How to Write a Great Hatch-Waxman Prompt for Relativity aiR

When you ask Relativity aiR to review Hatch-Waxman documents, a great rule of thumb is this - treat the software like a very smart fourth-grader. Tell it a clear story, use short words, and explain every special term the first time you use it. Below is an easy recipe our IntrepidX team follows after running live matters through aiR and seeing up to 60-70 percent time-savings over classic tech-assisted review (TAR).

1. Open with a easily understood case story

Start your Case Summary tab by answering five plain-English questions:

  1. Who is suing whom?
  2. What drug is involved? (give its brand name, generic name, dose form, and strength)
  3. Which patents are at issue? (list numbers and filing dates)
  4. Why are we in court? (e.g., The generic maker filed an ANDA saying its recipe won’t infringe these patents)
  5. What should the AI hunt for? (e.g., documents that compare the two formulas)

Keep this whole story under the 15 000-character cap that aiR sets for Prompt Criteria. (Relativity Help)

2. Write the way you’d teach a friend, not the way you’d brief a judge

  • Positive verbs. Say “Find documents that show bio-equivalence tests,” not “Do not find irrelevant docs.”
  • Lawyers love double negatives, AI does not. Avoid phrasing such as “not inconceivable” and stick with more straightforward, positive verbiage.
  • Short sentences. Aim for the length you’re reading now.
  • Bullet lists. They help a machine (like a child) see ideas fast.

These best-practice tips come straight from Relativity’s documentation. (Relativity Help)

3. Spell out every bit of insider lingo

Hatch-Waxman work is alphabet soup. The golden rule: never leave an acronym naked. Put the full phrase first, the shorthand in parentheses, then keep using the shorthand. Make sure you define common ANDA acronyms too, because unless you specify that FTF means “first to file” the AI may assume it is a more common acronym, like “face to face.”

Examples
“Active Pharmaceutical Ingredient (API)” the first time, then just “API.”
“First to File (FTF)”
“Launch Year (LY)- instead of the more commonly used “Last Year” ”
“Reference Listed Drug (RLD)”

Create a mini-glossary at the end of your prompt if you have more than a dozen special terms.

4. Give examples—tiny but true

aiR learns fastest when you paste one-paragraph snippets:

  • A snippet from your ANDA bio-equivalence report tagged Responsive.
  • A short outside-counsel memo.
  • A patent-office notice tagged Non-Responsive.

Two to three samples per tag are plenty. Remember, the longer your prompt, the less room you leave for your story inside the 15 000-character limit.

5. Use aiR’s explain-and-iterate loop

After the first run, aiR shows why each document hit or missed. Read those explanations like you would a junior attorney’s notes, tweak unclear criteria, and rerun. Because aiR needs no training set, you can refine in hours, not days. (IntrepidX)

6. Keep the structure super simple

A proven outline that fits aiR’s tabs:

  1. Case Summary – the story above.
  2. Key People & Companies – bullet list with nicknames and emails.
  3. Key Terms & Acronyms – your glossary.
  4. Responsive Definition – what “yes” looks like.
  5. Non-Responsive Definition – what “no” looks like.
  6. Examples – labeled snippets.

Copy-paste this into the Prompt Criteria box, press Run, and watch the AI fly.

Final thought

Good prompts are just good teaching. Tell your story plainly, define and code words, show a few examples, and let aiR handle the speed. When the machine understands like a fourth-grader, your Hatch-Waxman team can work at graduate-school pace.