User Manual · v1.0

Everything you need to operate Preemptify AI.

A complete operator's guide — feature walkthroughs, math behind every metric, an exhaustive glossary, and an FAQ. Stuck? Press the Ask Preemptify button bottom-right at any time.

Overview

Preemptify AI is the AI Predictive Risk & Balance Sheet Intelligence platform for retail wealth, corporate treasury, trade finance and mutual fund books. Instead of waiting for market events to materialise, the Neural Engine maps the current financial environment directly against the client balance sheet, identifying structural vulnerabilities and capital mismatches before they materialize — and fuses central-bank + treasury signals with proprietary balance-sheet math to surface preempted product recommendations, each one citing the source headline and the local math rationale.

It ships with two operating tracks (Retail Wealth and Corporate Treasury), an admin console (Policy Compiler, Signal Calibration, Schema Aligner, Product Catalog), and a built-in backtest engine for evaluating strategy variants.

The platform never says “trust me.” Every recommendation carries the numeric drivers that produced it.

Getting Started

1. Sign in at /login using one of the demo identities below. Clicking the demo card auto-fills + auto-submits.

# Demo identities
executive_demo    / executive_demo     → Retail + Corporate access
compliance_admin  / compliance_admin   → All access including /admin*

2. Land on the dashboard at /sandbox. Switch between Retail Wealth and Corporate Treasury in the top bar.

3. Click any row in the customer/corporate list to open the detail panel — metrics, decay clocks, NBA cards, AI Analyst chat, Smart Analysis, and the goal/life-event triggers.

4. Drill into admin tools via the top-right links: /admin (Policy Compiler), /admin/signals (Signal Calibration), /admin/aligner (Schema Aligner), /admin/products (Product Catalog).

Part I

The Four OS Modules

Four standalone Industry Operating Systems. Each is licensable independently. All four interoperate on one cryptographic audit spine with AsyncLocalStorage tenant isolation and a unified PII-redaction boundary.

🏛

Banking OS

/sandbox
Retail · Corporate · Treasury · Compliance
Standalone · Interoperable

The flagship balance-sheet module. Retail and corporate exposures on one chassis, wired to the cryptographic audit ledger.

4
Pillars covered
20+
Currencies (MCC)
21
Regulators bound
88+
Product templates
Lifecycle stages
Customer 360NBA EngineDecay ClocksTreasury MathHedge ComposerAudit Ledger
AI capabilities
Retail Customer 360 (3-column Liability · Asset · Insurance)
Liquid Cushion Decay Clocks (CASA · TD · RD)
Corporate Treasury Engine (CCC · FX-mismatch · runway)
Trade Finance Hedge Composer (LC · BD · IF · SCF · FX Fwd · IRS · XCS)
Multi-Currency Core (20+ currencies · ISO-20022 ↔ SWIFT MT)
Policy Compiler · Schema Aligner · Workflow + Cron Engine
Mortgage Repricing · Credit Card AI · Personal Loan UW
5-strategy Backtest (Static · Decay · Preempted NBA · Coefficient · Glidepath)
Featured Simulator · Signal-Driven NBA Backtest
Inputs
Signal regime · 6-60 month window · 5 strategies
Outputs
CAGR · Sharpe · max drawdown · hit-rate · cited rationale
Roles · Bank Admin · Compliance · RM · Executive
Regulators · RBI · BOT · MAS · ECB · Fed · OCC · BNM · FINMA · BoE · BoJ · PBoC
💳

Lending OS

/lending
Origination → Underwriting → Collections → Restructuring
Standalone · Interoperable

End-to-end loan lifecycle. AI from application to recovery with a prompt-driven product catalog and JSON rules engine.

8+
AI co-pilots
1
What-if simulator
4
Module roles
6+
Regulators bound
Lifecycle stages
ApplicationUnderwriting AIDisbursalDPD TrackCollectionsRestructure
AI capabilities
Filterable Customer Workbench (name · token · tier · city · product · risk · DPD)
Per-borrower 360 (loans · DPD · credit · exposure · mix donut)
AI Underwriting Co-pilot (decision · confidence · APR · risk factors)
DPD Early-Warning Briefs (severity · root cause · actions · settle floor)
Restructuring Proposals (tenor · moratorium · rate cut · haircut · settlement)
Collections Priority Queue (deterministic green/amber/red)
AI Risk Assessment + Multi-turn Ask-AI per borrower
Custom Product Catalog (6+ templates) · JSON Rules Engine (8+ rules)
Featured Simulator · Credit Stress Simulator
Inputs
Rate shock (bps) · GDP shock (%) · Unemployment shock (%)
Outputs
NPA delta · expected loss (LGD-weighted) · capital consumption % · DPD migration chart · AI commentary
Roles · Underwriter · Collections · Lending Admin · Lending Read
Regulators · RBI · OCC · MAS · BOT · BNM · ECB
🛡

Insurance OS

/insurance
Underwriting → 5-state Claims → Lapse Revival
Standalone · Interoperable

Cat-aware, IRDAI-aligned. AI from quote to claim to revival with persistency cohort dashboard and NatCat stress simulator.

7+
AI co-pilots
2
Simulators
5
Module roles
5+
Regulators bound
Lifecycle stages
QuoteUW ProfileIssuePremium / LapseClaim 5-stateRevival
AI capabilities
Filterable Policyholder Workbench (name · LoB · premium · overdue · claim status)
Per-policyholder 360 (policies · sum insured · claims · lapse risk · LoB mix)
AI Risk Profiler (uw_class · premium loading · medical UW · regulator flags)
5-state Claims Machine (filed → under_review → escalated_fraud / approved → disbursed)
AI Claims Anomaly Detector (score · patterns · investigation priority)
Lapse Revival Brief (urgency · grace · waiver % · retention escalation)
Per-policyholder AI Risk Assessment + Multi-turn Ask-AI
NBA: retention · cross-sell · claim follow-up
Featured Simulator · Catastrophe Stress + Persistency Cohort
Inputs
Cat event (flood/pandemic/cyclone/wildfire/cyber) · severity 1-5 · auto cohort
Outputs
Projected claims · new loss ratio · premium-at-risk · reserves-breach flag · m13/m25/m37/m49/m61 retention
Roles · UW · Claims Officer · Persistency · Insurance Admin · Insurance Read
Regulators · IRDAI · MAS · NAIC · Lloyd's · FCA
📈

Wealth OS

/wealth
Advisory · MF · PMS · AIF · Demat
Standalone · Interoperable

Mandate-locked, drift-free. AI from suitability to allocation to rebalance with a live What-If rebalance simulator.

7+
AI co-pilots
1
What-if simulator
5
Module roles
5+
Regulators bound
Lifecycle stages
OnboardRisk ProfileAllocateDrift WatchRebalanceCapital Risk
AI capabilities
Filterable Client Workbench (name · strategy · drift % · NAV)
Per-client 360 (portfolios · drift % · AIF drawn · XIRR · risk profile)
Mandate-Drift Analyzer + Capital-Risk Co-pilot
AI Rebalance Briefs (drift severity · proposed actions · correction %)
Risk-Questionnaire Classifier (qualitative → strict equity/debt/alts)
Suitability Scorer (mandate · drift · unhedged · churn risk)
Per-client AI Suitability Assessment + Multi-turn Ask-AI
NBA: rebalancing · AIF drawdowns · diversification · hedge triggers
Featured Simulator · Mandate Rebalance What-If
Inputs
Target drift slider (%) · action (sell / buy / both)
Outputs
Tx cost (25 bps) · tax drag (~6% × 15%) · NAV impact % · projected breach count · verdict (execute / phase / hold / abort)
Roles · RM · Wealth Admin · Compliance · Wealth Read
Regulators · SEBI · MAS · SFC · FINMA · SEC
Banking OS · Deep Reference

The sections below dive deeper into Banking OS internals — retail wealth, corporate treasury, NBA mechanics, signal preemption, goals, backtests, schema aligner, audit ledger and admin tools. Lending / Insurance / Wealth deep references live inside each module's /help tab.

Retail Wealth Pipeline

The retail pipeline tokenizes individual customers and surfaces preemptive recommendations across savings, lending, insurance, and investment products.

Key metrics

Liquid Cushion
Checking + savings + short-tenor term deposits. Foundation of the decay clock.
Decay Clock
Days until the cushion depletes at current burn. Emerald > 180d · amber 90–180d · red < 90d.
Risk Score
Composite 0–100 across liquidity, leverage, coverage, sector exposure, macro context.
Insurance Gaps
Unhedged exposure tags (LIFE, HEALTH, HOME, CREDIT_LIFE, DISABILITY) with severity + exposure amount.
Goals
Auto-generated from profile (Emergency Fund, Retirement, Home, Education, etc.) — see Goals & Life Events.

Corporate Treasury

The corporate track adds sector-aware treasury intelligence, working-capital diagnostics, and trade-finance product recommendations.

Cash Conversion Cycle (CCC)
DIO + DSO − DPO, in days. > 90d is amber.
Counterparty Concentration
% revenue from the top counterparty. > 45 % amber, > 60 % critical.
FX Mismatch
% of revenue/cost in a foreign currency vs reporting currency. Drives FX-hedge NBAs.
Trade Finance / LC Utilisation
% of available line drawn. Surfaces credit-tightening preempts.
Cash Burn / Runway
Monthly burn vs available liquidity → months of runway.

AI Predictive Recommendations (NBA)

The NBA (Next-Best-Action) engine is the heart of the product. Each entity detail returns a ranked list of recommendation cards. Three kinds:

  • Base NBA — derived purely from the entity's own metrics (low cushion, unhedged exposure, idle deposit).
  • Goal-driven NBA — generated when a goal projects below target. Shows the shortfall and required monthly contribution.
  • Preempted NBA — injected by the Signal Preemption engine. Carries the ⚡ amber badge, the source headline, and the math rationale (rates_pressure +0.43 · source weight 1.40× · confidence 0.91 · …).

The order is: Preempted cards first (sorted by priority), then base/goal cards, capped at 10 per entity.

Signal Preemption Engine

The four-stage pipeline:

1.  Ingest     → RSS pull from 24+ central banks + treasuries
2.  Classify   → Preemptify Neural extracts:
                  policy_direction, magnitude, time_horizon,
                  asset_impact[], affected_sectors[], confidence
3.  Aggregate  → weighted preempt-vector across rates / fx / credit
                  / equity / commodities + per-sector pressure
                  (source × confidence × magnitude × time-decay)
4.  Recommend  → pressure > calibrated threshold triggers a
                  "Preempted" NBA card with full math rationale.

Visit /admin/signals to: refresh the feed, tune source / asset / magnitude weights, view the live preempt-vector, and inspect each classified headline with its sectors and asset impacts.

Goals & Life Events

Each customer auto-generates a deterministic goal set from their profile (age, segment, tier, mortgage status). Default goals include Emergency Fund, Retirement Corpus, Home Down Payment, Wedding, Travel, Legacy Planning, and more.

Visit /sandbox/goals to pick a customer, review progress bars + projected future value vs target, then trigger any of 9 life events:

marriage
Adds Honeymoon Fund + LIFE insurance gap
child_born
Adds Child Education Fund + LIFE & HEALTH gaps
home_purchase
Removes the home-down-payment goal + adds HOME & CREDIT_LIFE gaps
job_loss
Pauses all goal contributions + adds DISABILITY gap
promotion
Boosts retirement contribution by 1.5×
retirement_5y
Shortens retirement horizon + adds HEALTH gap
divorce
Halves balances and targets + adds LIFE gap
inheritance
Distributes a windfall 60 / 40 between retirement and legacy
education_start
Adds an Education Loan Repayment goal + DISABILITY gap

Historical Backtest Engine

Five deterministic strategies, side-by-side, 6–60 months. Visit /sandbox/backtest.

Static Policy
Buy-and-hold synthetic balanced portfolio, no rebalancing — the baseline.
Liquid Cushion Decay
Defensive: exit equity when decay-days < 90.
Preempted NBA
Rotates based on the signal preempt-vector — this is the “hero” strategy.
Policy Coefficient
Tactical tilt toward lower-coefficient (less risk-weighted) sectors.
Goal-aligned Glidepath
Time-horizon weighted equity allocation — longer goal ⇒ higher equity.

Each result returns equity curve, drawdown trajectory, CAGR, Sharpe, max-drawdown, hit-rate, and annualised volatility. Output is deterministic per (entity × strategy) seed.

Schema Aligner

Visit /admin/aligner. Upload (or paste) a sample of your legacy CSV / JSON / fixed-width data. The aligner builds a column-to-canonical-field mapping manifest, you commit the manifest, and the commit is anchored on the Audit Ledger with a SHA-256 hash.

Manifests are versioned, addressable by schema_alignment_id, and re-usable across batches.

Audit Ledger

Every action that mutates state — policy compiles, schema commits, signal calibration changes, goal creates/deletes, life events, backtest runs, seed events — produces a ledger entry with: timestamp, action, prev_hash, current_hash, and event-specific details.

The chain is tamper-evident: changing any historical entry breaks the SHA-256 cascade. In production the chain replicates to your WORM (write-once-read-many) storage tier so any tampering is provably detectable by your own auditors.

Admin · Policy Compiler

The Policy Compiler at /admin lets compliance officers re-weight sector risk coefficients live. Type or paste a directive like “increase real-estate sector risk weights by 10 %” and the engine produces a JSON delta, attaches it to the active policy, and anchors the change to the ledger.

Affected sectors are surfaced in the top bar as N sector(s) under active policy — click through to review the live coefficients.

Localization · 80+ jurisdictions

The country picker in the top bar drives:

  • Currency formatting throughout the UI (via Intl.NumberFormat)
  • Product catalog filtering (target_country / target_currency)
  • Signal Console priority sorting (the user's country's signals surface first)
  • Regulator-pack citation style (RBI vs MAS vs OCC)

Security & Privacy

Zero data egress
On-Premise Pilot / Enterprise+ deploys inside your VPC. Only the LLM gateway makes outbound calls — and even that can be routed through your own air-gapped inference endpoint.
Tokenized PII
No raw customer identifiers leave the perimeter. All references use opaque tokens (customer_token, entity_token).
SHA-256 ledger
Every state change is hash-chain anchored. Replicated to WORM in production.
RBAC
Two demo roles — executive_demo (retail + corporate) and compliance_admin (all access including /admin*).
Right-to-erasure
GDPR / DPDP Act / CCPA — write to info@preemptify.com.

Keyboard Shortcuts

?Open the Help Agent chat
/Focus the customer search bar
escClose any open modal or sheet
g gGo to dashboard (/sandbox)
g aGo to admin (/admin)
g sGo to Signal Console (/admin/signals)
g mGo to user manual (/manual)

Note: shortcut handlers are progressively rolled out — verify availability in your build.

Glossary

Audit Ledger
SHA-256 hash-chained append-only event log. Every policy compile, schema commit, signal calibration change, life event, and seed event is anchored. In production, the chain replicates to your WORM storage so any tampering is provably detectable.
Backtest Strategy
A deterministic simulation rule used to compare hypothetical portfolio behaviour over 6–60 months. Five strategies ship out-of-the-box: Static Policy, Liquid Cushion Decay, Preempted NBA, Policy Coefficient, Goal-aligned Glidepath.
Cash Conversion Cycle (CCC)
For corporate clients — DIO + DSO − DPO, measured in days. A high CCC (> 90d) indicates working capital is tied up longer than ideal and is flagged amber on the dashboard.
Counterparty Concentration
For corporate clients — percentage of total revenue derived from the top counterparty (or top 5). > 45 % triggers an amber risk indicator; > 60 % is critical.
CAGR
Compound Annual Growth Rate — annualised geometric mean return of a portfolio or strategy over the backtest window.
Decay Clock
Projected number of days until a customer's Liquid Cushion depletes at their current burn rate. Three-stage indicator: emerald (> 180d), amber (90–180d), red (< 90d).
DPO / DIO / DSO
Days Payable Outstanding (you owe), Days Inventory Outstanding (you hold), Days Sales Outstanding (you are owed). Components of the Cash Conversion Cycle.
Drawdown
Maximum peak-to-trough decline in portfolio value during the backtest window, expressed as a negative percentage.
FX Mismatch
Percentage of revenue or cost denominated in a foreign currency vs the reporting currency. Drives FX-hedge product recommendations.
Goal Status
Forward-projected health of a financial goal: on_track (≥ 0 % surplus), monitor (−10 to 0 %), at_risk (−25 to −10 %), lagging (worse than −25 %).
Hit Rate
Percentage of months during the backtest where the strategy posted a positive return.
Insurance Gap
Unhedged exposure on a retail balance sheet. Types: LIFE, HEALTH, HOME, CREDIT_LIFE, DISABILITY. Each is shown with severity and exposure amount.
Liquid Cushion
Checking + Savings + short-tenor Term Deposits. The foundation of the Decay Clock and many NBA triggers.
Life Event
Triggerable customer-state change (marriage, child_born, home_purchase, job_loss, promotion, retirement_5y, divorce, inheritance, education_start). Each mutates goals and may add insurance gaps.
Magnitude (of a signal)
0–100 score the Preemptify Neural classifier assigns to a news headline reflecting its market-moving potential.
NBA
Next-Best-Action — a product or intervention recommendation generated from the customer's metrics, life events, and the active signal context. Cards may be Preempted (signal-driven), Goal-driven, or Base.
Policy Coefficient
A sector-level risk weight (e.g. Real Estate Dev = 1.2 ×, Pharma = 0.9 ×). Compliance admins can re-weight live via the Policy Compiler.
Policy Direction
Classification output of a news signal: hawkish | dovish | neutral | risk_off | risk_on. Drives the sign of the rates/credit pressure vector.
Preempted NBA
An NBA card that was injected by the Signal Preemption engine — carries the ⚡ amber "PREEMPTED" badge, the source headline, and the math rationale.
Preempt-Vector
Six-axis aggregate of all active signals: rates_pressure, fx_pressure, credit_pressure, equity_pressure, commodity_pressure, and per-sector pressure map. Positive values are tightening / risk-off; negative are easing / risk-on.
Risk Score
Composite 0–100 score per entity, weighted across liquidity, leverage, coverage, sector exposure, and macro context.
Schema Aligner
Tool at /admin/aligner that maps legacy CSV / data formats to Preemptify's canonical balance-sheet schema. Each commit is anchored on the Audit Ledger.
Sharpe Ratio
Risk-adjusted return — (mean monthly return − risk-free rate) / std-dev × √12. > 0.5 is good, > 1.0 is excellent on the backtest charts.
Signal Calibration
Admin tuning at /admin/signals controlling per-source trust, per-asset-class weight, magnitude amplifier, time-decay window, min-confidence threshold, and preempt firing thresholds.
Smart Analysis
AI-generated 3-paragraph balance-sheet briefing (risk band + drivers + recommendations) on a specific customer or corporate, available from each detail panel.
Tier (Retail)
Customer wealth band: Standard | Affluent | Private. Drives product eligibility and default goal templates.
Time-Decay Window
Hours after a signal is published before it stops contributing to the preempt-vector. Default 96 hours.
Volatility (annualised)
Standard deviation of monthly returns × √12 — used to size risk-adjusted return metrics.

FAQ

What does Preemptify do that a traditional CRM doesn't?
Traditional CRMs are descriptive — they tell you what your customers look like today. Preemptify is preemptive: it ingests live central-bank and treasury signals, runs them through the Preemptify Neural classifier, fuses them with proprietary risk math, and surfaces product recommendations BEFORE the loss event materialises.
What data do you collect? Does it leave my perimeter?
On the Public Sandbox (Free) you interact with synthetic data only — no real customer PII is processed. On On-Premise Pilot / Enterprise+ the engine deploys inside your VPC with zero data egress; only the LLM gateway makes outbound calls and even that can be routed through your own air-gapped inference endpoint.
Where do the news signals come from?
Real RSS feeds from 24+ central banks and treasuries: Fed, ECB, BoE, BoJ, RBI, PBoC, MAS, BOT (Bank of Thailand), BSP, SBV, BCB, Banxico, SARB, SAMA, SNB, BoK, BoC, RBA, BoI, plus US Treasury and others. The full list is visible in /admin/signals.
How does the AI know about my specific product catalog?
The Product Catalog at /admin/products is the source of truth. Each product carries target country, target currency, eligibility criteria, and risk envelope. The NBA engine joins live product entries to triggered recommendations.
What's the difference between an NBA card and a Preempted NBA card?
A standard NBA card is generated from the entity's own metrics (e.g. low liquid cushion, unhedged exposure). A Preempted NBA card is additionally driven by an external signal — it carries the ⚡ amber badge and cites the source headline plus the math rationale.
Can compliance officers tune sensitivity?
Yes — /admin/signals exposes 30+ sliders. Per-source trust weights, per-asset-class weights (rates/fx/credit/equity/commodities), preempt firing thresholds, magnitude amplifier, time-decay window, and minimum confidence cutoff. All changes are ledger-anchored.
How is the demo data generated?
Deterministic synthesis — 50 retail customers and 20 corporate clients are generated from seeded RNG using US cities and USD currency. Trade-finance metrics, decay clocks, insurance gaps, and goal templates are derived from each entity's profile.
What if a backtest disagrees with my preferred strategy?
Each backtest is deterministic (same entity + strategy ⇒ identical series), so disagreements come from the underlying math — review the equity curve, drawdown chart, and the Sharpe/CAGR table to understand why. You can also run only the strategies you want to compare.

Need help with something not covered here? Open the Ask Preemptify chat (bottom-right of every page) or write to info@preemptify.com.