Юристы запустили стартап с ИИ для унификации кадрового учета в ресторанах

For restaurant managers, employee problems have a way of happening at the worst possible time. An employee might arrive 10 minutes late on a busy Friday night, when the restaurant is packed and a manager has little time to stop and fill out paperwork. By the end of the shift, a manager might document the tardiness — or decide the employee worked hard enough afterward to let it slide. At another location, a different manager might write up the same infraction and cite the company handbook. It’s the kind of inconsistency experts say can create headaches for businesses, particularly if an employee later challenges a disciplinary action or termination. ADVERTISEMENT Two Connecticut employment lawyers believe technology can help address the problem. Ryan O’Donnell and Eliot Gersten, attorneys with Hartford law firm Pullman & Comley, have teamed with restaurant operators to launch COMP, a Farmington-based startup developing an app that uses artificial intelligence to help restaurant managers document employee incidents and apply workplace policies more consistently. The app’s AI assistant, called Kat, allows managers to describe an incident by voice or text. It uses the restaurant company’s policies, along with applicable state and federal rules, to help categorize and document the incident. The system can also track positive employee performance. O’Donnell said the idea grew in part out of his labor and employment law practice, where employers facing lawsuits sometimes insisted they handled an employee situation properly but lacked the records to prove it. ADVERTISEMENT COMP is designed to create those records while an incident is still fresh and before a busy restaurant manager moves on to the next problem. Gersten said that consistency can make COMP a risk-management tool for employers. The company, which is incorporated under the name Supersonic Limited, began developing COMP in early 2025 and tested early versions in restaurants later that year. It now has 15 restaurant groups representing 67 locations signed up for the platform. Some of those restaurant groups have locations in New York and Massachusetts, giving COMP a footprint beyond Connecticut. Restaurant groups pay a fee for each location that uses the app. ADVERTISEMENT The company has eight employees, including a full-time engineering staff, and also uses contract workers. The Connecticut Restaurant & Hospitality Association is also promoting COMP to its members. Not a lawyer Phil Barnett, co-owner of Hartford Restaurant Group, which operates the Wood-n-Tap chain, said his company has been using COMP for about six to eight months at some locations and is expanding its use. Hartford Restaurant Group has 11 locations in two states and about 850 employees, he said. O’Donnell said Hartford Restaurant Group and Newtown-based Market Hospitality Group, which operates restaurants in Connecticut, New York and Florida, were founding investors in COMP. Before the app, Barnett said, managers had to fill out performance improvement notices to document infractions. On a hectic shift, that paperwork could be delayed, completed inconsistently or skipped altogether. With COMP, a manager can record an incident as it occurs. Once the report has been reviewed and approved, the employee generally receives notice the following day and has an opportunity to respond. That delay can take some of the emotion out of the interaction, Barnett said, while giving employees a chance to explain their side. He said one general manager described COMP as his favorite among the various software platforms the restaurant group has implemented. However, the system isn’t intended to make employment decisions on its own, O’Donnell said. COMP uses a point system established by each employer for workplace infractions and progressive discipline. Managers still must review what the AI assistant, Kat, produces, and restaurant companies can require additional approval from another manager or human resources. COMP’s Chief Technology Officer Lucas Gray said Kat also doesn’t “learn” from one restaurant and apply that information to another. Customer data is kept separate, and the system draws from a defined set of materials, including the employer’s handbook and rules that apply in its jurisdiction. If Kat cannot find enough information to answer a question, Gray said, it directs the manager to an owner or human resources. O’Donnell stressed that Kat is not intended to serve as a substitute for an employment attorney. Allowing the AI assistant to practice law is a “big flashing neon no-no,” he said. “Frankly, a lot of our time was spent making sure that the app does not practice law,” he said. Potential risks When businesses use AI for employment-related purposes, human involvement is critical, according to Lisa Zaccardelli, chair of the labor and employment practice at Hinckley Allen in Hartford, who is not affiliated with COMP. Zaccardelli, who primarily represents employers, said AI can make employment processes more efficient, but it also introduces potential risks involving discrimination, privacy and a rapidly changing patchwork of state laws. Human oversight, she said, is “critically important,” along with training managers on how to use AI tools and evaluate the information they produce. Consistency in disciplinary practices and good documentation can help employers manage legal risk, Zaccardelli said, and a tool that simplifies those tasks could be particularly useful in service businesses such as restaurants, where managers face numerous demands on their time. But she cautioned against applying AI recommendations too mechanically. For example, an attendance policy might normally call for discipline after a certain number of absences or instances of tardiness. An AI system, however, might not know that an absence involves protected sick leave, a medical condition or another circumstance that could change how an employer legally should respond. Incomplete or biased information supplied by a manager could also affect the AI’s output. O’Donnell said COMP has sought to address those risks by building multiple levels of human review into the system. The app also does not collect demographic information such as an employee’s race. As COMP works to expand its customer base, it is also raising additional capital. O’Donnell said the company is not yet profitable but aims to reach profitability within the next 18 to 36 months. He declined to disclose its revenue or how much has been invested in developing the platform. The founders eventually hope to expand COMP nationally and see potential for the app beyond restaurants, particularly in industries where frontline managers face similar documentation and compliance responsibilities. For now, their focus remains on the hospitality industry. Barnett said restaurant operators are routinely approached by companies selling new technology, but COMP stood out enough that several industry operators have also become investors. “We get presented with software daily,” Barnett said, “but when you see something that actually makes your life easier and better — not just for the owner, but for the employees as well — it’s really cool to see that.”