LatestAll news

Home / Use cases / PCT-002

Underwrite · PCT-002 · Published note

Covenants out of the agreement

An associate on a new deal needs the covenant package, EBITDA add-backs, and key baskets from a publicly filed credit agreement before the term-sheet call.

Worked example: The Viavi term loan has no financial maintenance covenant at all (Article VI runs §6.01 to §6.09), and restricted payments are unlimited at pro forma Total Leverage of 3.50x or less. See the walk-through ↓

Direct lendingSpecial sits / distressedPublic + sample docsVendor + in-house
INPUTCredit agreement (200+ pages)AI STEPSRead the agreement · Table everycovenant · Cite each sectionOUTPUTCovenant and basket table with sectioncitesHUMAN CHECKVerify every cite against the agreement

Today vs automated

TODAY
  1. Read the whole agreementAssociate · manual
  2. Retype test levelsAssociate · manual, re-keying
  3. Hunt for definitionsVP / counsel · manual
  4. Email / meeting hand-offAssociate → VP / counsel
AUTOMATED
  1. A draft or final credit agreement lands in the data roomTrigger: starts on its own
  2. Read the agreementAutomated · Document parser
  3. Table every covenantAutomated · LLM
  4. Cite each sectionAI draft, human review · Model template
  5. Verify every cite against the agreementApprove / override · VP / counsel

AutomatedAI draft, human reviewHuman onlyTools: Data room, Document parser, LLM, Model template

Estimated valueestimate

160–1,320 analyst hours a year
$24k–$330k of time, vs an assumed $20k–$100k tool cost

Net: −$76k to $310k a year net of tool cost; at the low end of the assumptions it does not cover the tool cost.

Manual6–12 h per agreementWith AI1–2 h per agreement, review included
Assumptions
  • 6–12 analyst hours per agreement today; 1–2 hours with an agent, including review
  • 40–120 agreements 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

  • Build the reporting and test-date calendar from the same read
  • Load test levels and definitions into the monitoring template
  • Compare the draft against the term sheet and list every mismatch
  • Flag LME blockers and their gaps in the same pass
  • Produce a one-page term summary for the IC pack
Worked walk-throughKey catch: The Viavi term loan has no financial maintenance covenant at all (Article VI runs §6.01 to §6.09), and restricted payments are unlimited at pro forma Total Leverage of 3.50x or less.

1. The input

A public credit agreement: the Term Loan Credit Agreement dated as of October 16, 2025, filed by Viavi Solutions Inc. as EX-10.1 to its 10-Q on 2025-10-30, with Wells Fargo Bank, National Association as agent. Source: SEC EDGAR exhibit. A public-company agreement is used because private-deal agreements are not public; the structure is the same.

2. What the AI produces

ItemFindingSection
Financial maintenance covenantNone found. Article VI runs §6.01 to §6.09 with no financial covenant section (covenant-lite term loan).Art. VI (§6.01–6.09)
General debt basketGreater of $122,500,000 and 50% of LTM Consolidated EBITDA§6.01(a)(xiv)
General RP basketGreater of $61,500,000 and 25% of LTM Consolidated EBITDA, plus Available Amount and Available Equity Amount§6.08(a)(viii)
Ratio RP basketUnlimited if pro forma Total Leverage Ratio ≤ 3.50x and no Event of Default§6.08(a)(xiv)
Run-rate cap on (b)(ii)(b)(ii) savings from later Specified Transactions / restructurings: capped at 30% of Consolidated EBITDA per Test Period, calculated after the add-backs; Reg S-X-compliant adjustments sit outside the cap§1.01 “Consolidated EBITDA” cl. (b), 2nd proviso
J.Crew blocker: designationPresent. A Restricted Subsidiary holding Material IP cannot be designated unrestricted§5.15 (last sentence)
Chewy blocker (guarantor release)Present. Ceasing to be wholly owned does not release a guarantor unless the equity went to a non-Affiliate for a bona fide business purpose§9.15 (proviso)

Every row was read against the filed text (regex + section tracking on the filed HTML; every row hand-read against the filed text on 2026-10-16). Full table: ca-scan-2025a.json.

3. What a person still checks

  1. Open each cited section and confirm the number and the “greater of” grower both match.
  2. Read the defined terms the basket depends on (here, LTM Consolidated EBITDA) before you trust any capacity figure.
  3. Check for baskets that can be reclassified or stacked, and whether amounts reload.
  4. Confirm no later amendment changed the clause (search the borrower’s later 8-Ks and 10-Qs).

4. Where it breaks

  • A “None found” result is only as good as the search: a covenant can sit in a schedule or a separate document.
  • Long definitions with nested provisos get summarized too neatly; the second proviso is where caps hide.
  • Scanned or image-only exhibits defeat text extraction.
How you get thereFour steps from prompting to an agentic workflow; most teams start at step 1.
1 · PromptingAn analyst asks a chat tool about one agreement.
Needs: an approved AI tool, a confidentiality policy.
Risk: unchecked answers, data pasted outside policy.
Typical starting point
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.
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 (15 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 team6–12 h → 1–2 h per agreement (estimate); coverage goes from the sections someone had time to read to every clause and schedule checked against the term sheet.
ROLESAssociates move from reading and retyping to checking cited exceptions.
CYCLE TIME6–12 h → 1–2 h per agreement (estimate)
COVERAGEFrom the sections someone had time to read to every clause and schedule checked against the term sheet.
CONTROLSEvery output cites its source; approvals and overrides are logged.

Credit agreement scan demo →

How it works

Full title: Covenants, baskets and add-backs out of a 200-page credit agreement, every line cited to a section

“200 pages in, one table out, every line cited to a section.”

Scenario

Public: an associate on a new deal needs the covenant package, EBITDA add-backs, and key baskets from a publicly filed credit agreement before the term-sheet call.

What the human still decides

Verify every cite against the agreement. Check definitions and step-downs yourself; models miss cross-references and amendments.

Why it matters

Add-back caps, cov-lite springing tests, and LME blockers are the 2026 documentation fight. Many vendors sell this.

Before and after

Before

  1. Read a 100–200+ page agreement (illustrative length)
  2. Find each financial maintenance covenant and its definitions
  3. Key covenants, levels and cure rights into a term-sheet grid

After (agent first draft)

  1. Agent reads the agreement (HTML or PDF)
  2. Outputs a table: covenant, test level, test frequency, definitions referenced, equity cure, section cite
  3. Quotes key definitions verbatim and says 'not found' when unsure

Agent pattern: extract → table → cite section → human check

Tools

  • LLM with PDF/HTML input
  • EDGAR full-text search for credit agreement exhibits (EX-10)
Try it: prompt and data
From this credit agreement, list every financial maintenance covenant with level, test date, and section number. Quote the definition of Consolidated EBITDA add-backs verbatim. If unsure, say 'not found'.

Data

Use only agreements that are publicly filed (many are exhibits to 8-K, 10-Q or 10-K filings). Never paste a confidential agreement into a consumer AI tool.

Who uses it, vendors and build
Segments
Direct lending, Special sits / distressed
Who uses it
Associate / VP (deal team), Special sits analyst
Build
Vendor + in-house
In-house
LLM + document parser + an eval set of 20 EDGAR agreements. Estimate: 2–4 weeks for a small team.
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.

Part of the deal lifecycle: stage 4 of 7, Documents.

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.

Worked example: covenants in a filed agreement

What one filed credit agreement says: Viavi Solutions Inc., Term Loan Credit Agreement dated as of October 16, 2025
ItemFindingSection
Financial maintenance covenantNone found. Article VI runs §6.01 to §6.09 with no financial covenant section (covenant-lite term loan).Art. VI (§6.01–6.09)
Consolidated EBITDA definitionDefined term; add-backs in clauses (a)(i)–(xii), (b) run-rate, (c), (d) QoE/model add-backs§1.01 “Consolidated EBITDA”
Run-rate cost savings add-backTwo prongs. (b)(i) savings tied to the Transactions: no cap found. 24-month window; must be “reasonably quantifiable and factually supportable”§1.01 “Consolidated EBITDA” cl. (b)
Run-rate cap on (b)(ii)(b)(ii) savings from later Specified Transactions / restructurings: capped at 30% of Consolidated EBITDA per Test Period, calculated after the add-backs; Reg S-X-compliant adjustments sit outside the cap§1.01 “Consolidated EBITDA” cl. (b), 2nd proviso
Model / QoE add-backsAdd-backs reflected in the lenders' Feb 2025 model and QoE summary, or a later QoE report, are permitted with no stated cap§1.01 “Consolidated EBITDA” cl. (d)
General debt basketGreater of $122,500,000 and 50% of LTM Consolidated EBITDA§6.01(a)(xiv)
General investment basketGreater of $122,500,000 and 50% of LTM Consolidated EBITDA, plus Available Amount / Available Equity Amount / Available RP Capacity§6.04(n)
Investments in Unrestricted SubsidiariesGreater of $49,000,000 and 20% of LTM Consolidated EBITDA§6.04(z)
General RP basketGreater of $61,500,000 and 25% of LTM Consolidated EBITDA, plus Available Amount and Available Equity Amount§6.08(a)(viii)
Ratio RP basketUnlimited if pro forma Total Leverage Ratio ≤ 3.50x and no Event of Default§6.08(a)(xiv)
Junior debt prepayment basketGreater of $61,500,000 and 25% of Consolidated EBITDA, plus Available Amount§6.08(b)(iv)
J.Crew blocker: designationPresent. A Restricted Subsidiary holding Material IP cannot be designated unrestricted§5.15 (last sentence)
J.Crew blocker: transferPresent. No investment or transfer of Material IP to an Unrestricted Subsidiary (non-exclusive licences excepted)§6.04, final paragraph (ah)
“Material Intellectual Property” definitionIP “material to the business and operations of the Borrower and the Subsidiaries, taken as a whole”. Materiality is not quantified§1.01 “Material Intellectual Property”
Serta blocker (subordination)Present, with an exception: subordination needs each adversely affected Lender's consent unless every such Lender is offered its pro rata share of the Priority Debt (open ≥5 Business Days)§9.02(b)(x)
Pro rata sharing / waterfall sacred rightPresent: changes need each directly and adversely affected Lender§9.02(b)(ix)
Open-market / Dutch auction buybacksPermitted: non-pro rata open-market purchases and Dutch auctions by a Purchasing Borrower Party (the route used in some uptiers; read with the Serta clause above)§9.04(h)
Chewy blocker (guarantor release)Present. Ceasing to be wholly owned does not release a guarantor unless the equity went to a non-Affiliate for a bona fide business purpose§9.15 (proviso)

Source: EX-10.1 to Viavi Solutions Inc. filing 0000912093-25-000153, filed 2025-10-30 ↗. Every row hand-checked against the agreement text. One agreement, shown as a worked example, not a market survey.

FAQ

Can AI extract covenants from a credit agreement?
It can locate definitions, baskets and blockers and quote the section, but each item still needs a person to check it against the text. The table on this page was produced that way and every row was hand-checked.
What is a covenant-lite term loan?
A term loan with no financial maintenance covenant. The filed agreement in the table has no financial covenant section in Article VI.
What are J.Crew, Serta and Chewy blockers?
Clauses that restrict moving material IP to unrestricted subsidiaries (J.Crew), subordinating existing lenders without their consent (Serta), and releasing guarantors when a subsidiary stops being wholly owned (Chewy).
Where do grower baskets show up?
In the debt, investment and restricted-payment covenants, usually as the greater of a dollar amount and a percentage of LTM EBITDA, as in the table on this page.