Login Not registered yet?
Sing up

Iscriviti per ricevere i nostri articoli via email in tempo reale

Oops! Non è stato possibile trovare il tuo modulo.

Who We Are
Portolano Cavallo concentra la propria attività nel settore TMT sin dalla sua nascita nel 2001 ed è ora riconosciuto come lo studio legale italiano leader in questo settore.
The Firm
Portolano Cavallo Studio Legale was founded in 2001 by partners Manuela Cavallo and Francesco Portolano. The firm’s activity is mainly focused on the corporate/M&A sector and the media industry.
Eng / Ita
May 10



Our firm has recently celebrated the first anniversary of the implementation of a due-diligence [1] automation platform, which through the use of artificial intelligence [2] and machine learning [3] technologies, classifies and organizes documents by type of contract, governing law, language and various other criteria. We are writing this note to share thoughts on our own experience with this platform and how we envision these types of solutions impacting the M&A practice.

Some effects of the application of automated due-diligence solutions on M&A transactions are readily apparent (e.g., improved efficiency/accuracy and cost/time savings); however, we also believe that these could potentially have a deeper and more substantial impact on certain aspects of the M&A practice and dynamics.


Thanks to these types of technologies the parties involved in M&A transactions can significantly improve their knowledge of the target business. In our opinion, the degree of such “augmented knowledge” might be relevant for drafting and/or interpreting contractual provisions such as representations and warranties, and indemnification provisions. Given how new the use of these solutions is, those involved in transactions where machines support the due-diligence exercise should, in our opinion, carefully consider the effects the use of these technologies might have on the actual rights and obligations of the parties, both in the course of the negotiations and post-closing.

In M&A transactions, parties often engage in long negotiations to limit or expand the buyer’s indemnification rights. A buyer typically relies on the seller’s or target’s representations and warranties on the target business and/or entity, and breach of the representations and warranties triggers the buyer’s rights to seek indemnification from the seller or the target for the damages suffered as a consequence of such breach. Disclosures are qualified by the seller’s knowledge whilst representations and warranties are typically qualified by the buyer’s knowledge acquired either during the due-diligence exercise or thanks to the voluntary disclosures made by the seller.

Consequently the meaning of “knowledge” plays a critical role in the allocation of risks amongst the parties, and in fact the parties very often engage in long negotiations around knowledge qualifiers in particular when discussing whether knowledge should be limited to “actual knowledge” or should also include “constructive knowledge” or what a party should have reasonably known as a result of controlling the business and after conducting a sound due-diligence or enquiry on a specific matter. Knowledge qualifiers are relevant for both seller’s representations and buyer’s indemnification rights.

Any contractual provision qualified or impacted by a party’s knowledge should be carefully negotiated/drafted when an automated due-diligence solution comes into play. For example, the extent of both actual knowledge and constructive knowledge could be significantly expanded by artificially augmented knowledge. A liability a “normal buyer” could have reasonably missed, could become something a buyer supported by artificial intelligence could have reasonably known. As a consequence, the use of artificial intelligence could have the undesired effect of limiting the buyer’s indemnification rights by expanding the scope of the buyer’s constructive knowledge. In light of the interplay between actual knowledge, constructive knowledge and augmented knowledge particular attention should be paid while negotiating knowledge qualifiers in order to avoid undesired consequences on the allocation of risks amongst the parties. A buyer using an artificial intelligence solution and seeking to protect its indemnification rights might for example try to qualify knowledge as the knowledge an average buyer not using an artificial intelligence solution could have reasonably had of the target. Likewise, a seller using artificial intelligence to populate a virtual data-room should, in order to avoid a misrepresentation of a representation qualified by its best knowledge, carefully consider whether or not to disclose such a use to the prospective buyer and, if the use is disclosed, to define the knowledge as the knowledge of an average seller not supported by an artificial intelligence platform.

For the same reasons references to the use of these solutions in any part of the transactional documents (e.g. whereas clauses) should be carefully considered to identify potential consequences on the rights and obligations of the parties.

Some acquisition agreements include provisions for ”sandbagging” (i.e. a clause stating that the buyer shall be indemnified for breaches of representations and warranties made by the seller, whether or not the buyer had knowledge of such breaches prior to closing) or “anti-sandbagging” (i.e. a clause preventing the buyer from being indemnified for a breach of a representation and warranty that the buyer knew prior to closing). However, many acquisition agreements are silent on the impact of the buyer’s knowledge of certain issues prior to closing, thus the parties will look at the law governing the agreement to identify the answers to questions like: can a buyer with knowledge of a misrepresentation seek indemnification? And what about if the buyer should have reasonably had knowledge? Should a seller indemnify a buyer that knew of a liability which was unknown by the same seller? In some civil law jurisdictions, the seller’s indemnification obligations could be limited by the buyer’s actual knowledge and in some cases also by the constructive knowledge of a breach of representations and warranties.

In our opinion, the use of automated due-diligence solutions could have an impact on the answers to the questions indicated above. In a post-closing litigation, augmented knowledge could potentially be used by a seller to prove the buyer could have reasonably, or even easily, known of a specific liability; a court could order an expert to re-run automated due-diligence analysis to partially replicate the knowledge acquired by the buyer thanks to the automated due-diligence tool. More generally, while accessing a virtual data-room does not mean per se that buyers’ counsels actually reviewed all documents uploaded in the data-room (tracking features could provide evidence on whether or not a certain document was actually opened) and some circumstances could be used to reduce the extent of the constructive knowledge (e.g., if in a competitive bidding process documents are dumped in a virtual data-room at the very last minute right before signing one could reasonably argue to not have full knowledge of the information disclosed), however once an automated due-diligence platform is linked up to a virtual data-room one could argue that all documents were to some extent analyzed in a relatively short timeframe.

We mentioned above the sandbagging protection; a buyer that negotiated a pro-sandbagging clause may assume that it has very strong indemnification rights against the seller. However, in some jurisdictions the effects of a pro-sandbagging clause could be jeopardized if the buyer omitted to report to the seller it had knowledge of a specific circumstance unknown by the seller (and which the seller should not have reasonable known), triggering an indemnification right. Failure to disclose such knowledge could be construed by some civil law courts as a breach of the duty of good faith and the result could be the court not enforcing a pro-sandbagging provision. This is another area where augmented knowledge could potentially be used by a seller to prove the buyer’s actual or constructive knowledge of the facts underlying a certain claim; accordingly, it could be wise to clarify in the agreement how the augmented knowledge impacts (or does not impact) the application of the sandbagging provision.

Another possible consequence of the broader extent of the actual and constructive knowledge acquired thanks to the use of artificial intelligence could be a more frequent and more detailed use of special representations and warranties. Indeed, this might be a way for the parties to clearly allocate, and to avoid uncertainties around the extents of the parties’ knowledge, the liabilities arising from specific aspects of the target company.

Finally, the use of these platforms should raise higher confidentiality concerns. Combined use of artificial and human intelligence could provide a prospective buyer with knowledge of the target beyond the seller’s intentions.


Based on our experience we expect automated due-diligence platforms to play a role also in other aspects of deal-making:

Virtual data-room population. Preparation of the data-room is key for the success of an M&A transaction; but this can be a time consuming and somewhat cumbersome task. Failure to properly populate a data-room could slow down and potentially kill an M&A transaction or have uncontrolled effects on risk allocation. Diligence automation solutions could be used by lawyers/consultants to accurately classify documents based on desired criteria; this could significantly reduce the time lawyers invest on these low-value classification activities.

Impact prospective buyers’ monitoring. Virtual data-rooms usually include audit functions that allow for the tracking of documents accessed by specific users, time spent on documents and other information which is often used by sellers’ advisors to monitor and profile buyers in order to determine the most serious buyers. The use of automated due-diligence solutions could make such assessments misleading, as an interested buyer with artificially augmented knowledge could spend far less time in the virtual due-diligence environment as compared to a less interested buyer working without such artificially augmented knowledge.

Drafting disclosure schedules [4]. Preparation of seller’s disclosure schedules could be significantly simplified thanks to a diligence automation solution. Complete and accurate disclosure schedules are key to getting a deal done and are fundamental for the allocation of risk and liabilities amongst buyers and sellers: accuracy is key. The time saved thanks to a diligence automation solution could be invested by lawyers in refining, for example, special representation and warranties provisions.

Improved and faster decision-making process. The combination of an automated due-diligence solution and M&A lawyers/consultants could provide prospective buyers with more accurate and faster knowledge of the target thanks to a deeper and more efficient understanding of the underlying data-room documentation. As a result getting to signing a transaction may occur much faster; and parties with these tools may actually not only have an advantage from a diligence perspective, but also in an auction process as it would be easier for a buyer to pre-empt a deal and offer up a price and close out the competition if they can get their diligence done faster and more efficiently.

Faster underwriting of representations and warranties insurances. For all the reasons described above the use of automated due-diligence solutions could reduce the costs associated with negotiation of the policy and ancillary documentation for representations and warranties insurances and more in general make smother and faster the underwriting of those policies.


Although we are advocates of the use of these platforms we recognize that such application is very new to the legal industry and, therefore, the potential risks deriving by such are not clearly known ad of yet. In addition to the side effects discussed above there are other criticalities to be considered while adopting those new tools. Unfortunately, there are neither best practices nor industry standards to benchmark what is a safe use of these platforms; obvious risks include breach of privacy regulation, breach of confidentiality undertakings and potential risks associated to malpractice. Training of lawyers is also something which will potentially require some rethinking: lawyers of all seniority need to be trained to use these new solutions and to understand the implications of such use, but it is paramount to ensure that the use of artificial intelligence and machine learning does not prevent the most junior resources from getting enough exposure to the legal training deriving from the “manual” due-diligence process. In this respect, we believe that artificial intelligence and machined learning does not mean that there will be per se worse training of the junior legal force but rather that this will ultimately depend on how junior lawyers are instructed to use these new tools.


Exponential increases of computational capacity and better understanding on how to use it has recently created many new areas of application for new technologies in the practice of law. These, if used properly, can improve significantly the accuracy and efficiency of certain tasks currently performed without the support of machines. Law firms are client-centric and thus any tool that could actually improve accuracy and efficiency should be regarded as an opportunity for the legal profession. In addition, in most legal markets (the USA being the most notable exception to this) in-house teams will most likely have the financial resources and skill-sets to access these technologies before most law firms. Clients’ awareness of the benefits that artificial intelligence and similar solutions can offer means that lawyers will most likely need to embrace these technologies to meet the expectations of potential clients and to work on more complex transactions. At the same time lawyers should closely monitor the potential side effects the use of these technologies could have on several aspects of the legal profession.

[1] Due-diligence is the phase of an M&A transaction aimed at identifying potential risks in the target’s contractual documentation and gathering information on the target. Due-diligence is an extremely delicate activity as it is key for evaluation purposes, risk allocation and drafting risk-related sections of the transactional documentations (e.g., representations and warranties, disclosure schedules, indemnification provisions, escrows, hold backs, and many others).
[2] Artificial intelligence is defined by the Oxford English Dictionary as the “theory and development of computer systems able to perform tasks normally requiring human intelligence, such as visual perception, speech recognition, decision making and translation between languages.”
[3] Machine learning is defined by the Oxford English Dictionary as “the capacity of a computer to learn from experience, i.e. to modify its processing on the basis of newly acquired information.”
[4] A disclosure schedule is a document “that supplements the representations and warranties (and sometimes other provisions) contained in an agreement. The disclosing party (often the seller) typically uses disclosure schedules to disclose exceptions to, as well as to provide information that is too lengthy for inclusion in, the agreement. Disclosure schedules are typically attached to the end of the agreement and incorporated by reference.” [Practical Law]