FIELD ALERT  ·  May 15, 2026  ·  OpenAI personal finance launch  ·  12,000+ financial institutions in scope  ·  Agentic traffic timeline accelerated
Field Note v5 Companion to Silver Lake Series #2  ·  CF Companion ↗

OpenAI Just Pointed Agents at Your System.
Is Your IBM i Ready to Answer?

ChatGPT's personal finance launch is not a consumer product story. For the IBM i operator, it is an infrastructure alert. The agentic traffic is coming. The question is whether your system is in a condition to receive it.

Signal4i · Field Note v5 · May 2026 · Signal Stack v9.2 · 348 signals · 17 categories · Cat 16 · Cat 11
12K+
Financial institutions
in ChatGPT's scope
via Plaid · Day one
200M
Monthly users asking
financial questions
in ChatGPT today
0
IBM i shops with
a defined agentic
traffic policy
Q+4
OpenAI's signaled
timeline to agent
execution in finance

The Reframe

This Is Not an OpenAI Story.

On May 15, 2026, OpenAI launched a personal finance experience inside ChatGPT. Pro users in the U.S. can now connect their bank accounts, credit cards, investment portfolios, and loan accounts to the chatbot via Plaid. More than 12,000 financial institutions are in scope — including the lending institutions, servicers, and specialty finance companies that run their core operations on IBM i.

Most IBM i practitioners will read that news as something happening to the consumer side of their organization. A product story. A fintech story. A ChatGPT story.

The IBM i Read

OpenAI inserted a conversational agent between consumers and their financial institutions. That agent now holds a unified view of a consumer's financial life across every connected account.

Right now it responds with information. Within quarters, not years, it will respond with action — credit applications, payment arrangements, account servicing — all executed inside ChatGPT.

When that happens, the agent doesn't call a human at your institution. It calls your system. Your IBM i system.

The question for every IBM i practitioner is not whether this traffic is coming. It is coming. The question is whether your system is in a condition to receive it, respond to it, and make that response defensible.

Perplexity added Plaid integration one day before OpenAI's announcement. The race for the consumer finance relationship layer has already begun. Every participant in that race will eventually need to query the systems underneath. On IBM i, that means your DB2 tables. Your RPG business logic. Your core decisioning rules.

Agentic traffic is not a human browsing your portal. It is an AI system operating at machine speed, with full cross-institutional context, arriving at your door to transact.

The Wall

Four Reasons Your IBM i Isn't Ready Today.

Field Note v4 introduced the KD Wall — the point at which an organization's infrastructure physically prevents it from responding at agentic speed. For IBM i operators, the wall has a specific shape. Not a technology failure. An architecture and governance gap.

01
The data is in DB2 — but the schema isn't documented.
An agent can't query what it can't understand. The table structures, field definitions, and relationship maps that live in institutional memory need to be surfaced before an agent can navigate them. Documentation isn't overhead. It's the entry condition for agent access.
02
The business logic is in RPG — but it isn't callable.
Decades of decisioning rules, eligibility criteria, and workflow logic are encoded in production programs. An agent can't run them directly. They need to be exposed through an interface — MCP, a REST wrapper, a service layer — that the agent can invoke without human intermediation.
03
The integration surface doesn't exist yet.
Mapepire confirmed the path: IBM i can be reached from a modern stack through a clean API bridge. The capability is available. The implementation isn't done. The shops that build the bridge before the traffic arrives will be reachable. Those that build it after will be playing catch-up in a live environment.
04
The governance framework hasn't been defined.
What queries will you allow an agent to run against your system? What data will you expose? What actions will you permit without human approval? For regulated shops, these aren't product decisions. They are compliance and model risk governance events. They need answers before the traffic arrives — not after the first examination finding.

The Bridge

Three Tools That Make IBM i Agent-Accessible.

Silver Lake Installment 2 calls it the layer that doesn't commoditize. This is that layer, made concrete. The organizations building this infrastructure now are the ones agents will reach first — not because they're the largest, but because they're reachable.

Data Access

Mapepire

Opens IBM i DB2 to modern API consumers over a clean REST/WebSocket interface. An agent that reaches Mapepire reaches your data in real time — not a batch file from yesterday. The state of the account, right now, in milliseconds.

Logic Access

MCP

Model Context Protocol provides the standard that lets an AI agent discover, call, and interpret tools from an IBM i system as part of a reasoning chain. Your RPG programs — decades of decisioning logic — become callable tools in the agent's workflow.

Knowledge Access

Project Bob + Claude Code

Surfaces business logic embedded in your production codebase and makes it interpretable. The institutional knowledge encoded in RPG isn't locked or untouchable. It is documentation-ready and agent-accessible with the right tooling applied systematically.

Together, these three form the bridge. Mapepire surfaces the data. MCP exposes the logic. Project Bob makes the knowledge readable. The IBM i becomes what the Silver Lake Series has been arguing it already is: the orchestration layer the agentic stack runs on top of.


Practitioner Diagnostic

Three Questions to Answer Before June.

The ChatGPT personal finance launch is the most visible marker yet that agentic traffic in consumer finance is no longer theoretical. These three questions are worth answering before Installment 3 arrives — because I3 maps the technical path assuming your shop has at least started asking them.

01
Can an agent reach your DB2 data in real time?
Not through a batch job. Not through a manual process. Through a live API surface that returns structured data in a form a machine can consume and act on. If the answer is no — or not yet — Mapepire is the starting point. The data is already there. The surface isn't.
If yes → define what the agent can see. If no → Mapepire is the first gate.
02
Can an agent call your business logic?
The decisioning rules, eligibility criteria, and workflow logic encoded in your RPG programs — are they accessible through an interface that doesn't require a human to run them? If they're locked in production programs with no callable surface, the agent has no way to make a defensible decision using your rules. MCP is the protocol. Project Bob is the inventory.
If yes → map which programs are exposed. If no → Project Bob first, MCP second.
03
Is your governance framework defined?
What will you allow agents to do in your system? What requires human approval? What requires audit trail documentation? For regulated lenders, these are compliance questions with examination consequences. For every IBM i shop, they determine whether your agentic deployment is controlled or chaotic. The crawl/walk/run discipline isn't caution. It's the framework.
If yes → document it and version it. If no → this is the governance gate I3 will map.

Silver Lake Series · Installment 3 · June 2026

The KD Reduction Engine — The Technical Path

Field Note v5 is the alert. Installment 3 is the map. The KD Reduction Engine arrives in June — tracing the full technical path from orchestration layer to knowledge distance reduction at enterprise scale. The IBM Cyber Fraud case from IBM Think 2026 provides the validation: 90% reduction in investigation time, driven by the same architecture the Silver Lake Series has been documenting. The organizations that reduce the distance before the agentic traffic arrives will handle it. The ones that don't will see the traffic route around them.

Follow the Silver Lake Series → signal4i.ai
Permanent Reference · signal4i.ai

What Does Agent-Ready Actually Look Like?

Field Note v5 names the wall. The reference page maps what's behind it — five layers every IBM i shop and consumer finance operator needs to build through before agentic traffic can flow without breaking. Discovery, Comprehension, Access, Action, Governance. llms.txt, MCP, Schema.org, SR 11-7. The practitioner checklist before I3 arrives.

Open the Five-Layer Reference → agent-ready.html
Consumer Finance Companion · consumerfinance.ai

When Klarna Is Your Peer — The Orchestration Wall in Regulated Lending

Klarna's AI retreat from credit-adjacent conversations wasn't an operational setback. In regulated U.S. consumer lending, it is a compliance and fair lending event. The companion piece applies the orchestration layer argument to the sector where the wall carries regulatory consequences — ECOA, FCRA, UDAAP, model risk governance — and maps what agent-ready actually looks like for a regulated lender.

Read the Companion → consumerfinance.ai