A Practical Starting Point For Small Law Firms – AI for Lawyers Fireside Chat Feedback
We really dumbed down (AI for Dummies) the idea of getting started with an AI tool/agent/app or even just a standardise prompt (Play books) for a lawyer who has never used AI before, to getting the 1st working AI solution off the ground in their firm.
No one size fits all, or “right click” and install the app silver bullet, but we tried to make it simple and straight forward.
Step 1 – AI Exposure
- Two-week AI familiarisation period recommended
- Use one AI tool for first week, with Gemini suggested as starting point
- Use second tool during week two to compare capabilities and prompting results
- Suggested tools included Gemini, ChatGPT, Grok, Copilot, Claude, and Kimi
- Low-risk personal and administrative use cases encouraged
- Email drafting, document summarisation, language learning, article headings, graphics, recipes, DIY guidance, and email triage
- Private, personal, client, and firm information must be excluded from free AI tools
- Free-tool token limits can be managed by switching between AI LLMs
Step 2 – AI Policy
- Smaller firms urged to create formal AI usage policies
- Policy should address approved tools and permitted use cases
- Confidential and client information prohibited unless appropriate protections exist
- Human review required before AI-generated work leaves firm
- AI policy development proposed as an AI-assisted, cross-checked process
- Identify frequent AI user in your firm to draft initial policy
- Review draft with your different AI tool using fictional or generic information instead of private valuable information
- Final policy to cover training expectations and disclosure of AI assistance to clients
Step 3 – Pilot Solution
- Firms should select a paid AI plan with data protections before handling firm work
- Preferred tool should not use firm data for training, that is why you opt for a paid for plan.
- Pilot opportunity should be repetitive, high-volume, time-consuming, or costly
- Low-risk process with a clear human review point preferred
- Pilot candidates include legal research, document generation, contract review, discovery summarisation, client intake, and administrative work
- Additional options include billing support, precedent searches, marketing content, scheduling, form completion, and client updates
- Desired output and source data must be clearly identified before development
- Existing document libraries can support document-generation workflows
- Pilot performance should be measured over 60 days (2 months)
- Compare time and cost before and after AI implementation
- Include AI-tool cost in pilot assessment
- Record failures, fixes, and unresolved issues before deciding whether to expand, improve, or abandon pilot
Scaling And Support
- Successful pilots may expand from limited users to wider firm rollout
- Possible improvements include scheduled data retrieval, automated reports, CRM integration, calendar appointments, and document assembly
- Expansion decision follows two-month assessment
- Malcolm Pearson offered follow-up guidance on potential firm use cases
- Firms can send proposed processes for review
- Support may include selecting suitable applications and shaping pilot design
- Copilot agent-building course shared as additional learning resource (35:31)
- J said course remains free
- Skills and agent-building concepts can transfer to tools such as Claude
- Malcolm Pearson agreed to review resource
Sources of more information
Learn how to build an agent with Microsoft CoPilot
https://founderz.com/skilling/agent-a-thon-fz/sign-up/
Guidelines and structure for an AI Usage Policy in your firm
https://legalaigovernance.com/resources/policy-templates/
Give me a shout on malcolmp@tech4law.co.za is you want to discuss a solution in your firm – no strings, no charge.
