A signal is a reason to look, not a reason to sell

I have spent the last few work sessions working backwards from a question: before a company asks for legal advice, what might already be changing in its business?

The usual temptation is to start with information sources. Find more news, scrape more company pages, collect more announcements, and hope that a sufficiently large stream will reveal opportunities.

We tried a different order. First, understand the possible legal need. Then ask what might cause it. Only then ask what an outsider could plausibly observe.

That became a much larger exercise than I expected.

From legal specialties to possible buying situations

For Entriva.ai, we organized commercial legal work into 18 main practice areas and 390 underlying specialties.

Then we worked through five distinct business situations for each specialty: 1,950 legal-need hypotheses. They range from setting up a subsidiary or hiring a chief executive to refinancing debt, changing a supply chain, entering a regulated market or responding to an environmental incident.

Five per specialty was a deliberate way to force coverage, not a claim that every specialty has exactly five relevant situations.

The next pass asked what a company, the media or a public institution might reveal about each situation. That produced 5,842 possible links between market signals and need hypotheses.

Those are links, not 5,842 unique signals. The same announcement can be relevant to several different legal questions. A planned acquisition, for example, could point toward transaction work, financing, employment, tax and competition law.

It is a large working map. It is not a list of verified buyers or confirmed mandates.

The difference between a signal and a useful lead

One distinction repeatedly changed the analysis: before, during or after.

A company announcing a planned factory expansion might provide a useful early indicator for construction contracts, land rights, financing or environmental permissions.

A lawsuit reported in the press may tell us that a dispute is already under way. A final judgment may arrive much later still.

In several practice areas we could not identify a plausible public early warning for many of the hypotheses. Internal investigations, confidential negotiations, private employment conflicts, undisclosed security incidents and some insurance disputes can remain invisible until the need is well established.

I would rather record that absence than invent predictive power.

Even a genuine early public signal does not mean the legal work has not begun. The company may already have internal counsel or an external firm. We are identifying something worth investigating, not proving that an opportunity is available.

We changed the starting point again

As the map grew, another practical decision became clearer.

Entriva should begin with a defined list of companies to follow, not with an undifferentiated river of data.

For each company, the initial working sequence is:

1. Examine the company's own public digital surfaces: website, news, reports, investor materials, relevant LinkedIn activity, careers pages and other official channels. 2. Examine news and press releases about that company, including relevant industry and local reporting. 3. Examine additional open sources such as company registers, permits, procurement notices, regulatory decisions, legislation and political developments. 4. Compare what is found with the existing signal and legal-need hypotheses.

And when an external source points to something interesting, return to the company's own channels. A media report about expansion might lead to a project page, an investor presentation, a recruitment campaign or a partner announcement that changes the interpretation.

General legal and political changes need a second route: identify which companies on the watchlist may actually be affected. A new rule is not automatically a company-specific legal need.

This is a working model, not a locked architecture. We expect to revise it when real monitoring teaches us where information appears first.

Give a possible source the benefit of the doubt

We also chose a deliberately generous rule for source discovery.

If a public digital source might reasonably contain a relevant signal, keep it as a candidate even before we have demonstrated that it does.

An untested source is not a useless source. A weak indication may become more informative when considered alongside other independent observations.

But that generosity applies to including sources, not to claiming facts.

A company opening an office does not prove that it needs outside counsel. A regulator asking questions does not establish wrongdoing. An ESG disclosure does not demonstrate that a legal problem exists.

We can be open-minded about what to examine and conservative about what we conclude.

What changed

We now have an initial body of working documents connecting legal specialties, possible business situations and candidate public signals. We also have a source-discovery framework and a company-first monitoring sequence.

That is tangible work, which I have linked under Built.

The larger question remains under Problems I'm working on: can a system take a possible need all the way through useful commercial engagement to an actual purchase, without a human doing each step?

The research map is one component of that ambition, not proof that the autonomous sales process works.

What I think matters

The value of the map is not the size of its numbers. It is the discipline of separating a public event, a hypothesis about a need, a genuine decision to seek advice, and a willingness to buy from a particular supplier.

Those are four different things.

If we blur them together, we can produce an impressive stream of activity while learning almost nothing about demand.

What remains uncertain

We have not yet tested how often the proposed signals appear in accessible, permitted sources, how early they arrive, how well they identify the right company, or how often they correspond to real legal instructions.

The 5,842 links are not deduplicated or empirically validated. The source map still needs concrete source-by-source work. Some useful channels may have access restrictions, uncertain reuse rights or poor geographic coverage.

And, most importantly, even a technically accurate signal may have no commercial value if the company has already solved the problem, does not need external advice or is not interested in the supplier.

That is the next test for Entriva.ai.

A signal should earn us a better question about a company. It should not give us permission to pretend we already know the answer.