AI-native patent services test whether software, expertise, and liability can fit within one company without diluting evidence, training, or customer remedies. getty
Most AI companies sell intelligence as a capability to the professionals they serve. The new category combines intelligence and service providers within one enterprise. This is an AI native service company that hires experts and contracts them for results and answers to their jobs.
That changes the way you test your business model. Model functionality remains important. Liability, insurance, training and customer redress will also be part of the product.
Lightbringer, a Swedish AI-native patent service, provides one early example. This is evidence of the potential of the structure, rather than evidence that it works on a large scale.
The product hardens
Legal technology companies typically license their tools to law firms. Lightbringer owns its platform, employs patent attorneys, and contracts directly with companies seeking patent protection.
Co-founder and CEO Dominic Davis describes the business as a “full-stack services company.”
“We own both the software and the lawyers,” Davis told me.
Therefore, customers purchase services from an integrated provider rather than software licenses used by another specialized company.
This difference may sound larger than it actually is. Integration alone can amount to process automation within a traditional service business.
Stronger arguments depend on four more difficult questions: who contracts, who signs, who bears the losses if the job fails, and whether the economics will withstand expert scrutiny.
Responsibility is harder than integration
Professional rules uphold the core duties of lawyers.
The American Bar Association’s Formal Opinion 512 sets out obligations such as competency, confidentiality, and oversight when lawyers use generative AI.
The USPTO’s AI guidance also treats AI as a tool to be used by responsible practitioners.
Davis said Lightbringer’s attorneys reviewed the work, approved the work and maintained professional responsibility.
The company says designated attorney status will be granted before the relevant firm prepares, reviews and submits each application. For applications in Europe, a European or nationally qualified lawyer will be used. A US registered physician is used for US applications.
This will identify the responsible expert. It has not been proven that companies will absorb the full financial loss of every mistake caused by AI.
Lightbringer says there is coverage available for professionals. Its standard terms include liability limitations, but business customers negotiate terms on an individual basis.
Insurance details are not disclosed. The available records do not indicate how disputed AI errors are classified, covered, or assigned in any particular case.
Consolidation can make it easier to attribute responsibility. The terms of the contract and insurance will determine how much risk remains with the provider.
Automation changes the field of decision-making
Lightbringer’s AI breaks patent work down into smaller decisions. Identify claim components, compare each to the prior art, test for novelty, and follow structured steps used in patent attorney training.
This method reverses the compression caused by experience. Tasks that feel intuitive to senior experts are translated into sequences that the software can try and review by humans.
AI doesn’t need to replicate professional intuition to change workflows. Once there is reliable performance on enough component tasks, the human role can be shifted to review, exception, and client judgment.
Davis rejects the comforting idea that automation only eliminates low-level tasks. “This happens to everyone,” he said.
The claim should remain narrow. Decomposition can compress work beyond seniority. It does not demonstrate equivalent or superior patent judgment.
Indicators that still evaluate companies
In its June 2026 Series A, Lightbringer reported more than 200 clients in 17 countries and a 300% year-over-year increase in second-quarter revenue.
These numbers were provided by the company. Number of clients and countries are cumulative since commercial launch in 2024.
It is claimed to deliver approximately 50% cost savings when compared to a flat rate subscription and typical hourly rates for equivalent work. Submission time refers to the time from drafting to submission, not the time to approval.
These definitions make the argument more readable. They still measure commercial traction, velocity, and price.
Customer testimonials are added to the G2 rating and designated case studies. We cannot show you how patents will work through examination, opposition, and litigation.
Stronger evidence would be tracking attorney time, material modifications, allowance rates, and contested outcomes over time. That evidence has not yet been made public.
The current record supports the claims of workflow and price changes. It does not establish good legal judgment.
Automation eliminates training opportunities
Professional judgment is built, in part, through repeated work by junior students. Automation can eliminate these tasks before companies create another route to experience.
Reviewing an AI-generated draft is not the same as building an argument, making your first mistake, and learning why you failed.
Enterprises may require simulations, intentional rotations, monitored exceptions, and exposure to failure cases that automated systems typically hide.
This issue extends beyond patent law. Professions that move humans from production to review need to explain where the next generation of reviewers will acquire their judgment.
One company, one test
A full-stack model is important when expanding beyond a single company.
Durable versions must integrate software and experts without hiding risks in contract terms, weakening standards of evidence, or hollowing out training systems.
Lightbringer has shown that AI companies can hire experts to sell their services. It has not yet been shown that this model yields better legal outcomes or assumes risk more fully than a regular company.
The next proof point occurs when the application fails, the claim is contested, or the attorney rejects the system’s output.
