The Digital SME for Core Banking.
Forty years of COBOL. Three hundred million accounts. Nine-figure batch windows. A generation of mainframe engineers walking out the door. NCiS is the intelligent nervous system that captures, preserves, and modernizes what Hogan knows — before the people who know it retire.
Module Portfolio · by Category
◆ 45 · 6 categoriesDXC Licensing Posture
modernization gravityModernization Economics
traditional vs genaiLineage Pulse
live subgraphLive Intelligence Feed · Reconciliation & Operations
LIVEInside NCiS · six integrated views
click any panel to jumpTop Modules · by GenAI Opportunity Density
Σ (impact × complexity) across indexed casesFive failure modes converging on
every Tier-1 bank running Hogan.
Hogan — originally authored by Hogan Systems Inc. in the 1980s, acquired by CSC, now operated under DXC — remains the deposit and loan core for a meaningful slice of the top 20 US banks and credit unions worldwide. Its durability is its curse: every patch, every reg, every channel since 1988 is laminated onto a single COBOL/CICS/DB2 substrate. Five pressures now collide on it at once.
The SME Retirement Cliff
The engineers who wrote DDABAL, FEECAL, and CIFNAM are retiring at a pace that dwarfs new-hire training. Mainframe-fluent developers aged 55+ exceed those aged under 35 in virtually every bank's core team. Tribal knowledge — why a batch job kicks at 02:14 UTC, why GL10POST must precede STMTCYC — evaporates with every retirement letter.
The Shrinking Batch Window
Hogan was architected for a world where the bank closed at 17:00 local and the mainframe owned the night. 24×7 digital channels, Zelle, FedNow, ISO 20022 real-time rails, and global card presence have collapsed the nightly batch window from 8 hours to under 3 at most large institutions. EOD (End-of-Day) posting cycles, NSF determinations, and interest accruals now race the sun.
Regulatory Velocity
Dodd-Frank, CCAR, CECL, BSA/AML upgrades, Reg CC amendments, CFPB 1033, and ISO 20022 migration each demand core-system changes. Hogan absorbs them as overlay modules, nested IF-THEN legacy patches, and duplicate business logic scattered across programs. Each change increases surface area for undetected drift and regulatory breach.
AI-Native Challengers
Neobanks and fintechs build on event-sourced, API-first, cloud-native cores (Thought Machine, 10x Banking, Mambu, Temenos Transact R24). They ship product in weeks. Incumbents on Hogan ship in quarters. The gap is no longer feature set — it is time-to-market, personalization depth, and cost-to-serve. Every quarter the spread widens.
The Modernization Graveyard
Published outcomes of large core-replacement programs are sobering: 50–70% overshoot budget by more than 2×; a material minority are abandoned. TCS BaNCS, Finacle, Fiserv DNA, and Temenos each have live case studies of multi-year, multi-hundred-million-dollar programs that failed to cut over. Risk-averse CIOs defer the program — and the debt compounds.
Loss of Semantic Truth
After four decades of patches, "what the system actually does" diverges from "what the documentation says." Regulatory interpretations live in program comments. Business rules live in copybook defaults. Intent lives in 1994 design memos on a shared drive. Without a deterministic capture of current semantics, any modernization is gambling with the bank's books.
You cannot modernize what you cannot see.
Every failed Tier-1 modernization shares the same root cause: the organization began cutting before it could faithfully describe what the legacy system was doing. NCiS is designed to invert that order. Before a single line of Java is written, NCiS produces a complete, machine-verified ledger of Hogan's behavior — module by module, transaction by transaction, byte by byte — with every SME's tribal knowledge captured, attested, and promoted into executable tests.
Modernization becomes an audit, not a gamble.
With NCiS, every modernized component ships with proof: a reconciliation artifact showing bit-for-bit parity against the shadow legacy path over a regulator-defensible sample. Cutover becomes a ceremonial flip — the work was done already, under observation, with governance receipts. The $300M+ gamble becomes a staged, audit-preserving migration with rollback contracts and zero accounting surprise.
Six layers. One purpose.
Preserve every semantic of Hogan — auditably.
NCiS is not a translator, not a wrapper, not a chatbot. It is a purpose-built observability and modernization substrate for core banking — composed of six cooperating layers, each addressing one dimension of the preservation problem. Each layer is independently testable, independently deployable, and produces machine-readable artifacts that downstream layers consume.
What Hogan Already Contains परागृह्य — the sources of truth
A deterministic harvester for every artifact that describes Hogan's current behavior: COBOL source (including HLASM inserts), JCL, PROCs, copybooks, VSAM schema definitions, DB2 catalog (SYSIBM.SYSTABLES, SYSCOLUMNS, SYSVIEWS), CICS CSD exports, MQ channel definitions, batch scheduler (CA7 / TWS) job streams, IMS DBDs where present, operator logs, and — critically — SME voice, video, and chat transcripts captured under a structured interview protocol.
What Hogan Actually Does अन्वय — deterministic dependency truth
A graph-native semantic resolver that walks from ingested artifacts to a canonical directed acyclic graph of every entity, rule, and flow in Hogan. Copybook → program → paragraph → SQL call → table → column. CICS transaction → linked program → CALLed subroutine → commareas → MQ put. Business rule → regulatory obligation → program embedding. Provenance is preserved end-to-end: every node in the graph is traceable to an exact line of source.
Shadow Execution with Reconciliation प्रतिबिम्ब — proof under load
A production-grade replica of Hogan's execution semantics that receives a tap of real production traffic (anonymized where required), executes against both the legacy path and the modernized path, and reconciles outputs at the byte, journal, and balance-sheet level. A reconciliation variance triggers a structured diff and halts the candidate module's progression. Over time the twin becomes the authoritative spec; the source becomes commentary.
SME-Caliber Reasoning at Scale प्रज्ञा — LLM grounded in the graph
A constellation of LLM agents — grounded strictly in the lineage graph and digital twin — that perform the cognitive work previously locked in human SMEs. Explain why a program behaves as it does. Propose a modernized implementation with a generated test harness. Translate COBOL paragraphs to idiomatic Java/Kotlin with regulatory comments preserved. Answer "what breaks if I change this?" in seconds. Every answer ships with a citation set of source lines — no hallucination, no opacity.
Audit, Control, Regulatory Parity रक्षा — the bank's nervous system
An immutable ledger of every change, every reconciliation, every SME attestation, every LLM suggestion, every cutover receipt. Regulator-facing dashboards expose control effectiveness for SOX, FFIEC, SR 11-7 (model risk), BSA/AML coverage, CFPB 1033 readiness. When the OCC asks "are you sure?", the answer is a signed, cryptographically-anchored evidence package — not a PowerPoint.
Staged Cutover with Rollback Contracts संयम — the ceremony of migration
The conductor. Routes production traffic between legacy and modern implementations by percentage, by branch, by region, by customer cohort. Monitors latency, GL variance, and exception rates in real time. Executes dark launches, shadow routing, controlled rollouts, and — when needed — instant, cryptographically provable rollback to the previous known-parity state. Zero-downtime cutover becomes a scheduled event, not a weekend prayer.
Every layer emits receipts. Every receipt is an audit artifact.
NCiS's architectural rule, enforced by the Governance layer: no downstream layer may consume upstream output that is not cryptographically signed and provenance-linked. A Java microservice deployed by the Orchestration layer cannot exist without a Digital Twin parity receipt. A Digital Twin receipt cannot exist without a Lineage graph anchor. A Lineage anchor cannot exist without Ingestion provenance. The system is recursively audit-native — the way a bank general ledger is recursively debit-credit-native.
Forty-five functional territories.
Six categories. One Hogan installation.
Hogan is a confederation of modules — core accounting engines, DXC service wrappers, channel interfaces, regulatory feeds, and operational scaffolding. NCiS indexes every module's transaction surface, COBOL program footprint, DB2 table bindings, DXC licensing posture, and GenAI opportunities. Every module below is traceable to primary Hogan documentation. Click any tile to drill into sub-functions, technical details, simulation flows, and the GenAI opportunity set.
What the bank owns vs. what DXC owns.
Every Hogan module carries one of three DXC licensing postures that directly shapes the modernization path: LOCKED (source code sealed — modernization requires full replacement or DXC partnership), PARTIAL (configuration layer open, engine locked — parameter hygiene and overlay innovation possible), and OPEN (full source access — direct modification permitted). The mix dictates where NCiS focuses orchestration vs. where it proposes hollow-the-core strangler patterns.
The nightly choreography.
Fourteen EOD jobs. Nine message queues.
Every Hogan transaction memo-posts in daylight and hard-posts at night. NCiS maps the entire EOD choreography — the jobs that must run, their order, their dependencies, their MQ consumers. Miss one and the GL doesn't balance. Modernization starts here, because batch windows are the first constraint that collapses under 24×7 demand.
EOD Batch Sequence
click any job to trace upstreamMQ Queue Topology
real-time + async messagingBatch jobs run against the legacy core; MQ queues bridge the core to real-time channels. Any modernization pattern — from hollow-the-core to full replatform — has to keep this machinery running byte-for-byte. NCiS captures the full schedule + dependency tree so shadow runs can be verified in daylight, not in a 4-hour Saturday window.
Every Hogan transaction — executed twice.
Reconciled at the byte.
The Digital Twin receives a read-only production tap. For each selected transaction, NCiS simultaneously executes the legacy Hogan path (amber) and the modernized reference path (green). Outputs are compared at three levels: the COMMAREA byte image, the journal row, and the downstream GL posting. Variance at any level halts module promotion and opens an SME ticket with automatic root-cause bisection. Below: a live sample of four canonical flows.
Legacy Path · Hogan
Modernized Path · NCiS-Orchestrated Twin
Dual-path execution, not simulation.
This is not mock data running against mock systems. The legacy column represents CICS-mediated execution of the production Hogan program chain. The modern column represents the same logical transaction executed by a target-state implementation (Java microservice on a Postgres+Kafka substrate). Both paths run against their own journal store; the reconciler compares the outputs byte-for-byte.
Module promotion halts. Diff is signed. SME is paged.
If the reconciler finds a variance — even one-penny, even one-flag — the candidate module's promotion to production traffic is automatically withheld. A signed diff artifact is written to the Governance layer, the originating SME (or LLM agent) is paged with the bisected root cause, and the remediation is tracked with an audit ID. Regulators receive the event; nobody has to chase it down on Monday.
Every module. Every edge.
Every dependency. One graph.
The Lineage Engine constructs a canonical directed graph over every artifact ingested from Hogan. Below is the live 45-module module-call subgraph, centered on CIF — the master customer hub. Click any node to drill into its sub-functions, CICS programs, DB2 tables, and GenAI opportunities. In production, the graph also binds 16 COBOL programs, 25 DB2 tables, 14 batch jobs, and 9 MQ queues — the complete semantic surface.
Because Hogan's semantics live in four places at once.
A given business rule — say, "waive the NSF fee for the first two occurrences on a senior customer's account" — may be encoded partially in COBOL (FEENSF), partially in a DB2 parameter table (FEE.PARM_TBL), partially in a copybook default (FEEDEF), and partially in a CICS map (FEENSFM). Read any one and you see a fragment. The graph stitches them. NCiS doesn't just read code; it reconstructs intent.
The first deliverable every engineer wants.
Before NCiS, "what breaks if I change this copybook?" was a week of grep-through-JCL and cross-referencing old documentation. With the Lineage graph, the same question resolves in milliseconds: exact program dependencies, exact CICS impact, exact batch jobs affected, exact regulatory overlays disturbed. The engineer's week becomes an afternoon. The afternoon's change becomes defensible.
Two modernization paths.
Traditional vs GenAI — the economics, step by step.
Modernization is not one recipe. NCiS publishes two full playscripts authored from real Tier-1 engagements: a Cloud Native progression (6 steps, ~18–31 months traditional, 7–12 months with GenAI, ~63% savings) and an Agent Fleet progression (6 steps, ~24–38 months traditional, 5–9 months with GenAI, ~78% savings). Every step shows the real team composition, duration range, cost band, and specific tooling on both sides of the comparison.
The knowledge of forty years,
available in four seconds.
The SME co-pilot is grounded strictly in the Lineage graph and Digital Twin outputs. It answers only questions with citations back to specific source lines, journal rows, or twin receipts. Hallucination is structurally disabled: if the retrieval layer returns no provenance, the model refuses to answer.
Tap any prompt to replay a grounded SME-caliber response.
Every answer cites at least one of: source line, twin journal row, or graph edge. Ungrounded questions are declined, not guessed.
Hello. I am the NCiS Digital SME. I have indexed every COBOL program, copybook, CICS transaction, DB2 catalog entry, and SME transcript in your Hogan environment. Every answer I produce cites its source. Pick a prompt on the left, or ask me anything grounded in the graph.
Every modernization step produces
regulator-grade evidence — by design.
NCiS's Governance layer maps every module to the regulations that govern it, continuously measures coverage, and generates an evidence package on demand. The matrix below is representative; in production it spans every CFR part that touches the bank's core operations.
| Regulation / Module | CIS | DDA | SDA | TDA | IRA | LN | MTG | GL | FEE | INT | STMT | CHL |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| BSA / AML (31 CFR 1020) | ● | ● | ● | ● | ◐ | ● | ◐ | ● | ○ | ○ | ○ | ● |
| OFAC Screening | ● | ● | ● | ● | ◐ | ● | ◐ | ○ | ○ | ○ | ○ | ● |
| SOX 404 ICFR | ● | ● | ● | ● | ● | ● | ● | ● | ● | ● | ● | ◐ |
| FFIEC IT Handbook | ● | ● | ● | ● | ● | ● | ● | ● | ● | ● | ● | ● |
| FFIEC CAT (Cyber) | ● | ● | ● | ● | ● | ● | ● | ◐ | ○ | ○ | ○ | ● |
| Reg CC — Funds Availability | ○ | ● | ◐ | ○ | ○ | ○ | ○ | ○ | ○ | ○ | ◐ | ◐ |
| Reg D — Deposit Limits | ○ | ● | ● | ○ | ○ | ○ | ○ | ○ | ○ | ○ | ○ | ◐ |
| Reg E — Electronic Transfers | ○ | ● | ◐ | ○ | ○ | ○ | ○ | ○ | ◐ | ○ | ● | ● |
| CFPB 1033 — Data Portability | ● | ● | ● | ● | ● | ● | ● | ○ | ● | ● | ● | ● |
| RESPA (Mortgage Servicing) | ○ | ○ | ○ | ○ | ○ | ◐ | ● | ○ | ◐ | ◐ | ● | ○ |
| FATCA / IRS Reporting | ● | ◐ | ◐ | ◐ | ● | ◐ | ◐ | ○ | ○ | ● | ● | ○ |
| GLBA — Privacy / 501(b) | ● | ● | ● | ● | ● | ● | ● | ○ | ○ | ○ | ● | ● |
| CCAR / DFAST (stress) | ● | ● | ● | ● | ● | ● | ● | ● | ◐ | ◐ | ○ | ◐ |
| SR 11-7 (Model Risk) | ◐ | ◐ | ◐ | ◐ | ◐ | ● | ● | ● | ● | ● | ○ | ○ |
Legend: ● Full coverage ◐ Partial / in rollout ○ Not applicable. Coverage classifications are regulator-agreed and refreshed on every release train. Evidence packages include provenance receipts from Layers 02 / 03 / 05.
NCiS — Working Title for Patent Filing
System and Method for AI-Assisted, Audit-Preserving Modernization of Mainframe Core Banking Platforms — the working title for the substantive patent family covering NCiS's novel contributions. The following is an executive-grade summary of the invention's distinct claims. A formal specification with figures and full claim tree is under preparation with patent counsel.
What no prior art combines.
Mainframe modernization tools exist (AWS Blu Age, IBM Watsonx Code Assistant for Z, Micro Focus Enterprise Server, Asysco, Modern Systems). Digital twins exist (in manufacturing, energy, and aerospace). Lineage graphs exist (data-catalog domain). LLM code-translation exists. No prior system combines these into a closed-loop, audit-native substrate where (a) legacy semantics are captured as a first-class graph, (b) a production-tap digital twin reconciles byte-for-byte against the legacy core, (c) LLM agents operate only on graph-grounded retrieval, (d) every artifact is cryptographically chained, and (e) cutover is a contract-governed orchestration with instant provable rollback. That combination — and the interior mechanics that make it work — is the invention.
Representative Claims
Dual-Path Audit-Preserving Execution Substrate
A system comprising: (i) an ingestion subsystem that harvests COBOL source, job control, copybooks, database catalogs, and SME transcripts from a mainframe core banking platform; (ii) a lineage subsystem that resolves said harvest into a provenance-preserving directed graph of entities, flows, and regulatory obligations; (iii) a dual-path execution subsystem configured to consume a read-only tap of production transactions and execute each transaction simultaneously against (a) the legacy core and (b) a modernized reference implementation, emitting byte-level reconciliation receipts to an append-only ledger; (iv) a generative subsystem constrained to produce output only when grounded in said lineage graph or said reconciliation ledger; and (v) a traffic-orchestration subsystem that shifts production load between legacy and modernized paths under contract-governed rollback conditions anchored to the reconciliation receipts.
Graph-Grounded LLM Preservation of Tribal Knowledge
A method comprising: capturing a plurality of subject-matter-expert recordings; linking each utterance to one or more nodes in a lineage graph derived from a core banking mainframe codebase; indexing said utterances into a retrieval substrate; constraining a language-model inference pipeline such that no response is emitted unless at least one retrieved passage is cited with node-level provenance; and emitting, with each response, a cryptographically-signed citation manifest suitable for regulatory audit.
Reconciliation-Gated Module Promotion
A method for sequencing modernization of interdependent mainframe modules, comprising: computing a coupling, volatility, and regulatory-density score per module from a lineage graph; selecting a module for promotion based on said scores and an organization-capacity index; executing a minimum dual-run interval during which the reconciliation ledger is monitored for zero-variance journal parity; and promoting the module to primary traffic if and only if said interval completes without variance, otherwise withholding promotion and writing a signed halt-artifact to the governance ledger.
Cryptographically-Anchored Cutover Contracts
A method for executing a zero-downtime cutover between a legacy mainframe core banking module and a modernized replacement, comprising: generating a cutover contract that encodes a minimum dual-run duration, a maximum acceptable variance envelope, an evidence-export specification, and a rollback trigger set; signing said contract against a state root of the lineage graph and the reconciliation ledger; executing the cutover under enforcement of said contract; and, if any rollback trigger fires during or after the cutover, atomically reverting traffic to the legacy module and emitting a signed rollback receipt redeemable against the pre-cutover state root.
Regulator-Facing Evidence Aggregator
A system comprising: (i) a governance layer consuming receipts from all preceding layers; (ii) a regulatory-domain overlay mapping each receipt to zero or more regulatory obligations; (iii) a query interface operable by a regulator or auditor to produce a time-bounded evidence package including provenance receipts, reconciliation manifests, SME attestations, and orchestration decisions relevant to a named regulation and a named module, such package being independently verifiable against a published cryptographic root.
Prior-Art Differentiators
Translation-first tools produce modernized source; NCiS produces modernized source plus a continuous reconciliation substrate. The translation is an output, not the product. Prior art does not gate promotion on dual-path parity receipts.
Data catalogs (Collibra, Alation) map schema-to-schema. NCiS maps copybook-to-program-to-transaction-to-table-to-regulation across the entire mainframe surface, including SME voice. The graph dimension is fundamentally broader.
Industrial twins simulate physical processes. NCiS twins a semantic process — and enforces byte-level parity as the acceptance condition. Semantics are stricter than physics: a one-penny variance halts the program.
General code LLMs hallucinate. NCiS's generative layer is structurally incapable of answering outside its grounded graph, and every answer is citation-backed and audit-exportable. This is a regulatory-grade inference pipeline, not a productivity tool.
Confidential · Under preparation.
This disclosure is confidential to R.R. Enterprises AI and its authorized counsel. The formal filing targets a provisional application followed by a PCT-track non-provisional, with continuation strategy scoped to cover (i) the core dual-path method, (ii) the grounded-LLM constraint, (iii) the orchestration contract format, and (iv) the evidence-aggregator architecture. Inventor of record: Harsha Nagaraj, Founder, R.R. Enterprises AI.