.png)
A conversation with founder Nabin Banskota on building its own AI models, using the PEO as an entry point and creating a unified platform for global employment.
AI and global employment are probably the two biggest stories in payroll right now.
Most vendors are approaching them as separate product initiatives. They are adding international capabilities through partnerships while layering AI assistants over existing systems.
Niural AI is taking a much more foundational approach.
The company spent its first several years building its own payroll, payments, benefits, and compliance infrastructure. It has also built its own AI models, supported by human review and outside accounting expertise, to operate inside that infrastructure.
It is one of the more ambitious product strategies I have come across in the market.
During my conversation with Niural AI founder Nabin Banskota, we discussed why the company chose the harder infrastructure path, why general-purpose AI is insufficient for payroll, and why the domestic PEO market has become the company’s most important entry point.
Banskota’s interest in payroll started with payments.
Earlier in his career, he managed global teams while working in financial services. He later founded Harvest, a consumer fintech company that used AI to negotiate consumer debt and was acquired by Acorns in 2021. He also spent time at TMF, where he saw the infrastructure behind global payroll more directly.
That experience convinced him that payroll remained one of the least disrupted areas of financial technology.
“So we really saw a tremendous opportunity to disrupt this space by building the infrastructure from the ground up for the next hundred years,” Banskota said.
Niural AI’s original thesis centered on bringing payroll processing and payments into the same infrastructure while also incorporating benefits. The goal was to create a platform that could support a company from its earliest days through an IPO without requiring it to assemble several providers.
That required a longer building period than most startups are willing to accept.
Banskota said Niural AI spent roughly two and a half years developing and refining the underlying infrastructure before meaningfully entering the market. The team believed that “everybody else in this industry just uses somebody else's stack.”
The first generation of global payroll platforms largely aggregated local providers. Customers received consolidated reporting and a single interface, while the actual payroll calculations continued happening across separate in-country systems.
More recent vendors have invested in owning more of the experience. Banskota still sees much of that progress concentrated at the interface layer.
“Some of those companies started to unify the global payroll platform, but a lot of the companies focused on the front end,” he said.
Niural AI focused on the systems performing the actual work.
“Nobody really focused on the hard part, which is the back-end stack,” Banskota said.
That stack now supports U.S. payroll, PEO, global payroll, EOR and contractor management.
“And now we have one platform that can be used for US payroll, US PEO, global payroll, EOR and contractors with a single login,” he said. “That is very rare to find even today.”
A general-purpose AI model can be useful even if it is only 90% accurate when the job is drafting an email or summarizing a document.
The same is not true for payroll calculations, tax interpretations, and employee payments.
Banskota described Niural AI’s target as 99.999% accuracy.
That requirement shaped how the company approached AI from the beginning.
“Without AI it would have been very, very difficult for us to build that stack,” he said.
AI helped Niural process and structure the enormous volume of tax, compliance and regulatory information required to build payroll infrastructure across jurisdictions. It allowed the company to develop the platform faster than would have been possible through manual research and configuration alone.
It could not be trusted to operate independently.
Banskota previously built an LLM engine at Harvest before widely available foundation models existed. That experience taught him how much work happens after a general model produces its first answer.
“In this industry, that's not even enough, right? You have to get it 99.999% accurate,” he said.
Niural AI uses outside foundation models where they are useful, but the company is also building its own domain-specific models.
“We built our own models,” Banskota said.
That is a rare strategy in HR technology.
Most vendors are building applications on top of models from OpenAI, Anthropic or other frontier labs. Niural AI believes those models can provide a starting point, but payroll accuracy requires additional training, validation and controls that are specific to the domain.
The company maintains humans in the loop and works with outside experts to review its interpretations.
“We also work with consultants, you know, consulting firms like accounting firms,” Banskota said.
This may be the most thorough consideration I have heard from an emerging vendor about how AI should be applied to payroll.
Niural is using AI to accelerate the difficult work of building payroll infrastructure. It is also surrounding AI with internal models, human review, and external accounting expertise before changes enter production.
That level of caution is necessary when the output determines how much someone is paid or how a company reports taxes.
The most interesting part of Niural AI’s go-to-market strategy may be its focus on the U.S. PEO market.
There has been a major gap in this category.
Global employment platforms have historically been stronger internationally than in the United States. Domestic PEO providers built strong payroll, benefits, and compliance offerings, but most were slow to respond as their customers became more global.
There has been some recent movement.
TriNet partnered with Multiplier to extend its international offering. Justworks acquired Via to add global employment capabilities. These moves recognize the customer demand, but they still rely on connecting separate platforms and infrastructure.
No established provider has approached the opportunity through a fully unified PEO, U.S. payroll and global employment stack.
Niural AI sees that gap as its wedge.
Many of its customers initially arrive for the PEO. As they expand, they can add contractors, EOR workers, or entity payroll without introducing another vendor. Companies can eventually transition from the PEO into an ASO or standalone payroll model while staying on the same platform.
That creates a particularly strong value proposition for a company with 80 employees today that expects to have 250 employees next year.
The PEO solves the most urgent need while giving Niural AI the opportunity to become the primary employment platform.
Banskota described the current PEO market as heavily dependent on legacy infrastructure that was never designed to support global workforces. International capabilities are often added through partnerships after the fact.
“The partnership model has always been in place in this industry for many, many years,” he said.
Those partnerships can expand coverage, but they also introduce handoffs across onboarding, support, reporting and ongoing administration.
Niural AI believes global employment will become a native expectation for PEO customers as more businesses hire across states and countries.
The PEO market is also where Niural AI can become the stickiest.
Contractor management can be replaced fairly easily. A PEO sits much deeper inside the business. It manages payroll, benefits, tax filings, and compliance obligations that companies are reluctant to move without a compelling reason.
Winning that relationship gives Niural AI an opportunity to expand with the customer as its employment model changes.
Owning the infrastructure also changes what Niural’s AI can do.
Many HR technology assistants can retrieve policies, summarize documents, or guide a user toward the correct workflow. They still depend on a person to leave the conversation and complete the actual task somewhere else.
Niural AI is trying to remove that handoff.
“If we have our own infrastructure, we're able to control what our AI can do,” Banskota said.
A user can prompt the system to hire an employee, make a payment or process payroll. The AI guides the user through the required information and can execute the action once it has been reviewed.
“Today we're probably the only executional AI in the payroll system,” he said.
The distinction between conversational and executional AI is important.
Answering a payroll question is valuable. Applying that answer correctly inside the payroll system is far more valuable. It is also significantly harder because the AI has to operate within the product’s permissions, compliance rules and underlying calculation engine.
Niural AI’s infrastructure investment gives it a legitimate opportunity to deliver that experience.
Banskota expects software interfaces to move further in this direction.
“The UI will completely change and become more of a prompt-first approach,” he said.
That does not eliminate the need for payroll systems. It raises the value of the systems underneath the prompt.
The agent still needs reliable data, deterministic calculations, and the authority to perform the requested action. Without control of the infrastructure, the AI can advise the user but may struggle to complete the work.
Niural AI recently formalized its AI investments through NiuralAI Labs.
The dedicated team is focused on developing what Banskota calls long-horizon agents. These agents are intended to manage processes that require several steps, decisions and interactions over an extended period.
Payroll processing is one example. More complex global employment workflows could eventually include employee separations in countries where notice periods, negotiations and legal requirements must be managed over time.
The goal of Niural AI Labs is to “have a dedicated team of folks that are building long horizon agents,” Banskota said.
Niural AI also sees an opportunity to become a specialized training environment for AI models entering payroll, HR and payments.
General models are trained on widely available information. Much of the knowledge required to run payroll accurately exists inside regulations, operational processes and edge cases that are difficult to capture through public data alone.
Niural AI wants to help close that domain gap.
Niural AI is attempting several difficult things simultaneously.
It is building domestic payroll and PEO infrastructure. It is supporting global payroll, EOR, and contractors through the same platform. It is developing its own AI models for a category where small mistakes carry real consequences.
Any one of those projects would be ambitious.
What makes the company particularly interesting is how closely the pieces depend on one another.
The unified infrastructure gives Niural’s AI the ability to take action. Its AI made building and maintaining that infrastructure more feasible. The PEO gives the company a sticky entry point with growing businesses that will eventually need the broader global platform.
There are more established providers in every individual category Niural AI is entering. The company still needs to prove that its infrastructure and service model can perform consistently as its customer base grows.
The strategy itself is unusually coherent.
AI and global employment are reshaping payroll at the same time. Niural AI has spent several years building as though the two were always one problem to solve.
