This is one of the problems I am currently working on through MAS.ai.
The starting question was simple:
Can observable changes in a company indicate that a legal need may be forming before the company actively starts looking for a law firm?
I do not think we have a final answer. But we have observed enough to keep working on it.
Why I find the problem interesting
Most legal business development begins relatively late. The question we started exploring was whether some of the events that eventually create legal demand are visible earlier.
The early version of the idea was too simple
A single event is usually ambiguous. The problem gradually changed from “Which events predict legal work?” toward “Which combinations of observations make a particular commercial hypothesis worth investigating?”
What we are actually trying to detect
We are not trying to predict that a company will buy legal services. The practical question is whether enough has changed around an organization that it deserves earlier commercial attention.
Current working model
Observable company change → possible commercial pressure → signal context → convergence / contradiction → possible legal need → law-firm relevance → lawyer relevance → decision whether to investigate or engage → observed outcome → learning
What we currently think we have observed
- Single public events are often too ambiguous to be commercially useful on their own.
- Combinations of independent signals appear more useful than raw signal counts.
- A possible buyer need and supplier fit need to be evaluated separately.
- Law-firm fit and individual-lawyer fit can diverge.
- The closer we get to learning from outcomes, the more important identity, lineage and evidence become.
- A rejected commercial action does not automatically disprove the original need hypothesis.
What we still do not know
How many observations are enough? Which combinations generalize? How should timing be represented? How should need and activation outcomes be separated? Which outcomes genuinely resolve which predictions?
What we deliberately do not publish
The purpose is to explain the problem and how our thinking is developing, not to publish proprietary weights, customer-specific logic or the full engine.
Try the idea
A later version of this page can include a small exercise: Which of three synthetic companies gives you the strongest reason to investigate further — and why?