Overview
Delos AI Inc. has build an AI engine that offers 100% accuracy on calculating complex legal clauses that blend math and logic problems. Read more about it here. One example of such clauses are debt covenants in loan agreements. A common occurrence for PE-funded companies, those debt covenants require recurrent calculations (every month or quarter). Their complexity means that many debt funded companies and their lawyers struggle to calculate them correctly. The process is time consuming and error prone.
However, their nature - mixing math and logic - makes them perfect for Delos AI’s engine.
Link to the demo: Debt Covenant Demo
Use Case Details
For this demo we're interested in calculating the "Levarage Ratio". A common debt covenant used in loan agreements. It's purpose is to ensure the company is not taking on too much debt in comparison to its EBIDTA.
To illustrate the use case details we created a mock agreement with 3 key provisions:
- a rule defining the Leverage Ratio - limiting it at 5.00:1.00 of debt to EBIDTA
- a rule defining the EBIDTA for the given borrower
- a rule defining the Total Debt for the given borrower
The agreement itself is a mock one - containing just those 3 provisions. You can find it here. However the provisions themselve are based on real agreements and definitions found in loan agreements.
Agreement Provisions
Leverage Ratio
The Leverage Ratio provision in question is as follows:
For each fiscal quarter beginning with the quarter ending on the earlier of (i) March 31, 2020 or (ii) the last day of the fiscal quarter during which the Initial Period ends, the Borrower will not permit its Leverage Ratio of Total Debt to EBIDTA to exceed 5.00:1.00 at the end of any fiscal quarter, provided, however, that during any Acquisition Period, the Leverage Ratio shall not exceed 5.50:1.00 at the end of any fiscal quarter.
Simply put the total debt of the borrower cannot be bigger than 5 times its EBIDTA at the end of any fiscal quarter.
While this might seem straightforward, the devil is in the details. Specifically in how EBIDTA and total debt are defined for a particular company in a specific industry.
Though "EBIDTA" and "total debt" have a general meaning - the specifics will change from one agreement to another.
EBIDTA Definition
The EBIDTA definition is as follows:
“EBITDA” means, for any period, the sum of:
(i) Net Income for such period, plus
(ii) without duplication, the following consolidated expenses or charges to the extent deducted from Net Income in such period:
(a) interest expense,
(b) Taxes,
(c) depreciation,
(d) amortization,
(e) non-cash charges (but excluding, without duplication and to the extent included in Net Income, any cash actually paid with respect to such non-cash charges when paid),
(f) extraordinary, exceptional, unusual or non-recurring charges, expenses or losses,
(g) fees and expenses incurred in connection with the negotiation, execution and delivery on the Effective Date of the Loan Documents and the consummation of the Transactions,
(h) reasonable and documented fees and out-of-pocket expenses incurred in connection with any amendments, waivers or other modifications to this Agreement and the other Loan Documents,
(i) the annual administration fees (including without limitation any collateral management fee) paid to Administrative Agent pursuant to the Loan Documents, and
(j) fees, expenses and charges related to any Permitted Acquisition; provided, in no event shall the aggregate amount added to EBITDA for such period pursuant to this clause (j) exceed more than 5% of EBITDA for such period (calculated without regard to amounts added in the calculation of EBITDA pursuant to this clause (j)), minus
(iii) without duplication, the following consolidated gains or credits to the extent included in the calculation of Net Income for such period:
(a) income tax credits;
(b) all non-cash income or gain (but excluding any such amounts (A) in respect of which cash was received in a prior period or will be received in a future period or (B) which represent the reversal of any accrual of, or cash reserve for, anticipated cash charges that reduced EBITDA in any prior period);
(c) the proceeds of any life insurance policy; and
(d) any extraordinary, exceptional, unusual or non-recurring gains,
all calculated for the Borrower and its Restricted Subsidiaries on a consolidated basis in accordance with GAAP.
Total Debt Definition
The Total Debt definition is as follows:
“Total Debt” means, without duplication and on a consolidated basis for the Borrower and its Subsidiaries, the aggregate outstanding amount of:
(i) Borrowed Money, in particular:
(a) all loans
(b) notes
(c) debentures
(d) and other obligations for borrowed money (including intercompany and seller-financing).
(ii) Capital/Finance Leases – the present value of payments due under any lease classified as a capital or finance lease under GAAP.
(iii) Letters of Credit & Guarantees – (i) the undrawn face amount of all standby or commercial L/Cs, bankers’ acceptances, and surety or performance bonds, and (ii) all unreimbursed draws thereon.
(iv) Deferred Purchase Price – earn-outs, holdbacks, and other acquisition-related deferred consideration shown as a liability.
all calculated for the Borrower and its Restricted Subsidiaries on a consolidated basis in accordance with GAAP.
Delos Platform
The platform allows to easily transform complex debt covenant agreements into easy-to-use AI agents that can verify compliance:
Step 1 - Uploading agreements
The Delos platform allows to upload agreement and scan them to extract key terms. Users can add specific instructions for the AI as to what they'd like to extract.
Step 2 - Creating the Engine
The relevant clauses (those specific to debt covenant are converted with our algorithm into a logical format). Multiple clauses covering a given debt covenant are grouped into “Procedures” e.g. all clauses related to calculating EBIDTA and Leverage Ratios would be grouped into the “Leverage Ratio Procedure”.
Multiple procedures can be created for 1 agreement and then grouped into Agents.
The creation of this format requires a human:
- either a a Delos employee, or
- a trained user
However, once created – the agent can be re-used multiple times to re-run the same calculation for the given covenants, but with different data inputs.
1 Agreement -> 1 Agent -> Multiple Procedures Covered:
Delos Semantic Parsing Algorithm automatically converts agreement clauses into a logical format:
Step 3 - Adding Data Inputs
For this demo we need several pieces of information to calculate the EBITDA, the total debt and the Leverage Ratio: For the purpose of calculating this particular EBITDA, we need to use the following variables for the particular quarter:
| Item | Value |
|---|---|
| Net Income | $7,500,000 |
| Interest Expense | $1,000,000 |
| Taxes | $1,800,000 |
| Depreciation | $600,000 |
| Amortization | $400,000 |
| Non-cash Charges | $250,000 |
| Extraordinary Charges | $150,000 |
| Transaction Fees | $200,000 |
| Modification Fees | $50,000 |
| Administration Fees | $120,000 |
| Permitted Acquisition Fees | $300,000 |
| Income Tax Credits | $-100,000 |
| Non-cash Income or Gain | $-80,000 |
| Life Insurance Proceeds | $50,000 |
| Extraordinary Gains | $200,000 |
The total EBITDA for the quarter is $12,300,000
Total debt in turn is composed of:
| Item | Value |
|---|---|
| Loans | $50,000,000 |
| Notes | $10,000,000 |
| Debentures | $5,000,000 |
| Other Obligations for Borrowed Money | $2,500,000 |
| Capital Finance Leases Present Value | $3,000,000 |
| Undrawn Face Amount of Letters of Credit and Guarantees | $4,000,000 |
| Unreimbursed Drawn Amounts on Letters of Credit and Guarantees | $1,000,000 |
| Deferred Purchase Price | $750,000 |
The Total Debt with this formula will the be $76,250,000.
In accordance with the agreement, the Leverage Ratio for those numbers should be 6.2:1.00 (Total Debt / EBITDA = 76,250,000 / 12,300,000). Hence it would be in violation of the loan agreement.
Data can come from:
- user inputs in the chat
- "events" added by users or automatically via API
- a database of numerical data (for example specific data points on revenue at a given date)
In our example we defined a mock event with the quarterly data for Acme NYC in which we pass a table with the different inputs for EBITDA and Total Debt.
In a production setting, a company could easily connect their ERP or accounting system via API and transfer every quarter financial data which would then be used to calculate the specific ratios. The inputs just as easily be snippets from an excel spreadsheet or a table.
Step 4 - Using the Agent
Once in place – the Agent can be used to take questions in natural language and calculate the relevant ratios based on particular data points.
The Agents can then be configured to run their compliance automatically (e.g. every week, month, quarter) or whenever the user wants to use them.
Along with the answer (compliant or not compliant) the agent provides a "Compliance Track" explaining it's reasoning of the calculation.
Scaling to other clauses
The magic of Delos comes from the fact that it can be easily scaled to other clauses or used when clauses change.
If our definition of EBITDA changes, we can simply update the formula we're using to calculate it. If we want to add a new variable or element of the EBITDA definition, we can do so as well.
Conclusion
In this demo we've shown successfully how our Delos One engine can be used to accurately calculate even the most complex financial clauses found in loan agreements. The demo used in question shows a rudimentary example. However, our system can be easily scaled to calculate much more complex clauses.
Interested in Early Access?
We’re offering this early to a few select partners. Want to test it out? Contact:
- Witold Kowalczyk
- witold@delosintelligence.com
- www.delosintelligence.com
About Delos
Delos is building the world’s first AI clearinghouse for legal adjudication, starting with contract disputes. Our mission is to replace the traditional justice system with a faster, fairer, and more accessible AI-driven alternative. At the core of Delos is a powerful AI engine that can interpret contracts, assess facts, and deliver accurate, trustworthy decisions.