Trey SCHMITT

Give artificial intelligence six pieces of information about a business and you may be surprised by how much of that company’s world can be reconstructed. A person’s name, company name, business address, telephone number, website and claimed line of business can be enough to begin tracing corporate registrations, ownership connections, registered agents, licenses, regulatory histories, court records, affiliated companies, government contracts, securities filings and addresses shared across multiple entities. What once required an investigator to manually search dozens of disconnected databases can now be organized into a single systematic investigation. The important distinction is that AI should never be treated as the evidence. AI finds the trail. Original records establish the facts.

That distinction changes how artificial intelligence should be used for due diligence. The weak approach is asking, “Is this company legitimate?” The stronger approach is telling the AI exactly what needs to be tested. Start with the six identifying facts and instruct it to search the exact business name, abbreviations, spelling variations, former names and DBAs. Then identify owners, officers, directors, managers, members and registered agents. Search those names individually. Search every address individually. Search telephone numbers, email domains and related websites. Determine when the company was formed, where it was incorporated, whether it is currently active and whether it is registered to operate in every jurisdiction where it claims to conduct business. Every new fact becomes another search term. That is where ordinary research begins turning into network analysis.

Imagine a fictional company called Atlas Financial & Mortgage Group LLC. Its website claims that the company provides mortgages, investment services, business loans and payment services. Most people searching the company might simply type the name into Google and read the first page of results. AI can be instructed to do something far more useful: break every claimed service into a separate regulatory question. If Atlas provides mortgage services, search for mortgage licenses and licensed mortgage professionals. If it provides securities or investment services, determine whether the company or individuals appear in securities-registration databases. If it transmits money, identify the money-transmitter requirements of the jurisdictions involved. If it describes itself as a bank, determine whether it actually holds a banking charter or whether “bank” is simply being used in a business or marketing name. One company making four different claims can create four completely different regulatory investigations.

Licensing is one of the strongest uses of AI-assisted due diligence because the AI can be asked to determine not merely whether a license exists, but what licenses should exist. That is a much more important question. Suppose a mortgage company possesses one valid license but advertises services that fall under three additional regulatory categories. Finding the first license does not automatically authorize the remaining activity. A proper investigation asks what activity is being conducted, what jurisdiction governs it, what license is normally required and whether the company or responsible professional possesses that authorization. The same method can be applied to insurance, securities, lending, money transmission, real estate, construction, healthcare, transportation and dozens of other regulated industries.

Then comes corporate mapping. Secretary-of-state records may reveal formation dates, corporate status, registered agents, mailing addresses, managers and officers depending on the jurisdiction. Suppose Atlas Financial’s filing lists an address at 100 Example Street. The AI discovers three other corporations registered at the same location. That does not mean those businesses are controlled by the same people. The address could belong to a law firm, corporate filing service, registered-agent company or shared office. But it creates a new research question. Who owns those companies? When were they formed? Do they share officers? Do their websites use the same telephone numbers? Is the same individual named in their filings? Do they operate within related industries? One address can be meaningless, or it can become the first visible point connecting an entire network.

Registered agents should be treated the same way. Finding the same registered agent across multiple companies proves very little by itself because professional registered agents can represent hundreds or thousands of unrelated businesses. But combine that information with a common officer, identical mailing address, telephone number, domain registration, former company name or overlapping management and the picture begins changing. AI is particularly effective at assembling these pieces into a relationship graph instead of treating every search result as an isolated fact.

The graph might look something like this:

Person → Company A → Registered Agent → Address → Company B → Officer → Company C → License → Regulatory Action → Court Case → Government Contract

Every arrow represents another question that can be investigated.

Suppose the president of Company A appears in an old filing connected to Company B. Search Company B. Suppose Company B shares an address with Company C. Search Company C. Suppose Company C received a government award. Find the procurement record. Suppose one officer previously worked for another regulated company. Search that company and the individual’s professional registration history. Instead of researching one business, AI can progressively uncover the surrounding business ecosystem.

Court records can add another layer, but they require discipline. A lawsuit proves that a lawsuit was filed. It does not automatically prove that the allegations contained in the complaint are true. AI-assisted due diligence should identify the court, case number, plaintiffs, defendants, filing date and disposition. Was the case dismissed? Settled? Tried? Is it still pending? Was there a judgment? Were allegations ever established as findings of fact? The difference between saying “a company was accused of fraud in a lawsuit” and saying “the company committed fraud” is enormous. Serious due diligence preserves that distinction.

Regulatory enforcement deserves the same treatment. If an agency issued a consent order, administrative complaint, suspension, fine or cease-and-desist order, retrieve the actual document. Determine what the regulator alleged, whether the company admitted or denied the allegations, what penalty was imposed and whether the matter was resolved. AI should be instructed not merely to summarize the record but to identify the primary document behind the claim.

One of the most valuable AI techniques is what could be called a contradiction audit. Instead of simply asking what can be found about a company, compare what the company says about itself against what public records show. If a business says it has operated for twenty years but the corporate entity was formed eighteen months ago, was there a predecessor company? If the website claims nationwide operations, where are the corresponding state registrations or licenses? If someone describes themselves as a licensed investment professional, where is the registration? If a company advertises mortgage lending, securities, insurance and money-transfer services, does the regulatory footprint support all four activities?

A discrepancy is not automatically wrongdoing. Companies restructure. People operate under former entities. Businesses acquire other companies. Licenses may be held through subsidiaries. Corporate records can be delayed or incomplete. The proper response is not immediate accusation. The proper response is another question.

That is where AI can be exceptionally valuable.

Ask it:

What should exist if this company’s claims are accurate?

That question turns ordinary searching upside down.

Instead of searching only for information that exists, the investigation begins looking for the records that logically ought to exist. If a professional is licensed, which regulator should maintain the record? If a company issues securities, what filings might be expected? If it receives federal contracts, what procurement records may exist? If it operates through multiple states, where should foreign registrations appear? If it claims a particular executive history, what archived filings or previous corporate records could verify that history?

This is due diligence through negative space.

What is missing can sometimes be as important as what is present, although absence from a database must never automatically be interpreted as proof of illegality.

AI can also reconstruct timelines. Put every discovered event into chronological order: company creation, address changes, management changes, website launches, license issuance, license expiration, lawsuits, government contracts, acquisitions, dissolved companies, new companies formed by the same individuals and regulatory actions. Events that appear unrelated in separate searches can look very different once arranged chronologically.

Consider a hypothetical example. Company A dissolves in January. One of its managers incorporates Company B at the same address in February. Company B launches a nearly identical website in March. A license previously associated with Company A expires in April. Company B begins advertising substantially similar services in May. None of those facts individually prove wrongdoing. Combined, however, they create obvious due-diligence questions: Was the business transferred? Were customers moved to the new entity? Were required licenses transferred or reissued? Was there common ownership? Why was the original company dissolved?

AI did not prove anything.

It showed the investigator where to look next.

The most effective investigations therefore separate findings into categories:

Verified Fact: Directly supported by an official filing, court record, regulator, government database or other reliable primary source.

Documented Connection: Two entities are connected through a verifiable officer, ownership disclosure, address, transaction, filing or other record.

Probable Connection: Multiple independent pieces of evidence suggest a relationship but do not conclusively establish it.

Investigative Lead: Something worth pursuing but not yet sufficiently verified.

Allegation: A claim made by another party, plaintiff, regulator, whistleblower or publication that must remain attributed unless independently established.

That framework prevents one of AI’s greatest risks: turning correlation into accusation.

A good AI due-diligence instruction could therefore begin with only six facts:

1. Full name of the principal person

2. Exact company name

3. Known business address

4. Known telephone number

5. Website or internet domain

6. Claimed business activities

Then give the system a much larger mission:

Search corporate registrations, former names, DBAs, owners, officers, directors, managers, members and registered agents. Search every discovered person and entity separately. Determine all licenses reasonably associated with the services advertised by the business and verify them against the appropriate regulators. Search relevant federal and state enforcement databases. Search court records. Search securities filings where applicable. Search government contracting records where applicable. Search sanctions and exclusion databases where appropriate. Search archived web pages. Search previous addresses and telephone numbers. Identify companies sharing those identifiers. Build a corporate timeline. Construct a relationship map. Highlight inconsistencies. Separate facts from allegations and possible connections. And most importantly:

Provide the underlying source for every material conclusion.

That final command may be the most important part of the entire process.

If AI says a person owned another business, ask for the corporate filing.

If it says a license was suspended, ask for the regulator’s record.

If it says two companies are affiliated, ask what document demonstrates the affiliation.

If it identifies a lawsuit, obtain the case number and court record.

If it identifies a government contract, obtain the agency, award number, date and amount.

If it identifies a regulatory penalty, obtain the actual administrative order.

If it identifies a sanctions match, confirm that identifying information demonstrates that the person or business is actually the same entity rather than simply someone with a similar name.

That is how AI becomes useful without becoming reckless.

The future of due diligence is not simply putting someone’s name into a chatbot and asking whether that person can be trusted. The real power comes from giving artificial intelligence a small group of verified identifiers and forcing it to expand the investigation outward.

Name.

Company.

Address.

Phone.

Website.

Business activity.

From there the AI begins searching the branches.

Companies reveal people.

People reveal other companies.

Addresses reveal additional entities.

Entities reveal licenses.

Licenses reveal regulators.

Regulators reveal enforcement histories.

Court records reveal disputes.

Government databases reveal contracts and exclusions.

Historical records reveal timelines.

And timelines reveal questions that might otherwise never have been asked.

The result can be something substantially more sophisticated than an internet search. It can become a business-network intelligence map showing who is connected to whom, which companies exist, which licenses are required, which licenses can be verified, what public records support the connections and exactly where unanswered questions remain.

That does not replace investigators, journalists, attorneys, compliance professionals or human judgment. It makes those people faster.

AI can search.

AI can organize.

AI can compare.

AI can cross-reference.

AI can identify inconsistencies.

AI can construct timelines.

AI can map relationships.

But AI should not get the final word.

AI finds the trail.

Documents establish the evidence.

Humans determine what the evidence actually means.

And that may ultimately be the real revolution in AI-powered due diligence: not that artificial intelligence suddenly knows everything about a company, but that six seemingly ordinary pieces of information can now become the beginning of an investigation capable of exposing an entire business network.