Hobnob case study: AI recipe margin agent flagged 35 menu cost exceptions

What did the OPAG recipe margin agent do for Hobnob?

Restaurant margin moves quickly because ingredient cost, supplier availability, portion discipline, kitchen prep, delivery demand, promotions, waste, and staff execution can change faster than a monthly report can explain.

OPAG narrowed the workflow to one agent capability: recipe margin variance review before managers change recipes, menu prices, portions, supplier choices, prep plans, or guest-facing offers. The agent prepared 35 review packets so Hobnob teams could see which variance was supplier-driven, which was waste-driven, which needed kitchen validation, and which required finance or operations approval.

The answer-first summary is this: OPAG used governed AI to make restaurant margin review faster, source-linked, and auditable while keeping recipe, supplier, price, portion, and promotion decisions with accountable managers.

Why does recipe margin AI matter for restaurant groups?

Hobnob operates in a restaurant environment where the same menu item can be affected by branch demand, ingredient receiving, supplier price movement, kitchen yield, wastage, promotion pressure, and customer expectations.

The agent helped reviewers separate true food-cost pressure from avoidable waste, stale recipe cards, unapproved substitutions, supplier invoice variance, branch-level portion drift, and promotion leakage.

• Restaurant managers needed food-cost, demand, prep, and wastage evidence before changing a menu item.
• Kitchen leads needed recipe-card, portion, yield, substitution, and waste context before adjusting prep.
• Procurement owners needed supplier invoice, contract price, availability, and substitution evidence before switching vendors.
• Finance teams needed margin impact, promotion exposure, branch variance, and approval thresholds in one packet.
• Operations leaders needed repeatable source evidence before approving price, recipe, supplier, or portion changes across branches.

How did the agent flag 35 menu cost exceptions?

The workflow started with approved sources and role-based access. Branch teams saw their own variance packets, kitchen leads saw recipe and prep context, procurement saw supplier and purchase records, and finance saw margin and approval details.

Each packet included the menu item, branch, recipe baseline, supplier cost movement, POS demand shift, prep variance, waste signal, substitution note, margin impact, recommended reviewer, approval requirement, uncertainty note, and audit history.

• Scan: review POS demand, recipe cards, supplier invoices, purchase records, inventory movements, prep sheets, waste logs, promotions, and approval history.
• Score: rank exceptions by margin impact, waste trend, supplier price variance, branch concentration, substitution risk, guest impact, and urgency.
• Draft: prepare a source-linked packet with evidence, missing records, root-cause hypothesis, owner queue, and next accountable reviewer.
• Route: send prep/yield issues to kitchen leads, invoice gaps to procurement, pricing exposure to finance, branch drift to operations, and high-impact changes to management.
• Audit: record source retrieval, recommendation, reviewer edit, approval, rejection, escalation, guest-impact note, and override reason.

What governance kept menu margin decisions under control?

Restaurant AI should not silently change a menu price, reduce a portion, substitute an ingredient, switch a supplier, or publish a promotion. Those actions affect guest experience, food quality, purchasing commitments, revenue, and brand trust.

OPAG separated evidence preparation from decision authority. The agent could flag variance, explain likely causes, draft an internal review note, and route the packet, but managers approved recipe, price, supplier, portion, prep, promotion, and guest-facing decisions.

• Role-based access separated branch, kitchen, procurement, finance, operations, and management context.
• Source evidence showed why each exception was supplier-driven, prep-driven, waste-driven, demand-driven, promotion-sensitive, or approval-sensitive.
• Approval gates protected menu prices, recipes, portion rules, supplier substitutions, promotion changes, and customer-impacting commitments.
• Override logs captured when a manager accepted a lower margin for guest experience, availability, contract, or brand reasons.
• Audit logs supported branch coaching, supplier review, finance reporting, manager accountability, and model-quality monitoring.

Which OPAG services connect to recipe margin AI?

The recipe margin agent shows how OPAG connects restaurant evidence to accountable action. Predictive AI ranks variance risk, Conversational AI can answer source-linked questions about an item or branch, and Agentic AI routes each packet through the right approval path.

The same pattern can support restaurant chains, cafes, bakeries, cloud kitchens, catering operations, hotel food and beverage teams, and food manufacturing groups where recipe cost, supplier volatility, and waste affect margin.

• Predictive AI ranks cost variance, waste, supplier, branch, and demand signals.
• Conversational AI answers source-linked questions such as why one branch has higher food cost for a menu item.
• Agentic AI routes owner queues, approval gates, reminders, overrides, and audit logs.
• Governed workflow automation keeps operational recommendations connected to source evidence and manager authority.
• Supplier quality recovery AI helps recover value when supplier quality or delivery failures create food-cost leakage.

What can another restaurant group copy from this case study?

The practical pattern is to start with a visible margin queue rather than a broad restaurant AI program. Pick the highest-value menu category, define the recipes and source systems, review historical variance, then route new packets to the people who already own decisions.

OPAG usually recommends a read-only first release. Once managers trust the evidence, the workflow can expand into prep planning, supplier recovery, AP exception review, promotion governance, labor coverage, and executive reporting.

• Choose one menu category, branch group, or supplier-sensitive item set with measurable food-cost leakage.
• Connect POS, recipes, supplier invoices, inventory, prep sheets, waste logs, and approval policies before adding actions.
• Define protected actions: menu prices, recipes, portions, suppliers, promotions, and guest-facing messages require human approval.
• Measure margin variance, waste, review time, repeated root causes, override rate, and accepted recommendations.
• Expand only after managers can verify source links and trust the owner routing.