Exclusive: Restoke.ai brings AI ‘restaurant manager’ to the US

Melbourne-based restaurant automation start-up Restoke.ai is expanding rapidly into the US and other markets as it builds what it calls an "AI restaurant manager" to help venues protect margins and tackle day-to-day operational complexity.

Co-founded by CTO Ken Brand and veteran restaurateur, Assaf Stizki, the company uses machine learning, natural language processing and live data streams from point-of-sale and inventory systems to predict demand, automate back-of-house workflows and flag where restaurants are losing money.

Founders' mix

Brand grew up in Israel, working as a developer from a very young age and spending two decades in natural language processing and machine learning before moving to Australia in 2013.

His co-founder, Stizki, has spent around 20 years in hospitality, building and operating catering businesses, burger chains and other venues.

The idea for Restoke.ai emerged in Melbourne when the pair began eating their way around the city's restaurant scene. Brand's co-founder had built an early prototype in WordPress to tackle some of the day-to-day pain of running venues.

Brand saw a real operational problem that could be addressed with more advanced AI, leading the pair to turn the prototype into a full platform just as the Covid-19 pandemic disrupted hospitality. They used that period to pilot the system across multiple venues, learning from live operations and iterating quickly as restaurants faced intense financial pressure.

Brand describes restaurants as operating on "the tiniest of margins", with owners and managers often lacking formal training in how to run complex operations at scale. Restoke.ai's goal, he said, is to "bring the heart back to hospo" by using data and automation to remove some of the drudgery and firefighting from back-of-house work.

From insight to action

Restoke.ai initially focused on predictive analytics before moving into what Brand calls "true agentic automation". That shift was enabled by access to significantly more operational data as the customer base grew into the thousands, allowing the company to identify recurring patterns across venues.

Brand said the team built its own operational and orchestration layers on top of large language models to move from read-only insights to a system that can behave more like a restaurant manager - understanding workflows, creating tasks and triggering actions within fixed constraints.

The company positions its platform as a "hospitality solving machine", using data from many individual, traditionally siloed venues to identify what works and what does not across the sector. That includes common pain points such as stock management, prep timing and staff task allocation.

When AI steps in

Brand said Restoke.ai is designed around a "human in the loop" principle, gradually earning trust by starting with small corrections and moving carefully towards greater automation. The system weighs the criticality of decisions and its own confidence levels before acting.

In cases where a fast decision can materially affect the business - for example, an impending out-of-stock on a key ingredient during service - the platform can both inform staff and automatically create tasks, such as preparing more stock or adjusting orders, provided this fits within the restaurant's defined workflows.

If confidence is low, the system prompts staff for confirmation, even where venues have opted into greater automation. Brand said Restoke.ai is particularly focused on avoiding changes that touch the "creative, strategic" front-of-house experience, treating menu design and brand expression as the restaurant's intellectual property while concentrating on back-end optimisation.

Forecasting mushrooms

Restoke.ai's forecasting relies on live data streams from point-of-sale systems combined with operational inputs such as stocktakes, supplier orders and prep records. The company builds a constantly updated estimate of inventory, then models expected depletion and replenishment.

Brand gave the example of mushrooms used on pizzas: Restoke.ai can detect that a venue typically orders mushrooms on a Monday and again on a Thursday, but is predicted to run out on Wednesday. The system can then recommend bringing the next order forward and automate the communication with suppliers while triggering internal workflows.

He said performance improves the longer a venue uses the platform, as the model learns its patterns of sales, prep, wastage and ordering. However, the system is designed to provide useful recommendations "out of the box" using broader industry data and clustering similar venues together.

Engineering under pressure

Brand described a core engineering rule at Restoke.ai: "don't break the venue". That imperative has shaped the platform's data architecture and decision-making logic.

The team has shifted from batch data processing to streaming, enabling split-second decisions such as sanity checks on production volumes just before orders are placed. This is complicated by the messy reality of restaurant operations: delayed updates, missing records, unreported deliveries and recipes that change without the system being told.

To handle this, Restoke.ai includes internal cross-checks and a reconciliation layer to infer the most likely state of operations from incomplete information. Confidence scores gate automated actions and trigger human confirmations where needed.

Brand also stressed the need for transparency. Rather than relying on a "black box", the platform exposes its reasoning so operators can see how it arrived at particular conclusions or recommendations, which he said is essential to winning trust in a high-stakes, low-margin environment.

Beyond the menu

Brand said Restoke.ai does not automatically rebalance or alter menus, leaving product, flavour and customer experience decisions to chefs and owners. Instead, it surfaces insights on performance, such as dishes that consistently lose money or fall short of industry-standard food costs.

The system can propose tactical responses, such as bundling, portion changes or time-limited specials to clear overstock. It may recommend using surplus ingredients in new specials or limited-time dishes but stops short of automatically deploying those changes.

The platform also tracks operational metrics such as prep timing and procedures, especially for multi-venue groups where sites are implicitly experimenting with different approaches. Restoke.ai can identify which methods produce better financial results and automatically update certain back-of-house timings and processes.

Clusters and co-pilots

To understand the nuances of individual restaurants, Restoke.ai groups venues into clusters based on attributes such as cuisine, format, location and operational characteristics. This allows learnings from one set of venues to be applied, with care, to others that behave similarly, even if they serve different food.

Onboarding is designed to be fast, with Restoke.ai ingesting purchasing, scheduling and sales data, then automatically categorising and mapping it. Over time, the system refines its view of each venue's workflows, prep practices, shift patterns and staff performance.

Brand said many early customers are embracing the agentic features more quickly than expected, using the platform as a task manager for ordering, prep and general to-do lists distributed across staff iPads and phones. He likened it to "an extra pair of hands" that frees managers to focus on reviews, customer interactions and leadership rather than constant crisis management.

Global push

While most of Restoke.ai's early customer base has been in Australia, the company launched in the US in April and already counts hundreds of venues there. It is also active in New Zealand, Singapore and the UK. Brand said the US is now its primary growth market, with a go-to-market team on the ground and his co-founder having relocated there.

Brand sees Australia as an ideal testbed for global hospitality technology, citing Melbourne's high density of restaurants and strong local appetite for dining out. He noted that major international technology companies also often pilot new features in Australia before global rollout.

He pointed to a growing cohort of Australian hospitality technology firms operating globally and argued that local conditions - including a sophisticated but challenging restaurant economy - give founders a strong base to build from.

"Yeah, we're all hoping, right, but I think we have, we have a great track record of great hospitality products. There's something about the ecosystem here. Australian love hospitality. Melbourne is one of the biggest cities, if not the biggest, I think, in restaurant per capita, yeah, in the world, is that why I live there?" said Brand.