AI Adoption Playbook

Adoption Package

The set of documents a firm needs before it relies on an AI workflow: a process map, a governance memo, a use policy, a tested proof of concept, and a risk register. Each part answers a question the partner will ask, and none of them stands alone.

Process Map

A step-by-step account of the process from intake to closing, with the governing rule cited for each step and a judgment about whether the step should be automated, augmented, or left to a person. It is the foundation the rest of the package rests on.

Governance Memo

A short explanation of which laws, ethics opinions, and court rules apply to the firm’s use of AI on this process, drawn from the compliance mapping in Activity 2 and the duties in Activity 3. It tells the partner what the rules are before the policy tells the staff what to do.

AI Use Policy

A one-page statement of which tools may be used, what client information may enter them, who verifies output and how before a signer makes the Rule 2.515(d)(2) representation, how AI use is disclosed to clients, and how time is billed. It is the document 43 percent of firms do not have.

Proof of Concept

A working prompt chain or custom assistant for one step of the process, built with a skill document or standing instructions, grounded in the governing forms and rules, and tested on several inputs with the results recorded. It is what the partner can copy, paste, and try.

Evaluation Log

The record of each test run: the input, the output, what was right, what was wrong, how long verification took, and what changed in response. Without it the proof of concept is a demonstration, not evidence.

Risk Register

A table of what could go wrong, how likely and how serious each risk is, and what control addresses it: a hallucinated citation, confidential data in a public tool, a missed deadline, a biased output, an unbilled or overbilled hour. It is how the recommendation earns the partner’s trust.

Part 1: Map the Process

Write a process map of 500 to 700 words that walks the partner from the moment the client brings the debt to the firm through entry of final judgment. For each stage, say what must be done, cite the Florida Small Claims Rule or statute that governs it, and state whether the step is a candidate for automation, for augmentation, or for a person alone. Use the Process Mapper to draft and then verify every rule citation against the rules in the Resources.

Part 2: Write the Governance Memo and the Policy

Write a governance memo of about 400 words identifying the ethics opinions, court rules, and any statutes that apply to the firm’s use of AI on this process, including what must happen before client information enters a tool and what the signer of any filing represents under Rule 2.515(d)(2). Then draft the firm’s one-page AI use policy with the Policy Drafter, covering approved tools, client data, verification, disclosure to clients, and billing, and include the firm’s pre-signature verification checklist: the steps a person completes on every AI-assisted filing before anyone signs it.

Part 3: Build and Test the Proof of Concept

Choose one or two steps from your map that you rated automation or augmentation candidates, such as drafting the Statement of Claim from case facts, generating the filing and service deadline checklist, or drafting a motion for default judgment with its affidavit. Use the Proof-of-Concept Builder to write a skill document or standing instructions for a custom assistant, grounded in the clerk’s forms and the rules, and build it. Test it on three fictional cases with different facts, keep an evaluation log of every run, revise once, and run a confidentiality check (what client data the workflow exposes and to whom) or a bias check (whether outcomes differ for otherwise identical debtors) and record what you found.

Part 4: Risk Register and Recommendation

Use the Risk Register Builder to produce a table of the risks in your workflow with likelihood, severity, and the control that addresses each, then write a recommendation of 200 to 300 words: whether the firm should proceed, what a low-risk first use would look like, what training and double-check procedures to adopt, and the next two steps if the proof of concept holds up. Be honest; a recommendation not to proceed, well supported, is a good answer.

Part 5: Submit the Package

Compile a single PDF containing the memo to Catherine Reyes (the process map, the governance memo, and the recommendation), the AI use policy and verification checklist, the proof of concept as an appendix with the instructions, the three test inputs, the outputs, and the evaluation log, the risk register, and a reflection of about 200 words or a two-minute video on what you learned about applying AI to a real legal workflow, what was easier or harder than you expected, and what you would do differently. Post it in the Project discussion forum with your process named in the subject line.

Part 6: Share Chat Link

Include one AI chat link with a 1–2 sentence explanation of what the conversation shows and why you chose to share it.

Process Mapper

Act as a Florida litigation paralegal with years of small claims experience. Help me write a step-by-step map of a Florida small claims debt collection case from client intake through final judgment, for a claim on an unpaid contract, unpaid services, or promissory note. For each step, describe what is done and which documents are prepared, cite the Florida Small Claims Rule or statute that governs it, and note the deadline if there is one. Then, for each step, give your view on whether it is a candidate for automation, augmentation, or a person alone, with a one-sentence reason. Flag any rule citation you are not certain of so I can verify it against the current rules. Ask me for the county and any details of the firm’s practice before you begin.

A map with citations to verify. Requiring a rule for every step and a flag on uncertain ones makes the map checkable, and asking for the automation judgment step by step produces the analysis the partner asked for rather than a summary.

Policy Drafter

Act as a legal ethics consultant who writes AI use policies for small law firms. I will paste a governance memo describing the ethics opinions, court rules, and statutes that apply to a firm’s use of AI on one process. Draft a one-page AI use policy for the firm covering approved tools, what client information may enter which tools and with what consent, who verifies AI output and how, how AI use is disclosed to clients, and how time is billed when AI shortens a task. Then draft the firm’s pre-signature verification checklist for any filing prepared with AI: the steps a person completes so the signer can truthfully make the representation in Florida Rule of General Practice and Judicial Administration 2.515(d)(2) that every legal authority cited exists and is accurately cited. Tie each policy provision to the opinion, rule, or order in my memo, and do not add obligations that are not there.

A policy built from named duties. Tying each provision to a source in your memo keeps the policy from drifting into generic best practices, and drafting the verification checklist makes the abstract duty concrete.

Proof-of-Concept Builder

Act as a legal technologist who builds small AI workflows for law firms. I will describe one step of a Florida small claims process that I want to automate or augment, the clerk’s forms it uses, and the rule that governs it. Help me write a skill document or standing instructions for a custom assistant that performs that step: what to ask for, what the output must contain, which form it follows, what it must flag for a person to check, and where it must stop and wait for human verification, including the verification the signer of any filing needs before making the Rule 2.515(d)(2) representation. Then help me design three test cases with different facts, one of them deliberately tricky, and an evaluation log with pass-or-fail checks for each. Ask me clarifying questions before drafting.

Instructions, tests, and stops designed together. Writing the verification stop into the assistant’s instructions, and building the tests before the first run, is what separates a proof of concept from a demo.

Risk Register Builder

Act as a risk manager for a small law firm. I will describe an AI workflow for one legal process, the policy that governs it, and the results of testing it. Build a risk register as a table: each risk, its likelihood and severity on a simple scale, the harm if it occurs, and the control in the workflow or policy that addresses it. Include at minimum a hallucinated citation, confidential client data entering a public tool, a missed deadline, a biased output, and billing that does not reflect the time actually spent. Then tell me which risks my controls do not adequately address and what I would need to add. Be direct; do not soften a gap because the workflow is otherwise good.

Risks paired with controls. Naming the control for each risk forces you to check whether the package actually contains one, and the instruction to identify uncovered risks is what makes the recommendation trustworthy.