Startups & Funding

Marble raises $5.5M for AI restaurant back-of-house work

Marble raised $5.5 million and signed 150-plus restaurants in under four months. Early customers reported 40% less food waste and 20-plus labor hours saved each week.

Marble Draws $5.5M in Seed Funding After Signing 150+ Restaurants
Marble Draws $5.5M in Seed Funding After Signing 150+ RestaurantsPexels / pixabay
By Elena Vasquez6 min read

Why it matters

  • Marble announced a $5.5 million seed round on November 17, 2026, led by Contrary.
  • Marble reported more than 150 restaurant signups in under four months.
  • Contrary reported early results of 40% less food waste and more than 20 labor hours saved weekly.
  • An A&W case study reported 97% forecast accuracy and inventory counts completed 90% faster.
  • Managers retain approval authority over Marble’s purchasing and staffing decisions.

By Elena Vasquez

Marble reported signing more than 150 restaurants in under four months while raising a $5.5 million seed round.

On November 17, 2026, Marble announced the financing for its New York-based startup. The company said the money would support an AI-native operating system for restaurant back-of-house operations.

Contrary’s investment announcement included early-customer results: a 40% reduction in food waste and more than 20 labor hours saved per week. The figures came from Marble’s early customers and were published by its lead investor. The immediate issue is reproducibility as Marble adds locations and gives Remy more work across restaurants beyond the deployments that produced the first reported results.

Marble set out a longer-term goal of “a self-driving back of house that learns from each count, correction and order, taking on more operational responsibility within controls set by restaurant operators.” Managers would keep final authority over purchasing and staffing.

What did the $5.5 million seed round include?

Contrary led the round. Y Combinator, Robinhood Ventures and Restive Ventures participated, according to Marble’s announcement. The startup also named the angel investors backing the company.

Marble’s angel group includes Sweetgreen co-founder Jonathan Neman, Smash Kitchen and Jinx co-founder Sameer Mehta, Momos co-founder Sai Alluri, ResQ founder Kuljeev Singh, Breadcrumb founder Seth Harris and AtoB co-founder Tushar Misra.

Together, the round links Marble to three venture firms, Y Combinator and six named angel investors with founding experience across restaurants and software companies. The mix brings sector knowledge and capital into a company built around restaurant data and automated operational work.

The round arrives as Marble expands its engineering, operations and growth teams. The startup still has to show that its early operating gains travel beyond the deployments that produced them.

How does Remy turn restaurant data into orders?

Marble combines inventory capture, demand forecasts, purchasing, food preparation and scheduling in one system. It divides that system into perception, reasoning and action.

Employees capture inventory with voice AI or computer vision. Document extraction and third-party integrations pull invoices, deliveries, sales and staffing records into the same operating picture.

Together, those inputs create a current view of stock, costs and activity across each location, giving forecasts and recommended actions a shared operating context for the same restaurant. That shared data connects physical inventory with the purchasing and labor decisions that follow.

Demand forecasts use sales history, weather, reservations, local events and limited-time offers. Marble says those forecasts guide purchasing, food preparation and staffing, giving managers a forward-looking view of what each location may need.

Marble’s product documentation says its ordering workflow compares vendors and live prices with expected demand and available stock. Its scheduling workflow accounts for demand, staff availability and labor rules.

Remy, the company’s AI agent, turns those forecasts into draft supplier orders, preparation plans and schedules. It handles several parts of the back-of-house workflow without requiring restaurant staff to move information between separate systems.

Marble says Remy runs in the background without requiring a prompt for each task. That separates continuous processing from step-by-step prompting, while managers still review the resulting drafts before acting.

What do the first restaurant deployments show?

Contrary’s announcement paired the rapid customer count with two operating results. Early customers reported a 40% reduction in food waste and more than 20 labor hours saved per week, according to the investor.

At Sophie’s Cuban in New York, staff count inventory each day by speaking Spanish into the app. The operator texts Remy to draft vendor orders and transfers between locations. The workflow brings a multilingual count and cross-location ordering into the same daily routine.

A separate A&W case study was published on July 5, 2026. The first site took six days to onboard:

  • Day one: Connect the point-of-sale system and import historical sales.
  • Days two through four: Map the menu, recipes, vendors and par levels.
  • Days five and six: Activate the AI agent.

Subsequent locations went live in hours, according to the case study. The rollout therefore moved from a six-day setup to much shorter deployments once the initial work was complete.

The case study reports 97% accuracy for item-level demand forecasts. It also says inventory counts ran 90% faster through computer vision and voice AI.

Those faster counts enabled daily stock takes. The workflow could use current inventory data rather than figures gathered between infrequent manual counts, giving the forecasting and ordering system a fresher input.

The implementation sequence treats data setup as much a part of rollout as the AI agent itself, even though Marble activated the agent only on days five and six. The first site therefore required a full operating setup before the software could act.

Why do managers keep approval authority?

Remy also reconciles invoices and flags data inaccuracies. Marble presents those checks as support for a drafting layer, not the final decision maker for important operating decisions.

Managers review and approve its supplier orders, preparation plans and schedules before purchasing or staffing decisions move forward. This structure leaves the agent responsible for preparation while people retain control over execution.

Those checks matter because Marble’s forecasts depend on inventory and sales records reflecting what actually happened in each kitchen. Keeping people responsible for approvals gives Marble a clear operating boundary, but it also ties the company’s performance to accurate data and manager oversight at each location.

For restaurant operators, the stakes are practical: software can prepare orders and schedules, but managers still control purchasing and staffing. Marble’s market case depends on showing that this operating model produces repeatable waste and labor savings.

Who is building Marble?

Chief executive Aakar Khanna grew up managing his family’s multi-unit restaurant business. He later covered restaurants and food and beverage as an investment banking analyst at Goldman Sachs.

Chief technology officer Arjun Chaliha previously led AI and machine-learning initiatives for electronic trading at Bloomberg. His background centers on automated decision systems and market data, according to Marble’s account of his career.

Y Combinator’s profile lists Marble as founded in 2026 and a member of its Summer 2026 batch. The startup entered the accelerator with early restaurant deployments already underway.

Marble said it is expanding its engineering, operations and growth teams toward its longer-term goal of a self-driving back of house. The additional staff will support more deployments and a broader set of operational tasks.

Both founders’ backgrounds place Marble at the intersection of restaurant operating experience and financial-market technology, a combination that matches the software’s focus on pricing, demand and execution.

Can the early results travel to more locations?

The seed-stage test is straightforward: whether the early results reported by Marble and Contrary can be reproduced as Remy takes on more work across additional locations. The answer will test both the software and Marble’s rollout process.

The approach depends on accurate data from each kitchen and on managers remaining responsible for approvals. It also requires restaurants to supply voice, vision and document inputs that feed the shared operating picture.

If Marble holds the reported waste and labor savings while adding sites, the seed round will have produced evidence for a repeatable restaurant operating system. If those numbers fade, the company will need to show that its controls work beyond the deployments that generated the initial results.

The next checkpoint is whether the 40% food-waste reduction and more than 20 weekly labor hours saved hold beyond early customers as Remy assumes more operational work.

Source: unite.ai

Share this article:

More from Elena Vasquez

Elena Vasquez

Show full bio

Market editor covering technology and product news in artificial intelligence and machine learning at AI In Context.

708 articles

Related articles

  1. Ex-BitGo and Galaxy Execs Raise $11M to Finance AI Compute
  2. Spines Raises $64M to Scale AI Publishing Platform for 8,000 Authors
  3. Voyage Labs Raises $12M Seed to Take AI Travel From Chat to Booking
  4. Atlassian's AI Adoption Is High. The Revenue Isn't Following Yet
  5. Isomorphic Labs raises $5 billion, its second mega-round in six months

« Previous articleNext article »