Powered by the Patent-Pending DCR Engine

Deliver Personalized Financial Intelligence - Without Building an AI Platform

The DIY My Finances Decision Intelligence API gives you a hosted, privacy-first reasoning engine that computes facts, retrieves context, and explains decisions across tax, retirement, budget, investments, debt, and general finance - all through a simple API.

Request API AccessComing soon
See Example Requests

Why Personal Finance AI Keeps Failing

Personal finance is not a knowledge problem. It is a decision problem.

Decision Blindness
Dashboards don't decide.
Numbers show up, but they do not tell the user what matters now.
Brittle Logic
Rules break as life changes.
Rule engines collapse when user conditions shift.
Unverified AI
Confident language, weak truth.
Chat tools sound right but fail numeric scrutiny, and answers cannot be traced to computed facts.
Compliance & Privacy
PII risk shuts projects down.
AI that stores documents or trains on user data triggers compliance pushback.

This is not a UX problem. It is an architecture problem.

What It Does

The Decision Intelligence API combines deterministic financial computation, hybrid retrieval-augmented generation (HYBRID-RAG), and bounded AI reasoning to deliver personalized financial guidance grounded in verified facts.

This is not document chat. It is a decision system.

Deterministic KPIs
Compute verified financial facts for tax, budget, cash flow, and savings.
Verified Decision State
Transform user data into an authoritative decision state.
Controlled Reasoning
Explain decisions within constraints defined by verified facts.
Privacy & Compliance
No raw document storage. No full text retention. PII scrubbed before embedding.

Decision Intelligence, Not Document Search

Most AI systems attempt to find answers in text. The Decision Intelligence API does something fundamentally different:

It computes the decision first - then explains it.

User data and documents are transformed into a verified decision state using deterministic domain engines (tax, budget, cash flow, savings, etc.). Only after the facts are established does the system retrieve semantic context and generate explanations.

User data and documents to Deterministic KPIs to Verified decision state to Context retrieval to Controlled AI reasoning to Explanation of the decision

Language never determines truth. It explains what has already been computed.

Why This Is Different

From Retrieval to Decision-Centric Reasoning

Personal finance imposes constraints generic AI systems were not built for. Answers must be numerically correct, outcomes depend on changing user state, trust and compliance require explainability, and privacy limits document retention. That is why DCR exists.

Traditional RAG assumes the answer already exists in documents:

User question to retrieve text to generate answer
Traditional RAG works for:
FAQsManualsKnowledge bases
But it fails for:
Personal financeComplianceDiagnosticsPlanningOptimization
The DCR Shift: Decisions Come First
User data->
Deterministic KPIs->
Decision state->
Context retrieval->
Controlled reasoning->
Explanation

The authority is no longer text. It is the computed financial state of the user. The AI does not search for an answer. It explains a verified decision.

PropertyTraditional RAGDCR (This Platform)
What exists firstDocumentsDecisions
What is retrievedTextContext for a state
What changesDocumentsUser conditions
What must be correctLanguageComputation
Role of the modelInferExplain

In DCR:

  • The decision exists before language
  • The LLM is downstream of logic, not upstream

This is bounded reasoning.

The system still synthesizes, weighs tradeoffs, provides context, and explains implications.

But it does so within constraints defined by verified facts.

  • Not freeform
  • Not probabilistic guessing
  • Not tool-driven hallucination

The Mental Model Shift

RAG mindset: "Find text that looks like the answer."

DCR mindset: "Compute the truth - then explain it."

DCR is a new reasoning category, built for decisions where correctness, trust, and accountability matter.

DCR is for organizations where AI must explain decisions - not make them.

Fintech & Neobank Product Teams
Personal finance at scale
Fintech and neobank teams already compute KPIs, but users ask "what should I do?" DCR keeps the math authoritative and turns AI into a decision explainer.
Recommendations come from verified calculations, not AI opinion.
Advisory & Wealth Platforms
Human-in-the-loop decision systems
Advisors are accountable and need explainable, auditable outputs. DCR lets AI explain verified states without generating advice.
AI-assisted advice without AI-generated advice.
Compliance-Heavy Financial Software
Tax, credit, underwriting, insurance
Compliance-heavy systems require numeric correctness and audit trails. DCR keeps computation deterministic and layers explanation safely.
Explainability without inference.
Enterprise SaaS with Embedded Finance
Finance is not the core product - but still critical
Enterprise platforms want "what should I do?" guidance without building finance engines. DCR delivers decision intelligence as a service, without liability.
Add guidance without owning the risk.

Integration Options

We license our patent-pending decision intelligence engine (DCR) that computes truth first, then explains it - without storing your customers' documents.

  • No models to manage
  • No vectors to maintain
  • No prompts to secure
  • No documents to store

You call the API. We run the intelligence layer.

Platform License
(Business-first, predictable pricing)
What it includes
  • Hosted DCR platform
  • Secure tenant isolation
  • Privacy-preserving document intelligence
  • Deterministic domain KPI engines
  • Controlled AI reasoning layer
  • Support & monitoring

No per-token or per-call math in the contract. "Normal business usage" is included.

Compliance, Privacy & Audit Readiness

Compliance Overview: Built for Regulated Financial Environments

Financial institutions retain responsibility for governance, model risk, privacy protection, and operational control. The Decision Intelligence API is designed to support those requirements through deterministic decision authority, bounded explanation, and structured execution artifacts.

This is not a chatbot transcript system. It is a replayable, verifiable decisioning layer engineered for regulated personalization.

1. Alignment with Governance Expectations

Regulated financial institutions operate under established supervisory expectations and emerging AI risk frameworks, including:

  • Model Risk Management principles (e.g., SR 11-7, OCC 2011-12)
  • NIST AI Risk Management Framework
  • Sector-specific AI governance initiatives
  • Risk-based regulatory approaches (e.g., EU AI Act where applicable)

While each institution applies its own internal controls and jurisdictional requirements, the architecture is designed to support:

  • Reproducibility under version control
  • Effective challenge and independent review
  • Clear decision provenance
  • Data minimization
  • Operational transparency
2. Deterministic Decision Authority (DCR)

Decision-Centric Reasoning (DCR) ensures that:

  • Financial calculations occur in deterministic, versioned domain engines
  • The system produces a structured Decision Contract before generating narrative explanation
  • Engine version and assumptions version are explicitly attached
  • Identical input snapshots produce identical decision states

This allows replayability and audit inspection independent of language generation.

3. LLM Role & Model Agnosticism

The platform uses commercial or institution-approved LLMs as a bounded explanation layer. Decision authority remains in deterministic DCR engines. The LLM does not compute financial outcomes and cannot override the decision contract. The architecture is model-agnostic and does not require institutions to depend on a specific model provider.

4. Privacy & PII Controls (Default Posture)

Privacy is enforced as runtime behavior, not only documentation.

  • No raw document retention by default
  • No full-text document corpus
  • PII scrubbing before embeddings or reasoning context
  • Explicit scrub level and PII scrubbing flag recorded per execution
  • Tenant-level isolation of artifacts

Documents can be processed for extraction and then discarded. The system is designed to avoid becoming a document vault.

5. Hybrid-RAG: Bounded Evidence, Not Uncontrolled Retrieval

Hybrid-RAG operates subordinate to deterministic decisioning.

Retrieval is permitted to:
  • Surface policy references
  • Provide educational context
  • Attach structured evidence metadata
Retrieval is not permitted to:
  • Compute financial deltas
  • Infer personalized outcomes
  • Override deterministic decision states

This separation prevents probabilistic inference from becoming decision authority.

6. Execution Trace & Integrity Anchoring

A regulated decision system must answer four questions: what happened, what inputs were used, which policies were in force, and can we prove the output wasn't altered. Every run in DCR/Hybrid-RAG produces a trace record designed as audit evidence, not application logging.

A typical trace captures, at minimum:

  • A unique execution identifier for the run
  • Integrity anchors for the request and response (hash-based)
  • Runtime policy snapshot (privacy, retrieval, & processing)
  • The decision scope (which approved domains were invoked)
  • Evidence provenance (references + relevance signals)
  • The privacy posture applied (scrub mode / redaction enforcement)

This is the difference between "we log activity" and a verifiable system of record: the institution can replay a decision under the same versions, validate governance controls, and confirm the returned output matches the original execution - without turning raw customer documents or free-form transcripts into retained compliance liabilities.

7. Replayability & Version Control

DCR enforces deterministic replay under version control:

  • Input snapshot captured via Financial State Model
  • Domain engine version attached
  • Assumptions version attached
  • Narrative generated only after decision contract exists

Compliance teams can review the structured decision state without relying on model inference.

8. Operational Controls

Platform-level controls include:

  • Secure tenant isolation
  • API-based integration behind institutional identity layers
  • Monitoring & support
  • Version management for deterministic engines

Institutions retain control of:

  • User authentication
  • Input data sources
  • Policy thresholds (where configured)
9. What a Compliance Officer Can Actually Review

The platform supports inspection of:

  • Usage & activity logs
  • Privacy posture signals
  • Decision state object
  • Evidence references
  • Integrity hashes
  • Policy configuration at runtime

This enables governance teams to evaluate AI-assisted personalization without relying on probabilistic black-box outputs.

10. Scope Boundaries (Important for Risk Teams)

In scope:

  • Tax scenario deltas
  • Budget / cash-flow planning
  • Debt payoff prioritization
  • Retirement projections
  • Investment allocation alignment checks

Out of scope:

  • Credit underwriting
  • Approve/deny decisions
  • Credit scoring
  • Loan eligibility determinations

The system is not offered for credit underwriting or approval workflows.

Sample API

Domain
budget
Request
POST /api/hre/ask
{
  "domain": "budget",
  "question": "Given my budget, can I invest more money?"
}
Response
Your current budget shows a significant overspend relative to your income, resulting in a negative savings rate. This means you are not generating free cash flow to invest more money at this time. (more)
Your current budget shows a significant overspend relative to your income, resulting in a negative savings rate. This means you are not generating free cash flow to invest more money at this time.

Key Budget Insights
Monthly post-tax income: $8,967
Planned monthly spending: $11,854 (33% higher than income)
Savings rate: -42.0% (negative, indicating overspending)
Target savings: $1,793/month (20% of income)
Free-to-spend after savings: -$2,887 (deficit, no surplus cash)
Discretionary spending share: 14.3% of planned spend
Peer Comparison
Your monthly expenses exceed peers by $3,976 (50.5% more), indicating room to reduce spending.
Category-Level Spending Highlights (monthly)
Category  Your Spend  Peer Average  Notes
Food  $1,329  $1,329  At peer average
Utilities  $503  $503  At peer average
Transportation  $518  $518  At peer average
Domain
tax
Request
POST /api/hre/ask
{
  "domain": "tax",
  "question": "Which deductions or credits am I likely missing?"
}
Response
Your tax profile shows you are married filing jointly with an AGI of $413,446, using the standard deduction of $29,200, and claiming a Child Tax Credit of $3,300. (more)
Your tax profile shows you are married filing jointly with an AGI of $413,446, using the standard deduction of $29,200, and claiming a Child Tax Credit of $3,300. Your total tax is $77,180 with federal withholding of $68,605, leaving $8,575 owed. Based on this, here are potential deductions or credits you might be missing or could consider:

Potential Missing Deductions
Itemized Deductions:
Since you took the standard deduction ($29,200), check if your itemizable expenses (mortgage interest, state/local taxes, charitable donations, medical expenses) exceed this amount. If yes, itemizing could reduce taxable income further.

Retirement Contributions:
Maximize contributions to traditional 401(k), IRA, or other tax-deferred accounts to lower AGI and taxable income.

Health Savings Account (HSA):
If eligible, contributions to an HSA are deductible and reduce taxable income.

Education-related Deductions:
If you or dependents have education expenses, deductions or credits like the Lifetime Learning Credit or American Opportunity Credit might apply.

Potential Missing Credits
Additional Child Tax Credit (ACTC):
You have a Child Tax Credit of $3,300; verify if you qualify for the refundable portion (ACTC) if your tax liability is low relative to the credit.
Domain
retirement
Request
POST /api/hre/ask
{
  "domain": "retirement",
  "question": "Provide a summary of my retirement plan."
}
Response
Your retirement plan projects a nest egg of approximately $2.94 million at retirement, slightly below your target by about $40,600. (more)
Your retirement plan projects a nest egg of approximately $2.94 million at retirement, which is slightly below your target by about $40,600. With a safe withdrawal rate of 4.0%, your portfolio is expected to generate around $117,400 annually, but this leaves an income gap of roughly $40,600 per year compared to your desired retirement income. Your savings are projected to last 30 years, both nominally and inflation-adjusted.

Retirement Plan Summary
Projected Retirement Assets: $2,935,276
Shortfall vs Goal: -$40,611 (meaning your nest egg is slightly under your target)
Safe Withdrawal Rate: 4.0%
Annual Portfolio Income: $117,411
Total Annual Retirement Income: $117,411
Annual Income Gap: -$40,611 (income needed beyond portfolio withdrawals)
Longevity of Savings: 30 years (nominal and inflation-adjusted)
This indicates your current savings and withdrawal strategy will sustain your retirement for about 30 years but may require either increasing savings, adjusting spending goals, or supplementing income to close the income gap.

API Portal

Manage API keys, usage analytics, domain activation, and support in one place.

Access API PortalComing soon

Trust & Authority

Patent Pending
Patent-Pending Decision-Centric Reasoning Architecture.
Privacy-First
Designed to operate without storing raw user documents.
Architecture-Based Trust
Deterministic + Semantic + Controlled Reasoning.
Enterprise Ready
Built for regulated environments, multi-tenant security, and auditability.