LatestAll news

Home / Use cases / PCT-001

Monitor portfolio · PCT-001 · Published note

Loans marked below 90

An LP analyst covering seven large BDCs wants the list of holdings marked below 90% of cost from the June 30, 2026 filings.

Worked example: 930 holdings across 7 BDCs are marked below 90% of cost, from 199 at BCRED to 45 at OBDC. See the walk-through ↓

BDCsDirect lendingResearch & ratingsPublic dataVendor + in-house
INPUTSEC BDC data setAI STEPSPull the SEC data set · Rank every markbelow 90 · Link each row to its 10-QOUTPUTRanked list of marks below 90 withfiling linksHUMAN CHECKSpot-check at least 5 rows against thefiled HTML

Today vs automated

TODAY
  1. Open each 10-QAnalyst · manual
  2. Copy holdings to ExcelAnalyst · manual, re-keying
  3. Compute FV ÷ cost by handPM · manual
  4. Email / meeting hand-offAnalyst → PM
AUTOMATED
  1. A BDC files its 10-QTrigger: starts on its own
  2. Pull the SEC data setAutomated · Document store
  3. Rank every mark below 90Automated · LLM
  4. Link each row to its 10-QAI draft, human review · Monitoring model
  5. Spot-check at least 5 rows against the filed HTMLApprove / override · PM

AutomatedAI draft, human reviewHuman onlyTools: Inbox / portal, Document store, LLM, Monitoring model, Loan system

Estimated valueestimate

12–210 analyst hours a year
$2k–$52k of time, vs an assumed $20k–$100k tool cost

Net: −$98k to $32k a year net of tool cost; at the low end of the assumptions it does not cover the tool cost.

Manual4–8 h per quarterly reviewWith AI0.5–1 h per quarterly review, review included
Assumptions
  • 4–8 analyst hours per quarterly review today; 0.5–1 hours with an agent, including review
  • 4–28 quarterly reviews a year for one team
  • $150–250 per hour loaded cost
  • Tool or build cost $20,000–100,000 a year (both)

Estimate, based on the assumptions shown; your numbers will differ. How we estimate

What else it can do for the team

  • Cross-check your own marks on shared names
  • Track names that fall below 90 two quarters running
  • Feed the list into the watchlist
  • Draft the quarterly LP research note
  • Compare manager marks on the same loan
Worked walk-throughKey catch: 930 holdings across 7 BDCs are marked below 90% of cost, from 199 at BCRED to 45 at OBDC.

1. The input

The SEC BDC Data Set for the period ending 2026-06-30 (SEC BDC Data Sets), which carries each BDC’s tagged Schedule of Investments. Example filing: ARCC 10-Q on EDGAR.

2. What the AI produces

BDCHoldings below 90% of costFiling
BCRED19910-Q filed 2026-08-12
ARCC19810-Q filed 2026-07-29
GBDC17210-Q filed 2026-08-03
HTGC13110-Q filed 2026-07-30
MAIN11610-Q filed 2026-08-07
FSK6910-Q filed 2026-08-06
OBDC4510-Q filed 2026-08-05

Computed by the site build from the SEC data set and re-derived by an independent check. Row-level detail is on the BDC marks demo.

3. What a person still checks

  1. Spot-check at least five rows against the filed HTML schedule.
  2. Make sure subtotal and duplicate rows were not counted.
  3. Check that the cost used is amortized cost for that position, not a total for the issuer.
  4. Do not infer distress: a low mark is what the filing reports, nothing more.

4. Where it breaks

  • The SEC has flagged inconsistent BDC tagging; some filers tag the schedule differently.
  • Equity and warrant positions can sit below cost for reasons unrelated to credit.
  • Positions split across several rows can show different marks for the same loan.
How you get thereFour steps from prompting to an agentic workflow; most teams start at step 2.
1 · PromptingAn analyst asks a chat tool about one quarterly review.
Needs: an approved AI tool, a confidentiality policy.
Risk: unchecked answers, data pasted outside policy.
2 · WorkflowA fixed pipeline extracts the same fields into your template every time.
Needs: field definitions, a template, document parsing.
Risk: breaks silently when formats change.
Typical starting point
3 · AgentIt watches the inbox or data room, runs the checks and drafts the output.
Needs: inbox/doc-store access, system write-back, logging.
Risk: acting on a misread; needs review gates.
The automated lane above
4 · Agentic workflowSeveral agents work across the deal or book; people approve exceptions.
Needs: shared data model, permissions, audit trail, exception queue.
Risk: errors compound across steps; needs monitoring.
Solution optionsBuy (21 vendors listed), build or hybrid, with the trade-offs of each.
Build
An LLM API, a document parser, a vector or document store, your data warehouse, and a review queue.
Fits your definitions exactly; you own maintenance and controls.
Hybrid
Vendor extraction feeding your own checks, templates and systems.
Common middle path; two things to integrate and govern.

Listing is not endorsement.

What changes for the team4–8 h → 0.5–1 h per quarterly review (estimate); coverage goes from a sample of borrowers, quarterly to every borrower, every delivery.
ROLESAnalysts move from re-keying packs to reviewing exceptions and calling borrowers.
CYCLE TIME4–8 h → 0.5–1 h per quarterly review (estimate)
COVERAGEFrom a sample of borrowers, quarterly to every borrower, every delivery.
CONTROLSEvery output cites its source; approvals and overrides are logged.
BCRED199ARCC198GBDC172HTGC131MAIN116FSK69OBDC45
Holdings marked below 90% of cost, by BDC Source: SEC BDC Data Sets, period ending 2026-06-30. Derived from SEC filings; may contain extraction errors. Not investment advice. See the rows →

BDC marks demo →

How it works

Full title: Every loan a BDC marks below 90 cents, ranked and linked to the filing

“198 holdings in one BDC's book are marked below 90% of cost. Every one links to the 10-Q.”

Scenario

Public: an LP analyst covering seven large BDCs wants the list of holdings marked below 90% of cost from the June 30, 2026 filings.

What the human still decides

Spot-check at least 5 rows against the filed HTML. Check that subtotal or duplicate rows were not counted. Do not infer distress: a low mark is what the filing reports, nothing more.

Why it matters

The SEC staff statement (28 Sep 2026) asks for marks calibrated to market evidence and entity-specific disclosure. Markdowns are concentrated in software. A ranked, linked list of low marks is the starting point for both.

Before and after

Before

  1. Open the 10-Q and scroll the Schedule of Investments, often hundreds of rows
  2. Copy the table into a spreadsheet and clean merged cells and footnotes
  3. Compute fair value ÷ cost per holding by hand
  4. Sort and write up the outliers

After (agent first draft)

  1. Agent downloads the SEC BDC Data Set for the filing month
  2. Filters the SOI table to one BDC's latest filing
  3. Computes fair value ÷ cost for each holding and lists those under 0.90, ranked
  4. Writes a 5-bullet factual summary with each row linked to the filing's inline XBRL

Agent pattern: download → filter → compute ratio → rank → cite filing → human spot-check

Tools

  • Python or any agent with a shell
  • SEC BDC Data Sets
Try it: prompt and data
Using the SOI table from the SEC BDC data set for CIK {cik}, period {date}, compute fair value/cost for every holding, list those under 0.90 sorted ascending, show the inline XBRL URL per row, and summarize in 5 bullets. Do not speculate on causes.

Data

Public data only. The SEC states it cannot guarantee the accuracy of the data sets and that they are not a substitute for the filings. The SEC has also flagged inconsistent BDC SOI tagging.

Who uses it, vendors and build
Segments
BDCs, Direct lending, Research & ratings
Who uses it
Portfolio manager, Credit analyst, LP / credit research
Build
Vendor + in-house
In-house
This site's open SOI pipeline (Python). Estimate: days.
Last verified
2026-10-15

Vendors that claim this capability

Tags reflect each vendor's own public claims at the linked page (checked 2026-10-15). Not tested, ranked or endorsed.

Sources and related
In the news.

Next step

Get new walk-throughs

Subscribe to the newsletter

Not investment advice. Names marked SAMPLE are fictional. Found an error? See the corrections log.