Merlin: Most Innovative Tech of 2024 with Gen AI at the Heart
Merlin's Disruptive Technology
Merlin Search Technologies is making waves in the legal tech sector after recently winning the ACEDS 2024 award for most innovative new technology of the year. Their SaaS platform, DiscoveryPartner, offers unparalleled Gen AI capabilities that set it apart in the eDiscovery market.
Here’s why Merlin Technologies stands out in the industry.
The Birth of Merlin Technologies
First, let’s begin with a bit of history about Merlin. John Tredennick, the owner of Merlin Technologies, is a renowned figure in the legal tech industry and a serial entrepreneur. In 2019, he sold his cloud-based eDiscovery platform, Catalyst, to OpenText for $75 million. Catalyst was one of the first platforms to implement CAL that could analyse and rank 1 million documents in milliseconds.
However, rather than choosing to rest on his laurels, John saw further opportunity for Gen AI in the legal tech industry and launched Merlin Technologies in 2020.
Sherlock: A Next-Generation AI Algorithm
DiscoveryPartner is an end-to-end eDiscovery and investigations platform that handles everything from processing to production. While it is a SaaS platform designed for firms to use in-house, Merlin also collaborates with industry partners who can provide project management support if required.
Sherlock is Merlin’s proprietary AI algorithm. This post focuses exclusively on this feature given this is a dedicated blog for all things related to AI. Sherlock delivers a powerful combination of functionality, flexibility and cutting-edge technology which is redefining document review.
Sherlock’s Key Features
1. Flexible Search and Review Options
Sherlock has three distinct approaches for first-pass review and AI-driven workflows:
Algorithmic Keyword Searching
Unlike traditional keyword searching, Sherlock enables users to ask factual questions (e.g. “What were the key issues in the case involving Company X and Company Y?”) for enhanced precision. The system presents 100 documents at a time and continues until no further relevant documents are found. Importantly, Sherlock avoids hallucinations and clearly indicates when data is unavailable.CAL Algorithm
Documents can also be fed into the CAL classifier based on simple user input, such as thumbs up/down selections or by boosting/suppressing documents that contain specific keywords. This helps to improve the AI classification accuracy.Prompting with Generative AI
You can create prompts to isolate relevant documents within minutes. While this might sound similar to some of the other Gen AI tools currently available in the market, there are the following significant distinctions:Choice of LLMs: You can select which LLMs to use for searching and reporting, such as ChatGPT 4o and ChatGPT 4o Mini. For example, smaller models handle early case assessment (ECA) investigations or document summarisations more efficiently at 30-40% faster speed and are 30-40% cheaper, while larger models are better for complex prompting or analysis at an additional cost. Alternatively, you might simply choose another LLM because one is too busy and slow on a particular day.
Customisable Precision and Recall Settings: You can adjust recall and precision settings before you run your prompt, e.g. stop when there are fewer than 6% relevant documents or stop if a certain number of documents is found, up to a maximum of 1,000 documents.
Full Transparency on Costs: You can easily check your LLM usage report that states your token usage on a case within a date range that helps with billing, e.g. 2 million tokens were used for this client for this case last month.
2. Key Document Summaries
Sherlock generates document summaries and snippets that highlight critical sections of documents. Snippets are summaries of the keywords within each document which provide you with additional context for your specific keywords. The snippets also contain clickable links to the original text in the document for users to cross-check and review.
3. Green Architecture for a Better Future
Hosting is charged on a per minute per GB model and Merlin's infrastructure is environmentally friendly as the workspace can be powered down when jobs are complete.
You can set timers for inactivity ranging from 30 to 120 minutes for the workspace server to switch itself off and save energy as well as reduce your hosting costs by up to 70%. It’s also easy to schedule dates for a matter to power down to save additional hosting costs, e.g. during the weekend, and then quickly power back up again when required, e.g. on Monday morning. The system can also be turned on and off manually at any given time and works within minutes.
This feature is a unique selling point that helps to reduce your carbon footprint and will contribute to a more sustainable and eco-friendly future.
4. Multi-Language and Advanced OCR
Leveraging AWS Textract, Sherlock offers AI-powered OCR capabilities that are far better than standard OCR for text extraction. It also works with documents in multiple languages and you can interact with Sherlock in one language and ask for a response in another to handle multilingual document sets much easier.
5. Enhanced Security
Merlin uses AWS’s Virtual Private Cloud (VPC) architecture and each case is isolated within its own VPC. This means the risk of data breaches is minimal because incidents are restricted to a single workspace and not exposed to the client’s entire list of cases.
Clients retain full control over data within their VPC which provides compliance with data sovereignty and privacy regulations.
It’s also important to note that your data is not used to train the LLMs used by Merlin and LLMs can’t remember or learn from your prompts. When you instruct the LLM to respond to a query, it is basing its answers only on the source materials within your workspace.
Conclusion
Merlin Technologies is redefining eDiscovery and investigations with its innovative AI-powered platform. By integrating CAL, advanced LLMs and green architecture, it addresses critical industry challenges while offering unmatched flexibility and security.
We expect Merlin to continue breaking barriers in the future. To learn more, we recommend scheduling a demo or watching the following short video on their website: Announcing Gen III: Merlin's Third Generation Generative AI Platform - Merlin Search Technologies.