Supply-chain planning has long depended on spreadsheets, forecasts, and teams of employees making decisions about what a company should buy, how much inventory it needs, and where that inventory should be kept. However, an AI Startup Wants to Take Supply Chain Decisions Out of Human Hands
Atomic, an AI startup founded by former Tesla employees, believes many of those decisions can now be handled by software.
The Boston-based company is building an AI-powered supply-chain platform designed to analyse changing business conditions, simulate possible scenarios and determine how much inventory a company should have and where it should be positioned. In some cases, the system can go beyond making recommendations and make the decision itself.
That approach is already being used by companies including DoorDash and HelloFresh, as Atomic moves from testing its technology with pilot customers to handling real operational decisions.
Atomic’s origins can be traced back to Tesla’s difficult Model 3 production ramp in 2018.
Its co-founders, Michael Rossiter and Neal Suidan, were working at Tesla when they developed an early version of the system. At the time, rapidly changing production requirements were difficult to manage with traditional spreadsheets, creating a need for software that could continuously evaluate different planning scenarios.
That experience eventually became the foundation for Atomic.
Rather than relying on employees to manually analyse thousands of possible combinations, the company’s software is designed to model the operating environment and determine possible paths through it.
Rossiter, Atomic’s CEO, describes supply-chain management as an enormous optimisation problem in which conditions can change constantly. The company’s approach is to use AI to search through those possibilities and identify potential decisions.
The distinction is important because Atomic is not positioning its technology simply as another AI assistant.
Traditional enterprise software can tell an employee what is happening. AI tools can also recommend what the employee should do next. Atomic is attempting to take the process one step further by allowing its software to make operational decisions automatically.
That shift is already happening among some of its customers.
Jon McNeill, a former Tesla president and founder of DVx Ventures, where Atomic was incubated, told TechCrunch that DoorDash is using Atomic for about 90% of its purchasing across hundreds of sites. The platform is also being used by HelloFresh.
For businesses handling food and other products with limited shelf lives, inventory decisions can have immediate consequences.
Ordering too little can leave products unavailable when customers need them. Ordering too much can create waste, spoilage and unnecessary costs.
Atomic’s software is designed to continuously adjust those decisions as conditions change rather than requiring employees to repeatedly analyse the data themselves.
The company is also expanding beyond food and delivery.
Atomic says it is working with consumer packaged goods companies as well as businesses in mobility and manufacturing. The manufacturing focus has particular significance given that the technology itself emerged from the production challenges its founders encountered at Tesla.
One of the challenges with enterprise AI has traditionally been getting new systems to understand how individual companies actually operate.
Many organisations have decision-making processes that exist only in the experience of employees. The rules may never have been formally documented, even though staff follow them every day when deciding what to purchase or how much inventory to maintain.
Atomic says its agentic AI can identify those decision rules and use them to operate within a company’s existing processes.
According to McNeill, this has helped reduce the amount of work required to onboard customers. Once the system began understanding how employees were making decisions, customers started asking why the AI could not simply make those decisions itself.
That is the larger bet behind Atomic.
Instead of creating another dashboard that requires an employee to constantly monitor supply-chain data, the company wants AI to become part of the decision-making layer itself.
The idea also fits into a broader shift in enterprise technology, where AI systems are increasingly being designed to perform tasks rather than simply generate information.
For supply-chain teams, that could mean moving away from a model where employees spend significant amounts of time updating spreadsheets, comparing forecasts and manually adjusting purchasing plans.
Atomic’s growth suggests that businesses are beginning to experiment with that model. McNeill said the company’s annual recurring revenue has increased fivefold since the beginning of 2026.
That growth has also attracted fresh capital.
Atomic raised $12.5 million in a Series A funding round led by Klass Capital and Madrona Venture Group, taking the company’s total funding to slightly more than $15 million. The startup has also brought in Jeff Goodrich, a longtime Tesla planning director, as chief technology officer and its third co-founder.
The new funding gives Atomic more room to expand its technology across industries where supply-chain decisions remain heavily dependent on human planning.
Its ambition is ultimately larger than automating inventory spreadsheets.
The company is attempting to build software that understands a company’s operating model, evaluates changing conditions and determines what should happen next, with less need for human intervention.
That could change the role of supply-chain teams.
Instead of spending most of their time deciding what to order, where to put it and when to reorder, employees could increasingly oversee the systems making those decisions and focus on the exceptions that require human judgment.
For Atomic, the lesson began at Tesla, where rapidly changing manufacturing plans exposed the limitations of traditional planning tools.
Years later, the former Tesla team is applying that lesson to a much broader market.
The goal is not simply to put AI inside the supply chain.
It is to make AI part of the decision-making process itself.
