Business Observability · Loan Origination

Illustrative sample data — not a live customer book

Open Pipeline Exposure Briefing

The loan value you have in flight right now — and where the clock is running out — assembled live by an AI asking Outcome Owl in plain language, every figure cited and its limits disclosed.

Live · read-only connection data as of 2026-07-23 data confidence: high 27 typed tools

One plain-language question — “what's my open loan exposure, and what should I worry about?” — resolved in seconds: $3.43B of loan value in the origination pipeline across 5,792 active applications. And the clock is not kind — $927.7M of it is already past its expected close.

The headline

Open loan value in origination

$3.43B

exact: $3,427,653,216.14 · sum of Loan Amount · function=sum · status=Active

$927.7M already overdue27% of the pipeline

The clock is the story

$927.7M across 1,576 loans is already past its expected close. Each loan is judged against its own deadline — business start plus its configured ~60-day duration — so “overdue” is a real per-loan clock, not a blanket age cutoff.

5,792

Active applications in origination

99.7% carry a loan amount — near-complete, no silent undercount

$2.29B

Jumbo exposure (> $766,550)

2,311 loans · 67% of the pipeline

~$594K

Mean loan in flight

$3.43B ÷ 5,773 · range $50K – $2.0M

Concentration & risk

Where the exposure concentrates

Two-thirds of the pipeline sits in jumbo loans above the conforming limit — larger files, tighter secondary-market and non-QM scrutiny.

$2.29Bjumbo · 2,311 loans > $766,550

Ask it any threshold: $766,550 is the conforming limit, passed as a parameter — the same query answers for $1M, or a state high-cost line, without a new report. Each loan is counted once at its recorded amount.

What's slipping — and breaching

1,576loans overdue · $927.7M at stake

1,157 active rule violations sit on origination files — stalls, expired locks, thresholds — each drillable to the loan it's on.

A 60-day clock, not a 2-year one: unlike a long-tail claim, a loan pipeline turns in weeks — so “overdue” is the live signal here, and the surface times it off each file's own deadline.

Your single largest exposure — and it's stuck

Loan #ML-11198 · consumer mortgage

$1,999,964.27

“Consumer mortgage loan origination.”

Overdue 245 days Stalled 261 days 1 threshold breach severity 4

Exposure and risk on the same file. Your largest loan in flight — a $2.0M mortgage — reached Conditional Approval on Nov 4, 2025, then went silent. 261 days later it's still there: no clear-to-close, a rate lock long expired, an open threshold breach, 245 days past its expected close. The money and the problem are one file — and it took one question to surface it.

Application Submitted

Sep 21

Loan Estimate Issued

Sep 24

Rate Lock

Sep 25

Closing Disclosure

Sep 28

Document Collection

Oct 4

Credit Check

Oct 5

Identity Verification

Oct 6

Income & Asset Verify

Oct 13

Appraisal

Oct 27

Underwriting Review

Nov 1

Conditional Approval

Nov 4 — stalled

completeness: indeterminate intake: CSV upload impacts: customer experience, operations

Eleven steps completed, then nothing since Nov 4 — the timeline reads indeterminate (still open, not a false “closed”), and the stall is the last step whose successor never arrived. Honesty holds all the way down to the single loan.

How it was assembled — the live tool trace

ask “which of my attributes are open exposure?”

→ attribute_catalog(role='exposure') // Loan Amount = the pipeline value

ask “total the open origination pipeline”

→ aggregate(Loan Amount, sum, status=Active) // $3.43B · 5,773 of 5,792

→ aggregate(…, overdue_only=true) // $927.7M overdue · 1,576 loans

→ aggregate(…, Loan Amount ≥ 766,550) // $2.29B jumbo · 2,311 loans

→ list_violations(count_only) // 1,157 active

→ list_processes(sort by Loan Amount) // top exposure → Loan ML-11198

→ process + timeline(ML-11198) // 11 steps, stalled at Conditional Approval

Why an agent can act on this

cited

Every figure traces to a call

No number appears without the tool, filter, and population that produced it — a loan opens straight to its timeline.

bounded

Cost-governed at scale

An over-broad query is refused with the exact filter to add — it never silently runs to a timeout or returns a partial total dressed as complete.

coverage

Complete vs. partial, stated

Here, 5,773 of 5,792 files price out — the surface says “near-complete” rather than implying it. Elsewhere it will say 46%. It reports the number, whatever it is.

clocks

The right clock per file

“Overdue” is each loan's own deadline — business start plus its configured duration — not a blanket age cutoff. Selection is near-real-time; amounts are as of the daily refresh.

method

Its own limits, disclosed

Each loan counted once at its recorded amount; any threshold you name is a parameter, not a hard-coded report. No false precision.

facts

Facts, never judgments

It reports the exposure, the overdue count, the stalled loan; it never calls a file “bad” or tells anyone to act. That judgment — and the action — stay with you.

See this on your own pipeline

Every figure above came from a read-only connection and a plain-language question — no integration project, no data leaving your control. The same works on your originations, your reserves, your book.

Book a walkthrough →