Agentic AI in Law

A Live Working Session with Prof. D. James Greiner from Harvard Law School

Recently, I worked through the first phases of the A.G.E.N.T. Framework with Paski. The process was not about building “more AI.” It was about understanding where legal workflows actually break down.

One insight became very clear:

In evidence-based immigration, the real bottleneck is not drafting.

It is evidence collection, contextual preparation, and structured evaluation before strategy even begins.

That is where Meritocrat is focused.

The attorney frames the evaluation context.

Applicants prepare through structured questions, evidence organization, portfolios, and readiness evaluation.

Over time, that preparation becomes evaluation metadata.

Not just uploaded documents.

Not just intake forms.

Structured context that stays with the case throughout the pre-filing journey.

This creates a better foundation for:

• legal research

• petition drafting

• strategic guidance

• gap analysis

• even future RFE preparation

One thing I appreciated from the framework was stress-testing the workflow against real-world constraints:

• What is automatable?

• What still requires human judgment?

• Where does nuance matter most?

For immigration, especially EB-1A, EB-2 NIW, and O-1A, the answer is clear:

Some cases are straightforward.

Some cases are deeply evidence-heavy and require nuanced judgment.

That “gray area” is where structured evaluation systems can actually help.

Not replacing attorneys.

Helping organize complexity before legal strategy begins.

Grateful for the opportunity to think through the workflow more deeply and continue building systems around context-aware evaluation in legal tech.