There is a lot of noise about AI in contract work right now, and most of it is not useful to the lawyer who actually has to ship the deal. Vendor pitches make every product sound essential. Internal champions oversell what the tools can do today. Skeptics undersell what they can do tomorrow. The lawyers stuck in the middle have to figure out which tools to pick, how to evaluate them, and how to get people to actually use them once the contracts are signed.
How to Contract hosted a webinar that cut through that noise. Host Matt Kohel, Partner at Saul Ewing and head of the firm's AI practice group, sat down with three colleagues who use these tools every day. Ricky Brooman is Director of Litigation Support at Saul Ewing and runs the firm's evaluation work on AI products. Tal Unrad is a Partner in the firm's Boston corporate transactional group who handles M&A and outside general counsel work. Tom Goodwyn is Counsel in the Chicago office whose practice covers real estate finance and the full range of real estate transactions. The mix of perspectives, evaluation lead plus two transactional lawyers using the tools daily, kept the conversation grounded in what actually works.
The conversation moved across where to start with AI, how to evaluate output, how to build playbooks and prompt libraries, how to structure a real pilot, how to drive adoption past the early adopters, and how to know when a tool has stopped earning its keep.
Here are our top ten takeaways from the speakers' comments during the webinar:
Start with the tasks that hurt the most. Look for high-volume, lower-risk work where the time cost is real and the consequences of a mediocre output are manageable. Redline summaries, abstracts, first drafts from templates, and provision searches across large document sets are the places to start. The early wins also build comfort with the tool, which matters when we want broader adoption later. We can always layer in higher-stakes work once the team trusts the tool.
Decide whether you want quicker or better on each task. These are different goals, and they require different workflows. Quicker output needs heavier back-end review from us. Better output requires more upfront vetting of the system. Knowing which one we are after on a given task is what keeps the tool from disappointing on the wrong axis.
Build a playbook that sets your boundaries. A playbook captures the limits we have already accepted, like insurance liability ranges and other negotiation boundaries, and becomes the gold standard we redline incoming drafts against. The progression is crawl, walk, run. Start with AI-assisted document analysis. Move to a curated database of agreements you actually like. Then build the playbook so the boundaries are baked into the workflow, not relearned every deal.
Save your prompts and share them across the team. A good prompt is intellectual property. The first version is rarely the right one, and by the tenth iteration you have something genuinely useful. Tools that let us save and share prompts turn one person's hard-won expertise into team-wide leverage. Use that feature deliberately.
Treat a new AI tool like a new hire whose only reference is its mother. The salesperson will tell us everything wonderful about the product. We will not see the negative things until we have tested it on tasks we already know the answer to. Start with questions where we can verify the output instantly, then progress to questions where a trusted colleague can double-check the work. The AI never gets above third or fourth-year associate quality, so human review is permanent.
Curate your inputs as carefully as you evaluate the outputs. Uploading a hundred-point term sheet and expecting a clean draft of a contract is a mistake. Pull out the four or five points that actually need to flow through and feed those instead. Same for long contracts. Feed the AI the sections that matter, not the whole boilerplate-heavy document. The output quality is downstream of the input quality.
Run a structured RFI to POC pilot, not a sales demo. Cull from five providers down to two for proof of concept. Use real users and gnarly real-world use cases for at least three weeks. Score on a one to five scale across consistent KPIs and capture the data in a spreadsheet so we can compare tools on the same paces. Pair the quantitative scores with weekly qualitative feedback meetings. The vendor relationship matters as much as the feature set.
Drive adoption with one-on-one coaching, not just group training. Different users sit at different points on the comfort spectrum on any given day. The lawyer asking for training wheels and the lawyer asking complex prompting questions both need support, and they need it in different formats. Group continuing education has its place, but adoption happens in one-on-one sessions and in colleague-to-colleague conversations. Someone hearing that a peer saved three hours on a task they also do is a more persuasive case than any internal training deck.
Pick a platform, commit to it, and stop changing horses. Software fatigue is real. If we keep migrating users to new tools, even good ones get rejected because people are tired of relearning the workflow. Drive awareness across multiple mediums, double down on the platform we have, and accept that the ten percent who will never adopt are not the audience. The eighty percent in the middle is where the work pays off.
Watch how vendors are integrating AI, not just whether they have it. The cloud transition produced winners and losers based on who handled the migration well. The AI integration moment is doing the same thing. Look for vendors that are integrating AI thoughtfully into the software's existing workflow, with frictionless input and output. Look for providers with longevity, versatility across domains, and the financial backing to invest in the latest LLMs as they come. Point solutions chasing one use case are higher risk in the current market.
Subscribe to Stay in the Loop
Whether you caught this webinar live or are reading the recap after the fact, our weekly newsletter keeps you current on upcoming How to Contract events, new recaps like this one, and the practical insights that come out of each conversation. Subscribe now so the next set of takeaways lands in your inbox.
