What we mean by wave two
This is our framing, not an industry standard. Wave one was people trying chatbots on legal questions and finding out where they fail. Wave two is quieter: firms put AI into the middle of existing workflows with rules around it. Large firms staff innovation teams for that. A small firm needs the same discipline in lighter form.
The one-person innovation team
- 1
Name an owner
One partner or senior lawyer is accountable for AI use. It can be you.
- 2
Choose workflows
Pick two or three: for example, summarising incoming correspondence, drafting client updates, preparing meeting notes.
- 3
Write one page
Approved tools, what data may be entered, who reviews outputs, how problems are reported.
- 4
Pilot
Run for 30 days. Record time saved after review and any errors.
- 5
Decide
Keep, adjust or stop each workflow. Then repeat.
Make the data ready first
AI works better and is safer when your information is structured. If matters, deadlines and tasks are scattered across inboxes and folders, there is little for any tool to work with. Putting them in clear databases is the unglamorous step that makes later AI useful. Tools like Notion's Custom Agents, introduced in February 2026, operate on workspace data, so organisation comes first. See https://www.notion.com/help/custom-agents.
That is what a template like LawyerOS Pro with AI ($39, one-time) is for: structured databases and a Custom AI Agents page. It does not supply an AI model; Notion AI features need a Notion plan or add-on that includes them.
What to avoid
- Buying several tools at once.
- Letting each person adopt their own tools with no list of approved ones.
- Measuring enthusiasm instead of time saved.
- Skipping the review step because the first few outputs were good.
Frequently asked questions
- Do small firms need an AI policy?
- A short written policy is wise. It clarifies approved tools, data rules and review, which supports your supervision duties.
- Who should lead AI adoption?
- A named lawyer with authority over how the firm works.
- How long should a pilot run?
- Around a month is enough to see real use and errors.
- Where do I start?
- With a repeated, low-risk task and a clear way to measure time saved.
Sources
About the author
Raghav Handa trained as a lawyer, then spent years building legal technology, from enterprise legal products to custom systems for law firms.
Keep reading
- InsightsLegal tech trends to watch in 2026 (and what to ignore)Legal tech trends for 2026 explained without hype: agents, workspace AI, consolidation of tools and verification, plus how small firms should respond.3 min read · 17 October 2026
- InsightsThe real cost of DIY AI inside a law firmBuilding your own AI workflows in a law firm looks cheap, but the real costs are time, supervision, security review and maintenance. A framework to count them.3 min read · 17 October 2026
- AIHow to automate parts of a law firm with AI, safelyWhich law firm tasks can be automated with AI and which should not be: intake triage, summaries, reminders and reporting, plus guardrails and a staged rollout.3 min read · 17 October 2026
- AINotion Custom Agents for law firms: what they can and can't doWhat Notion Custom Agents are, how law firms can use them for intake triage, deadline digests and matter updates, the costs, and the safeguards lawyers need.4 min read · 30 October 2026