As we rewrite the rules of business in this decade, artificial intelligence (AI) assumes the limelight. Legal professionals believe that a lot of promise offered by AI in law is unproven with several use cases unable to deliver. However, one area easier to embrace, adapt, and prove is digital discovery (erstwhile called e-discovery). It involves AI, other analytics, automation of workflows to handle new sources of data, expanding volumes, and additional capacity for legal professionals.

AI-Enabled e-discovery

Given the volumes of data, both structured and unstructured that lawyers must otherwise manually analyze, AI can have a drastic impact on the actions, outcomes, costs, and efficiency of each case. According to American Bar Association, 62% of lawyers already use some variant of traditional e-discovery solution and this level of adoption is easier to AI-enable and automate unlike other legal use cases of AI.

AI Use Cases in e-discovery

AI can reduce the burden on legal professionals by identifying patterns, relevant phrases, sentences prone to omission when traditional methods. There are many low hanging opportunities with AI and e-discovery, let us discuss a few below:

Case Research

AI helps sift through high volumes of data to identify key documents and associated information adding speed and efficiency to legal teams. It can also benchmark whether a claim is legit, strong, or worthy. If terabytes of data can be analyzed to chart the strength of the case to focus on what matters, it is worth its weight in gold.

Structuring Data

Managing an e-discovery case is complex, there are numerous documents and facts with a lot of details e.g., contract scope, date of the breach, ramifications of violation, etc. Using clustering techniques, AI can group patterns of terms that appear together while creating multiple angles on each case issue. Unstructured data being clustered into logical packages for analysis simplifies the data and imparts speed that humans cannot achieve.

Technology-Assisted Review (TAR)

The most cumbersome and expensive part of e-discovery is the review stage. AI can expedite classification, tagging, relevance, etc. AI also learns from human actions for additional keywords or facts of interest and this additional data can train review models for higher ranges of accuracy which limits the time humans have to spend on these mundane tasks.

Digital Discovery Analytics

The discovery process is intense and deep, people can lose sight of their depth of analysis and go past the point of diminishing returns. These efforts add costs are rarely quantified, compared, or improved. KPIs correlation activities, costs, and financials are seldom analyzed for productivity, efficiency, or returns. AI can compile this information in real-time and roll it up into a visual layer eliminating recurring problems and enhance the cost efficiency of the entire function.

Data Redaction

Data redaction is one of the key concerns e.g., PII, consent, etc. Withholding data and privileged, anonymous, redacted, etc. is time-consuming and AI alleviates this issue. Shielding data from disclosure with high accuracy and greater levels of compliance is another important use case for AI in digital discovery.

Risks and Opportunities

There are several risks and opportunities with using AI in digital discovery, we can however categorize them into two broad themes with opacity being a perceived risk and cost being the opportunity.

Cost Reduction

AI enables substantial savings on the e-discovery spend by cutting human effort, avoiding rework, reducing errors, and imparting speed. Smaller and less complex cases can be optimized by automating research, reviews, redaction, etc. AI can also bring forth effort and spend patterns through real-time analytics enabling focus and operational decisions.

Explainability Risks

When data is limited, expensive to collect, or in unknown formats, it becomes prohibitive, and hence, traditional rule-based, automation workflows could prove better than AI. On top of this, cost, access, and limitations of data can affect accuracy. It takes massive amounts of data to train AI models. These constraints push legal departments back to legacy methods.

Finally, the opacity within algorithms creates a lack of explainability. Clarity by explainability is one of the most important factors in decision-making and a major area of discomfort for lawyers.

Future of AI-Enabled e-discovery

AI use cases are rapidly evolving and changing the way the e-discovery process is designed, operationalized, and delivered. The future brings multiple shifts to this space.

Native AI

Today, AI is layered on top of existing discovery processes i.e., it is not native to it. The integrations into ESI (Electronically Stored Systems) must be deeper. The extraction of information, categorization of data, clustering, and contextualization by case will be seamless in the future. Deep learning algorithms will simulate human thinking and the e-discovery software (not the AI module) will learn to learn driving exponential value.

Upstream Adoption

AI must be leveraged more upstream in the EDRM process, deeper integrations will enhance visibility into data sources before collection. In the future, AI will deploy sophisticated data mining algorithms to widen scope outside of ESI and correlate them with context, custodians, and anomalies. These algorithms will inform AI-based recommendation engines to identify additional custodians to be involved or placed on legal hold without compromising defensibility. Pushing AI more upstream will create significant downstream efficiencies by reducing data volumes sent out for review. Integrating AI upstream will also allow legal teams to set case strategies and tactics as legal teams transition from insight to foresight.

AI-Enabled Orchestration

AI will evolve from its tactical roles today i.e., collector, curator, advisor, etc. to an orchestrator. AI will strengthen its role as an advisor through recommendation engines proposing additional documents, custodians, etc. In the future, AI will be an orchestrator learning from past actions, correlating actions and decisions across multiple channels. For example, if the relevance rate of document review was X% then AI will help fine-tune the collection parameters, this learning can be leveraged to target the collection and map them to the right reviewers. AI, as an orchestrator will also help automate tasks, predict case, and sentence outcomes based on its own learning.

Concluding Thoughts

AI creates a range of opportunities for legal professionals to drive value in digital discovery. E-discovery is more suitable with acceptance higher than other areas of Legal Tech for using AI as it eliminates mundane human tasks like targeted collection, early data assessment, relevance, issue categorization, and quality control while preserving human bandwidth in areas of higher value.

About the Author

Nitin Kumar is a two-decade veteran in the Hi-Tech industry. His career has spanned many intrapreneurial and entrepreneurial roles in executive positions across Software, Services, and Management Consulting. His passion is propelling organizations to greater levels of success through strong relationships and differentiated products. He is considered a business builder, thought leader, and pioneer of many innovative approaches. During his operating and consulting roles, he has grown, transformed, and pivoted many software businesses and scaled businesses successfully. Nitin is currently the CEO of Ligl, a pioneer in digital discovery orchestration and is an author of many articles, and is often seen speaking at conferences on AI, IoT, SaaS, and Blockchain. He is a veteran CEO and has added value to 1,000 plus M&A transactions in his career.

Introduction With any new technology, adoption and usage make the journey from emerging to being mainstream and voice is clearly crossing that chasm. Apple, Google, Baidu have reported hundreds of millions of devices using voice, and Amazon has 200 million users. Adoption is typically uneven, but the penetration levels of smartphones accelerate that journey when it comes to Voice Tech. Given voice is not used in the same way by everyone, it creates opportunities to spread across context and experience continuums to find a sweet spot for deeper adoption. Use cases, context and method of access matter and playing a big role- voice adoption also differs by how people access it. While voice assistants are the media favorites, smart phones, in car assistants etc. are also very prevalent and scaling.

Altering Business Paradigms Voice is literally changing how business works and customers think about products. Especially given we are in the middle of a global pandemic, the contact-less use cases have suddenly scaled by replacing touch intensive graphical user interfaces or buttons. With platform players like Siri, Alexa, Cortana, and Bixby creating broader use cases, the enterprise players like Nuance, Houndify etc. create niche opportunities with white labelled solutions. The crowding of this market creates risk of rapid commoditization, but value is always created by context, user relationships and customer experiences which build the importance of skills like VUX (Voice User Experience) and VUI (Voice User Interface) specialists. Poor experiences can jeopardize use relationships, contaminate the brand, and impede adoption. Misunderstanding voice-based requests (poor ASR or NLU) can pollute the context itself e.g., answering an out of context question due to misinterpretation. Let us examine a few such improvements, these will enable various business models in the voice arena.

Boost Customer Interactions One of the goals of voice technologies is to develop deeper personal relationships with the users and customers. The level of personalization possible with voice is much higher than the graphical user interface or the chatbot era. The sheer variety of systems, context and behaviors are the biggest challenge to automation or unified customer experience, voice can play a huge role in integration the technology ecosystem of hardware, software, and service together.

AI Enabled Automation The convergence of Voice Tech and Artificial Intelligence create a myriad of opportunities as Voice Tech benefits from machine learning. AI is going to be embedded deep into the context enabling voice interfaces to easily leverage the scale. For examples, voice assistants can teach themselves how to solve problems better and quicker with your customers.

Sonic Marketing Voice Tech is well on its path to be front and center of search engine optimization. It has started to migrate organizational thinking from text-based to voice-based SEO. As voice-based searches continue to increase, sonic marketing strategies must weave themselves around voice. Sonic marketing is more than search, it also spans voice content e.g., podcasts and sonic branding i.e., using audio to reinforce brand recognition. Voice-enabled ads provide advertisers with real-time measurement capabilities, enhancing the appeal of sonic marketing. Lack of in-house skills and knowledge make it a bit hard for early adoption of voice-based marketing strategies, but there is wide acknowledgement of the large opportunity it presents.

Voice Enabled Products Voice has morphed from being a small feature on a smartphone to being able to control large ecosystem of products and services. They can eliminate the issue of multiple apps on multiple devices and siloed operation of ecosystems. Customers expect products to have native integration with their platforms of choice. All new generation products will have to be voice enabled for them to stay relevant and competitive.

The COVID-19 Catalyst Coronavirus has provided an ideal catalyst to voice adoption. Given concerns on the spread, everyone from retail stores, to banks and hotels etc. have either created or planning contactless products, services etc. Voice technologies provide the perfect alternative to touching buttons or screens to interact and transact.

Business Models Voice Tech is the unifying glue between a single ecosystem or multiple ecosystems enabling efficient distribution and monetization. Voice is also a way to quickly weave service economics around hardware or software. Voice can create voice enabled hardware infrastructure to drive larger adoption, create software modules to help developer communities to voice-enable applications or enhance existing services to provide better integration with products in the ecosystem.

Platform Based Platform plays have been around for a while, it is about creating of a seamless interaction on a two-side marketplace and monetize transactions. Ability to gain critical mass on each side of the platform and engineer interoperability, interconnections through meaningful curation is the key. The hardware market is very disparate e.g., bulbs, plugs, home appliances etc. Voice assistants can make the integration seamless; the business model eliminates the need of many separate mobile applications into one voice interface. The key is to get as many hardware vendors to adopt the platform while driving adoption by the users leveraging platform business model. Monetization can be attained through the hardware vendor or the user, for now most cases are the former paying for services while user paying for voice enabled hardware like Alexa or Google Assistant.

Voice First Business Models As voice adoption scales with voice only applications or voice enabled hardware, platforms will amplify reach, monetization, and scale by enabling third party solutions like marketing, analytics, security, automation etc. Almost all the Big-Tech players have embarked on creating such ecosystems to scale. KPIs to measure these ecosystems have been around # of skills e.g., Alexa, usage of skills, transactions enabled by ecosystems, # of ecosystems enabled e.g., smart home, smart car etc.

Transaction Based Business Model The transaction-based business models have been popular for a while, we can see voice-based transactions scaling across value chains such as payments, advertising, marketing, food delivery etc. The basic economic value underpinning the business model is a piece of the transaction fee, this can get to significant levels of scale as ecosystems grow their reach. The transaction-based business model can scale across hardware, software and services creating new experiences across complete value chains for companies and consumers alike.

Voice-Data Interchange Model The voice and data interchange are a great way to monetize Voice Tech. The difference between voice and data being most data generated by sensors today are monetized by sensor creator, but voice is monetized by parties other than the creators. In future, I see the advent of voice brokers aggregating queries from multiple sources and monetizing them as packaged intelligence. For examples, voice archives can become an archived investment to corelate with data patterns as augmented intelligence for better targeting etc. There is also a market for real time voice commentary, such as weather, disaster, traffic jam information etc. One of the emerging business models is micro- monetization of voice e.g., AaaS or “Answers as a Service”

Conclusion Voice is a fast-growing interface enabling better experiences through automation, integration and weaving service economics around hardware and software ecosystems. Voice Tech is at the crossroads between multiple context, intelligence and experience continuums and will continue to scale and evolve. There do exist a few barriers and challenges to overcome such as voice privacy when it is aggregated by cloud players, lack of fully baked data sets in the right volumes to train models in the enterprise, voice security still being immature and limited open source communities etc.