Restaurant operators rarely have two days that look alike. New challenges emerge constantly, and keeping up means making dozens of decisions with limited time and incomplete information: Why did food costs spike at one location? Which limited-time offer actually drove profits? Why is one store consistently outperforming another?
Finding those answers often means digging through multiple reports, exporting spreadsheets and piecing together data from different systems, a process that takes time and distracts restaurant operators from their key hospitality role.
Surfacing data and pattern recognition are where AI excels. However, when restaurants use a generic AI model, it can’t dive into what your restaurant needs to know about its own operation, explains Matt Sundberg, SVP, Product at Craftable. That’s the impetus behind “Crafti,” the AI model from Craftable, a software platform that replaces disconnected spreadsheets and manual processes with automated, real-time visibility into food costs, vendor spend and operational margins.
“Crafti’s conversational layer allows general managers to ask anything and receive a grounded answer, along with the path forward, without having to wait on a colleague or a support ticket,” Sundberg says.
From Reports to Conversations
Through conversational AI, GMs can interact with operational data naturally, uncovering insights that help them make better decisions across every aspect of the business.
“Restaurant operators are surprised to find that Crafti doesn’t just give an answer, but actually drafts solutions,” Sundberg says. For example, if a GM asks it to describe next week’s order, Crafti will draft a purchase order against the actual vendor catalog, pricing and pack sizes. Or, after providing a cocktail name, photo or PDF, it will draft a costed recipe matched to actual inventory.
Wondering how Crafti can level up your own operations? Here are some examples of actual questions operators have asked to spark ideas about how it can uncover insights.
Cost and recipe diagnostics
- Why is our chocolate cake costing twice as much at this location compared with others?
- Which specialty cocktail delivers the highest margins?
- Which limited-time offers generated the most profit over the past 12 months?
Cross-store analytics
- Where did we have our highest sales week in 2025 compared to 2024?
- Comparing the last 30 days across all my stores, which locations are lagging?
- Rank stores by their sales of LTOs.
Operational insight
- Where was our biggest source of waste this audit period?
- How much Patrón Reposado have we sold since April?
- What accounts for the differences in server performance (i.e. upsells, more table volume, etc.)
Menu engineering
- Which menu items have the highest sales volume, but the lowest profit margins?
- Where can we raise prices that is least likely to hurt sales?
- Which menu items have seen the biggest decline in sales over the past six months?
Labor and scheduling
- Which shifts have the highest labor costs as a percent of sales?
- How would adjusting staffing levels on specific shifts have impacted labor costs last month?
- What days and dayparts have the greatest opportunity to reduce labor costs without impacting service?
Why Crafti is Craftier than a Regular LLM
While tech-savvy restaurant managers might have already tried general models, the results were likely lackluster, since they don’t know an individual business. Crafti is built differently from a general chatbot as a purpose-built assistant—a frontier model wrapped in Craftable-specific tools, product knowledge and guardrails.
“Because it runs on your actual operational data—POS, labor, invoices, inventory—it’s not reasoning about hospitality in the abstract; it’s answering from your vendors, your recipes and your variance,” says Sundberg.
Trust is Baked into Every Answer
AI is only valuable if operators trust the answers, and Crafti shows the work behind every response.
That means when Crafti runs an investigation, it queries the data directly and returns the underlying numbers and the math it used to get there. So rather than seeing a vague conclusion like “your food cost looks high,” a GM gets the actual figures the conclusion rests on, such as “theoretical is 28.4%, actuals are running 31.9%, a 3.5-point gap.”
That transparency is reinforced through the platform’s architecture, with data access confined to be “read-only” by default, so data isn’t altered by an investigation. And because access is isolated per store at the database layer, Crafti only ever sees what the signed-in user is permitted to see, such as only one location and not all.
On the two actions Crafti can take—drafting a recipe or a purchase order—every draft is previewed for the GM to review and approve, and every action lands in an audit log. “The point isn’t to expect you to trust the AI, but to give you a number you can check and a trail you can follow,” Sundberg says.
And, he notes, this is just the beginning, with Crafti getting sharper as the overall platform does. “More agentic actions, broader module coverage and tighter pairing with Craftable’s other features are all on the path forward.”
Are you ready to spend less time searching for answers and more time making better decisions? Find out more about Craftable here.
