Reverse Triangular Merger: 5 Checks for Reliable AI Analysis
M&A Transactions in Practice · AI Lab 2, Part 1
A reverse triangular merger can leave the target company alive while its shareholders lose control. AI can explain that distinction and draw the ownership flows. In this case, however, its confident first answer came after reading only about 262 of 407 KB of the agreement.
The useful question is therefore broader than “Did it cite a section?” You also need to establish which document it used, which provisions it actually retrieved and which numbers it calculated itself. This practical example follows the original Synopsys–Ansys agreement signed in January 2024. It is a historical teaching case about reading a contract with AI, not a statement of the eventual closing terms.
1. What is a reverse triangular merger?
In a reverse triangular merger, an acquisition subsidiary merges into the target. The target survives as a subsidiary of the buyer. Do not read “surviving corporation” as “economic buyer”: survival identifies the legal entity that remains, while ownership tells you who controls it.
In this agreement, Synopsys is the parent, ALTA Acquisition Corp. is its merger subsidiary and Ansys is the target. Section 1.1 provides for the merger subsidiary to merge into Ansys; Ansys continues. The share-conversion provisions explain how the parent ends up owning the surviving company. Those mechanics, rather than the deal’s headline label, support the structural description. See the original merger agreement, Sections 1.1 and 1.5.
Keep three questions separate when reviewing a reverse triangular merger picture: which entity disappears, which entity remains, and who owns the entity that remains. A diagram can answer all three, but a single arrow labelled “merger” usually cannot. For your own transaction, write out the answers before choosing the diagram’s labels.
Watch AI explain a reverse triangular merger
The original 1:37 explanation shows two uses of Claude: a cited explanation of the structure from project documents, followed by an animated diagram with three stages. It ends by returning to the first answer’s disclosure that the file read was incomplete. The next part of the Lab deals with reading the rest; this video does not show that repair being completed.
The value of this reverse triangular merger demonstration is that the answer, the picture and the limitation remain visible together. A readable diagram helps you understand the sequence; it does not expand the evidence the model actually saw. Treat the initial output as a working explanation to test against the document.
2. Five checks for a reverse triangular merger explained by AI
| Check | Ask AI to show | Your verification |
|---|---|---|
| 1. Parties | Parent, acquisition subsidiary and target | Match the names to the document’s definitions. |
| 2. Survival | Who merges into whom, and who remains | Open the cited merger provision. |
| 3. Consideration | Cash, shares and relevant qualifications | Check the original date, units and cross-references. |
| 4. Ownership | Source figures and calculation behind each percentage | Distinguish a contract term from derived arithmetic. |
| 5. Coverage | What it read and what it did not read | Resolve the missing material before wider reliance. |
2.1 Map the reverse triangular merger parties
Start with a small entity list. Put the legal name next to the document’s shorthand, such as Parent or Merger Sub. Then ask for an ownership description at the start of the sequence. That prevents the model from using the buyer’s name for a subsidiary that has a separate role in the agreement.
Keep shareholder groups separate from the companies. Former target shareholders may receive buyer shares without continuing to own the target directly. If an answer moves between the company and its shareholders without saying so, the resulting ownership explanation can sound plausible while answering a different question.
For a fresh reverse triangular merger case, your review note should record the document, its date and the relevant entity names. It should also flag aliases or renamed entities. This is a useful first task for AI because it turns repeated defined terms into a short working map. You still check the map against the source.
2.2 Check the reverse triangular merger direction
Ask the model for a complete sentence: which entity merges into which other entity, and which corporation survives? Compare that sentence with the cited provision. Reversing the two entities changes the structure even when the same three names remain on the slide.
The survival question is especially important in a reverse triangular merger. A reader may assume that the surviving company has bought the entity that disappears. That conclusion does not follow. The ownership relationships before and after the merger must be read alongside the legal sequence.
In your reverse triangular merger diagram, use different labels for a merger step and an ownership relationship. A merger arrow describes an event; an ownership arrow describes a state at a particular time. Ask for a legend if the distinction is unclear. Then compare the labels in every stage instead of checking only the final image.
2.3 Check reverse triangular merger consideration
The January 2024 contract’s base exchange terms were $197.00 in cash and 0.3450 of a Synopsys share for each eligible Ansys share, subject to the agreement’s qualifications. The amount and exchange ratio should appear separately in a review table, with the agreement date beside them. The source is Section 1.5(a)(iii), not the later diagram.
Do not silently turn signing terms into a claim about what holders ultimately received. The original agreement is one point in a transaction’s history. A later amendment or closing disclosure can change the answer to a question about actual payment. This exercise uses the original terms to explain the mechanism.
A precise initial prompt for a reverse triangular merger is useful here: ask for the cash amount, stock exchange ratio, relevant section and any cross-referenced exceptions. If the task needs a closing calculation, make that a separate request tied to the appropriate later documents. Do not combine different dates merely because the company names match.
2.4 Identify numbers the model calculated itself
The animated output includes an approximate 83.6%/16.4% split. In the source chat, Claude explicitly identifies those percentages as its arithmetic, not figures stated in the agreement. That distinction belongs next to the percentages, not in a remote footnote that disappears when the diagram is copied.
A useful reverse triangular merger analysis shows the inputs, their source dates, the equation and exclusions. An unsupported percentage can become more persuasive once it is placed on a clean ownership chart. Ask the model to identify every calculated label before treating the chart as a source document.
This is the same habit used when checking an AI-generated weighted decision matrix: inspect the inputs and calculation before relying on the ranking. Here, the output is an ownership percentage rather than a preferred strategic option.
2.5 Check reading coverage before widening the conclusion
The first answer reported that the project tool had cut off the file. It described reading Sections 1 through 5.7 and said later provisions and the Exhibit A definitions were not read. The video highlights about 262 KB read out of 407 KB. These are observations from this particular run, not a universal Claude file limit.
A section citation tells you where to check a claim. It does not establish that the entire document, every definition or every cross-reference was considered. A useful next question is what the model did not retrieve. Record the answer before converting the output into a memo or presentation.
The initial explanation also asserted that the portion read was sufficient for its structural analysis. That is the model’s judgment, not a substitute for your review. A broader question about conditions, termination or the consequences of a defined term may require material outside the retrieved portion. Do not present a limited first read as a comprehensive agreement review.
3. Rebuild the reverse triangular merger ownership calculation
The following reproduces the simplified calculation shown in the source demonstration. It is an exercise in tracing a number, not a reconstruction of actual closing ownership. The inputs came from the agreement’s share-count representations, and the model excluded equity awards and possible exchange-ratio changes.
- Start with 86,917,171 Ansys shares and multiply by the base exchange ratio of 0.3450.
- The result is approximately 29,986,424 new Synopsys shares.
- Add the 152,521,036 existing Synopsys shares used in the demonstration.
- Divide the estimated new shares by that simplified total: about 16.43%.
- The complement is about 83.57%, rounded in the diagram to 83.6%.
These steps explain where the picture’s labels came from. They do not make the exclusions disappear. In particular, the dates of the share counts, treatment of awards and any applicable adjustment need to be consistent with the question being answered. Keep the calculation labelled as simplified whenever it is reused.
There are also two different ownership percentages on the page. One concerns the parent’s ownership of the target after the transaction. The other concerns former target shareholders’ estimated ownership of the parent. State the denominator for both. Confusing them is an easy way to misread a reverse triangular merger diagram.
To adapt this reverse triangular merger exercise, ask for a compact calculation table with columns for input, value, source date, source location and whether the input was included. Then independently recalculate the fraction. If an input is unavailable, leave it unresolved; a precise-looking chart should not force a guessed value into the model.
4. Repeat the two original exercises
The reverse triangular merger practice guide adapts exercises 1 and 2 from the existing M&A Institute AI Lab 2 material. It keeps the original prompts, adds preparation and checking notes, and puts the MADEX course information on page 2. Use the explanation video first, then repeat the workflow with the historical public filings.
4.1 Exercise 1: ask for a cited structural explanation
Load the agreement into a project and keep the follow-up in the same chat. Ask who merges into whom, which company survives, what each eligible target share receives and who owns the target afterwards. Require a section number for each point. The objective is a short explanation you can verify, rather than a generic summary of what mergers usually involve.
After receiving the answer, open the actual cited clauses. Check that the entities, transaction date and share terms match. Then inspect any statement about document coverage. If it reports a partial read, record the limitation and obtain the missing material before expanding the conclusion.
4.2 Exercise 2: ask for three distinct stages
Request a diagram for before signing, the effective time and after closing. Identify the companies and both shareholder groups, and label the cash and stock flows. Ask which labels are quoted terms and which are the model’s calculations. In the original demonstration, the diagram opens as an interactive artifact with stage buttons and a Play control.
Check all reverse triangular merger stages. Before signing should not show a completed acquisition. The effective-time stage should show the merger event. The final stage should make the surviving company and shareholder relationships clear. A diagram is helpful when it preserves these distinctions; animation alone does not verify them.
Download the five-page practice guide and repeat the original prompts →
5. Put the reverse triangular merger review in a short working note
A good working note lets another reader reproduce your checks without reopening the whole AI conversation. Keep it separate from the diagram so the picture can remain readable while the evidence is still easy to find. For this reverse triangular merger exercise, a useful note has the following fields.
- Question and document: explain the structure using the January 2024 agreement, with the filing date recorded.
- Entity map: identify Parent, Merger Sub and the target; distinguish companies from shareholder groups.
- Event: state the direction of the merger and the surviving company, with the provision that supports it.
- Payment: identify the original cash and stock terms and preserve the relevant qualifications.
- Derived labels: attach the simplified ownership calculation, inputs, dates and exclusions.
- Coverage: record what was retrieved, what remains unread and which questions are still open.
These fields are a proposed review format, not an additional output shown in the original video. Their purpose is to stop a limited first read becoming a broader claim when someone forwards the picture. Mark unresolved items plainly rather than allowing a completed diagram to imply that all checks are finished.
When the next reviewer reads the missing material, update the note with what changed and why. Keep the earlier diagram for comparison, but identify the current version clearly. That creates a practical trail from the reverse triangular merger explanation to the source evidence and the later corrections.
Questions to resolve before using the answer
Does uploading a document mean Claude read it all?
No. The recorded case demonstrates a partial read despite the file being available in a project. Current Claude documentation on project retrieval also explains that RAG retrieves relevant material rather than loading all project content at once. The recorded truncation and the documented retrieval process are different observations; neither supports a blanket claim that every uploaded page was considered.
Can a citation prove that a reverse triangular merger analysis is correct?
A citation gives you a verification route. Open the provision and compare its actual meaning with the answer. Check cross-references and definitions where the conclusion depends on them. The model’s section number can be useful even when its overall document coverage remains incomplete.
Do reverse triangular merger percentages describe the completed deal?
No. The percentages reproduced here are the original AI demonstration’s simplified arithmetic using historical inputs. They exclude items identified in the source chat. They should not be relabelled as actual closing ownership or used as a current transaction forecast.
Can the same prompt be used on a different agreement?
Yes, as a starting structure for the exercise. Replace the source and entity names, retain the questions about survival and ownership, and require evidence for each answer. Different documents can contain different definitions, exceptions and mechanics. The checklist organises your review; it does not predetermine the legal outcome.
What should go into the handoff note?
Record the document and date, the question asked, the relevant citations, calculations and exclusions, and any material not yet read. Keep open issues visible when passing the diagram to another person. For the next stage of the workflow, use the same discipline when reviewing an AI-written board memo: the source limitations must travel with the recommendation.
The practical standard: use AI to make the reverse triangular merger easier to see and the agreement easier to navigate. Before relying on the result, verify the clauses, recalculate the derived labels and establish what remains unread.
Sources: January 15, 2024 agreement filed January 16, 2024; original M&A Institute AI Lab 2 exercises and saved demonstration; current Claude project documentation checked November 6, 2026. This educational example does not determine tax treatment, contractual consent requirements or the accounting acquirer for a live transaction.
