
• User can see cumulative goal progress and current-month pacing at the same time.
• Goal summaries clearly indicate when the user is ahead, on track, or behind.
• A consistency score is visible and updates over time based on contribution behavior.
• If the user misses a target, the system presents at least one recovery option.

​

Definition of done

• Users can tell in seconds how each goal is performing.
• Monthly targets remain aligned to the current plan state.
• Missed contributions trigger a recovery path rather than silent failure.
• Goal completion creates a clear celebratory moment and next-step suggestion.
• Multiple simultaneous goals are understandable without confusion.
• Edge cases are handled:
  • User creates a new goal mid-cycle
  • User pauses a goal
  • User completes a goal early
  • User has no surplus available for goals

​

8.4 predictive warning system

​

Purpose

​

High-level prediction logic

• Spending pace versus time elapsed
• Remaining unallocated money
• Remaining goal contribution required
• Historical responsiveness to interventions
• Income reliability for the rest of the cycle

​

Time-based triggers

​

Warning types

​

Suggested corrective actions

• Reduce discretionary spending for the remainder of the cycle
• Move unused funds from lower-priority categories
• Lower the current-cycle contribution target and extend timing
• Protect a priority goal while pausing a lower-priority goal
• Accept a smaller recovery action that prevents full failure

​

Acceptance criteria

• User receives a warning before the end of the cycle when projected behavior indicates likely failure.
• Warning type changes based on level of risk.
• Every warning includes at least one direct corrective action.
• User can act on a warning in one or two taps.
• The product explains what will happen if the user ignores the warning.

​

Definition of done

• Warnings arrive early enough to enable realistic recovery.
• Users can distinguish soft nudges from strong alerts at a glance.
• Each warning clearly links current behavior to future outcome.
• Corrective actions are concrete and achievable within the current cycle.
• The system avoids false urgency for small or temporary deviations.
• Edge cases are handled:
  • Missing or delayed income data
  • End-of-month shortfall that is too large to recover
  • User with multiple at-risk goals
  • User repeatedly ignores warnings

​

8.5 transaction intelligence layer

​

Purpose

​

Required insight types

• Time equivalent: how much work effort the purchase represents
• Opportunity cost: how the purchase affects goal timing or goal contributions

​

Trigger rules

​

Presentation model

• Insights must appear inline or as an expandable card, not as a chat response.
• The first layer must be visual and instantly scannable.
• The message must connect the transaction to a real-life trade-off.

​

Example presentation formats

​

Behavioral intent

• Slow down impulse spending
• Make abstract costs feel concrete
• Increase awareness without shaming
• Strengthen the connection between daily transactions and long-term goals

​

Acceptance criteria

• User sees transaction intelligence for materially relevant discretionary transactions.
• Insight appears in a visual, inline format that does not require typing.
• At least one insight connects the transaction to effort or goal trade-off.
• User can dismiss the insight or view more context.
• Repeated low-value prompts are reduced when the user shows disinterest.

​

Definition of done

• Users can instantly understand why the transaction insight matters.
• Insights feel helpful rather than punitive.
• The system focuses on high-value moments rather than every purchase.
• Goal impact framing is clear for users with active goals.
• User control exists to dismiss or ignore insights without breaking the experience.
• Edge cases are handled:
  • User has no active goals
  • User has irregular work pattern, making time-equivalent less useful
  • High-frequency low-value purchases create noise risk
  • Transaction category is uncertain or missing

​

8.6 gamification system

​

Purpose

​

Mission system

​

Reward system

​

Difficulty scaling

• Mission difficulty must increase when the user consistently succeeds.
• Mission difficulty must decrease when the user repeatedly fails or ignores missions.
• Missions must reflect the user’s current plan health and financial pressure.
• The system must avoid giving stretch missions when the user is already in risk recovery mode.

​

Psychological drivers

​

Acceptance criteria

• User receives AI-generated daily, weekly, and monthly missions tied to the active plan.
• Missions reflect current risk level, goals, and recent behavior.
• Mission progress is visible in the product.
• Rewards are granted when a mission is completed.
• Difficulty adjusts over time based on behavior and response patterns.

​

Definition of done

• Missions feel relevant rather than generic.
• Rewards reinforce healthy planning behavior rather than pure app activity.
• Difficulty scaling prevents both boredom and discouragement.
• Progress, streaks, and completions are visible without adding cognitive load.
• Users in recovery mode are not overloaded with unrealistic challenges.
• Edge cases are handled:
  • User misses several missions in a row
  • User has a no-spend streak but fails a goal contribution
  • User completes missions but ignores rebalance prompts
  • User wants low-intensity coaching

​

9. AI behavior and interaction model

​

9.1 AI triggers

​

9.2 AI actions

​

9.3 interaction model

​

Required interaction patterns

• Cards with one clear recommendation
• Quick actions such as Accept, Adjust, Skip, or View impact
• Sliders for allocation changes
• One-tap confirmations
• Expandable detail for users who want more context
• Minimal typing by default

​

Interaction rules

• Each prompt must have one primary action.
• Each prompt must present only the most important trade-off.
• High-priority prompts must override lower-priority prompts for the same cycle.
• If multiple interventions exist, the system must queue and consolidate them.

​

9.4 feedback loop

​

10. AI interaction flows

​

10.1 overspending recovery flow

Overspending detected
        |
        v
AI assesses severity and projected impact
        |
        v
Prompt card appears with recommended adjustment
        |
        +--> Accept --> plan updates --> confirmation shown
        |
        +--> Tweak --> user adjusts slider --> plan updates --> confirmation shown
        |
        +--> Reject --> plan remains --> AI reduces repetition and monitors outcome

​

Step-by-step

1. The system detects that spending pace in a flexible category exceeds the defined threshold.
2. The AI determines whether the issue is a soft drift or a strong recovery need.
3. The user receives a card explaining the issue and one recommended action.
4. The user accepts, tweaks, or rejects the suggestion.
5. The system updates the plan state and shows the immediate effect on goals and safe-to-spend money.
6. Future prompts adapt based on the user response and whether the risk persists.

​

10.2 goal risk prevention flow

Goal shortfall risk emerges
        |
        v
AI forecasts likely miss before end of cycle
        |
        v
Soft nudge or strong alert delivered
        |
        v
User chooses corrective action
        |
        v
Goal pace and plan status are refreshed

​

Step-by-step

1. The system identifies that the current pace is unlikely to meet the monthly goal target.
2. The AI selects the appropriate warning level.
3. The warning explains the likely outcome and suggests one or more corrective actions.
4. The user takes action or dismisses the warning.
5. The system updates the forecast and reflects whether the goal is back on track, still at risk, or reprioritized.

​

10.3 income event planning flow

Income arrives
        |
        v
AI creates or refreshes allocation proposal
        |
        v
User reviews plan with sliders and quick actions
        |
        v
Plan activated
        |
        v
Goals, missions, and checkpoints begin for that cycle

​

Step-by-step

1. A new income event or cycle start triggers the planning experience.
2. The AI proposes an allocation plan based on user context and active priorities.
3. The user accepts or adjusts the recommendation.
4. The system confirms the active plan and highlights what is protected, flexible, and goal-directed.
5. Goal pacing, warning readiness, and missions begin for the cycle.

​

10.4 transaction intelligence flow

Relevant transaction detected
        |
        v
AI determines whether insight adds value
        |
        v
Inline insight card shown
        |
        +--> Dismiss
        |
        +--> View impact --> see effort and goal trade-off

​

Step-by-step

1. A materially relevant discretionary transaction occurs.
2. The AI evaluates whether the transaction should trigger insight based on amount, pattern, and current plan health.
3. A lightweight visual insight appears inline or inside the transaction detail.
4. The user dismisses it or views more context.
5. The AI adjusts future insight volume based on user engagement.

​

11. edge cases and constraints

​

Business constraints

• Recommendations must be explainable in plain language.
• Users must retain final control over plan changes.
• The experience must minimize typing and avoid chat-style dependency.
• AI action volume must be managed to prevent fatigue.
• Suggestions must remain realistic for the user’s financial context, not aspirational to the point of failure.

​

12. success criteria

​

12.1 user behavior metrics

​

12.2 engagement metrics

​

12.3 business metrics

​

13. release acceptance summary

• Users can create or activate a plan with minimal typing.
• AI recommendations are understandable and actionable in one to two interactions.
• The system can proactively suggest, warn, explain, and celebrate through UI-led interactions.
• The six core features operate as one connected planning experience rather than isolated tools.
• Risk and recovery flows are clearly defined for normal, low-data, and high-volatility scenarios.
• Success metrics and operational targets are agreed by stakeholders before launch.

​

14. open product decisions

• What level of coaching intensity should be default for new users?
• Which rewards have the strongest product value: unlocked spending freedom, visual status, or goal acceleration?
• How aggressive should prompt suppression be after repeated user rejection?
• Which goal types should be prioritized in the first commercial release?
• Should Plans be available to all users by default or introduced after a readiness signal such as income detection?

​

15. downstream team guidance

​

For design

• Design the experience as a sequence of lightweight decisions, not a dense dashboard.
• Prioritize clear trade-off communication over decorative complexity.
• Make warning severity and plan safety instantly recognizable.

​

For engineering

• Preserve the business order of operations defined in this document.
• Ensure each user-visible state can support accept, adjust, reject, dismiss, and confirmation patterns where required.
• Maintain consistency across planning, warning, rebalancing, and reward surfaces.

​

For stakeholders

• Evaluate Plans as a habit-forming guidance system, not only as a budgeting utility.
• Judge early success by response quality and repeat engagement, not only by setup volume.
• Prioritize trust, clarity, and realistic guidance over overly aggressive optimization.