Gemini Enterprise Legal Preview
Malcolm Pearson reported that Google Gemini Enterprise for Legal is available in South Africa in preview, meaning it remains in an evaluation and testing phase rather than a full release.
The service operates within Google Cloud and is intended for sensitive-data environments such as law firms and other regulated industries.
Setup is expected to require a Google Cloud authorized partner, rather than a simple self-service installation by an individual lawyer.
Preview pricing was described as $35 per month, approximately R630, or $30 per month when committing annually through a $360 once-off payment; final pricing is expected to change and may become usage-based.
Legal AI Integration Scope
The product was described as supporting legal tasks including contract review, redlining, playbook creation, legal research, document redaction, NDA drafting, data discovery, and contract negotiation acceleration.
Its MCP connectors are intended to connect Gemini with systems and applications including Microsoft 365, Google Workspace, DocuSign, iManage, NetDocuments, Everlaw, LAGORA, Harvey, and Thomson Reuters.
The stated objective is to let agents work with data held in connected systems, local data, and network shares.
Malcolm Pearson noted that privacy controls and training restrictions depend substantially on using the enterprise licensing tier rather than a free version.
Local AI and Thomson Reuters Progress
Thomson Reuters was reported to have launched an LLM based on the Chinese Qwen 3.5 and 3.6 models, with an open-weight option.
The model was described as having initially been trained on 10% of Thomson Reuters’ data.
Open weights can allow installation on a local desktop, subject to sufficient processing capability and storage; one demonstration reportedly required approximately 25 GB.
J added that local-hosted AI is advancing rapidly and that an example reportedly ran major models on a laptop using 2 GB of RAM, provided the necessary data was available on a connected hard drive.
Gemini Deployment Complexity
A key implementation blocker is that Gemini Enterprise for Legal runs within Google Cloud, which may be too complex for ordinary lawyers to configure directly.
Google reportedly recommends using a certified or authorized Google Cloud partner for setup and configuration.
The preview status creates uncertainty because the current offering is not yet a final release.
Pricing is also unsettled: the current $35 monthly structure may be replaced or supplemented by usage-based billing.
AI Billing Model Tension
Malcolm Pearson described a potential shift from standardized or fixed legal fees toward variable AI-related costs driven by usage and token consumption.
This creates a tension between clients’ preference for predictable pricing and the possibility that intensive AI use increases the cost of a specific matter.
F raised the counterargument that AI may reduce total service-delivery time and cost, potentially offsetting the AI expense.
The suggested pivot was to use transparent engagement terms explaining that AI may be used where beneficial, while retaining human review and responsible oversight.
Confidentiality, Legislation, and Integration Gaps
PM raised concerns about intellectual property and strategic case-planning information entered into ChatGPT, even where uploaded pleadings and notices are generally public.
PM also identified uncertainty about forthcoming South African AI legislation or regulation.
A practical workflow gap remains between email, ChatGPT, and AJS: the desired process would interpret an email, prepare a response using the relevant matter context, and record billing in AJS, but that interface was not available.
PM noted that AI outputs, particularly case-law research, require careful verification because the system may cite case law that does not exist; accurate prompting and human checking remain necessary.
Evaluate Gemini Enterprise for Legal
Review the URLs and product information Malcolm Pearson offered to share.
Consider a short trial of the preview offering at $35 per month, with the option mentioned of using it for a month and then cancelling.
Assess setup requirements, authorized-partner dependency, legal workflow fit, privacy controls, and possible future usage-based pricing.
No deadline before the next meeting was explicitly set.
Share Local AI Research
Owner: J. Investigate the reported laptop deployment of current Claude, Gemini, and ChatGPT models using 2 GB of RAM and locally connected storage. Identify who conducted the demonstration and how the local setup was implemented.
Send the findings to Malcolm Pearson by email.
Review Transcription and Workflow Tools
PM may assess alternatives to the current transcription workflow, including Fireflies, which Malcolm Pearson recommended based on perceived accuracy and action-note quality.
Continue human review of transcripts where accuracy is important before feeding them into downstream systems.
Explore future integration between email, AI matter analysis, response drafting, and AJS billing.
Action items
Malcolm Pearson
- Share Gemini Enterprise URLs
Gemini Enterprise for Legal
https://cloud.google.com/blog/products/ai-machine-learning/introducing-gemini-enterprise-for-legal
Video:
https://www.youtube.com/watch?v=ct5JDCb0-BA&t=57s
AI Legal Insurance Cover
https://news.bloomberglaw.com/legal-exchange-insights-and-commentary/legal-ai-insurance-exposes-that-practice-of-law-needs-defining
Thomson Reuters new Thomson 1.0
https://www.thomsonreuters.com/en/press-releases/2026/august/thomson-reuters-leverages-its-world-class-data-assets-to-launch-its-own-frontier-model
