Skip to content
All articles
Contracts & Overflow

How to combine AI contract tools with experienced counsel, and what to automate

The Goodlawyer Team

Combine AI contract tools with experienced fractional counsel by letting the tool do the repetitive first pass and having the lawyer own the judgment: what the risk means, what to accept, what to negotiate and what to tell the business. Use controlled automation where errors can be reliably detected and corrected, considering the consequences of the actual use. Keep anything that needs weighing competing interests, or that carries serious consequences if wrong, with a lawyer. This guide is not legal advice.

What professional guidance says

The American Bar Association's Formal Opinion 512, issued in July 2024, discusses competence, confidentiality, communication and fees when lawyers use generative AI, as summarized in the ABA announcement. The ABA is a professional association; its guidance is not a rule binding in every jurisdiction. Check the current professional rules and regulator guidance applicable to the lawyer and engagement.

Which tasks to automate and which need a lawyer

Sort each task by two questions: how easy is an error to spot, and how costly is it if missed?

Good candidates for automation or tool assistance

  • Extracting dates, parties and terms as a first pass; verify consequential renewal and notice dates against the source before relying on them
  • Flagging clauses that differ from your standard
  • First-pass summaries of long documents, checked by a person
  • Comparing versions and listing changes
  • Drafting routine correspondence for review
  • Classifying and routing incoming requests

Use the tool, with careful review

  • Redlining against a playbook, where a lawyer checks every proposed change
  • Drafting a first version from a template, where a lawyer reviews it in full
  • Research starting points, with every source verified

Keep with a lawyer

  • Advising the authorized business decision-maker on which risks to accept
  • Negotiation strategy and trade-offs
  • Advising the business on what a clause means for it
  • Anything involving privilege, regulators or disputes
  • Interpreting unclear or conflicting law
  • Final sign-off on non-standard terms
Output is easy to verifyOutput is hard to verify
Lower consequence

Controlled assistance

Use approved tools with a defined human check.

Check more carefully

Require source verification and allow time for review.

Higher consequence

Experienced review

A qualified lawyer verifies the output before use.

Human-led work

Keep the decision with qualified counsel; limit tool use to controlled support.

Data permissions, confidentiality and tool approval are prerequisites in every quadrant. This is a workflow guide, not permission to automate legal decisions.

How a fractional lawyer fits

An experienced fractional lawyer can play three roles in a tool-enabled workflow.

  1. Designer. Helps decide where the tool fits, writes the playbook the tool applies and sets what counts as acceptable output.
  2. Reviewer. Sets a risk-based review plan, verifies legally consequential outputs before reliance, and records errors. Sampling for workflow evaluation does not replace required review of individual advice.
  3. Decision-maker. Owns the judgment calls and explains them to the business.

Our Fractional InHouse service adds experienced counsel to your team. If tooling is part of the brief, agree which approved systems the lawyer will use, who supplies access and who reviews the output.

Rules for handling data

  • Decide which tools are approved and by whom.
  • Check what the vendor does with your data, including whether inputs are stored or used for training.
  • Do not put confidential or privileged material into an unapproved tool.
  • Tell outside counsel and providers what your rules are.
  • Keep an audit trail of what the tool did and who reviewed it.

Build the workflow in four steps

  1. Pick one contained task, such as contract data extraction or first-pass review of a standard agreement.
  2. Write the playbook the tool and reviewer will use.
  3. Run it side by side with your current process for a month and compare accuracy, time and issues.
  4. Expand only if the results hold, and keep a sample check.

Questions for a provider or candidate

  1. What tools have you used, and what was your review process?
  2. When did the tool get it wrong, and how did you catch it?
  3. How do you handle confidentiality with a tool?
  4. Which tasks would you not delegate to a tool?

Checklist

  • Task list sorted by error visibility and consequence
  • Approved tools and data rules written
  • Playbook ready
  • Reviewer and sampling plan assigned
  • Month-long pilot scoped
  • Measures: accuracy, turnaround, rework and reviewer time