Marc Lore has long built his reputation on letting numbers drive decisions, and his newest venture Wonder takes that philosophy to the next level by entrusting an artificial intelligence system with promotion recommendations. In a recent conversation with Fortune’s Allie Garfinkle, Lore explained that the food‑tech platform aggregates input from at least a dozen colleagues who rate each employee on performance, behaviors, and leadership every six months. Those quantitative scores, together with qualitative written feedback, are funneled into an AI model that produces a consolidated performance portrait. The approach is designed to move away from gut‑feel judgments and toward a repeatable, data‑backed process that can be scrutinized and refined over time.
The peer‑review component is deliberately broad to capture a well‑rounded view of an individual’s impact. Rather than relying solely on a manager’s perspective, Wonder solicits assessments from a diverse set of coworkers who interact with the employee in different contexts. This multiplicity of inputs helps surface blind spots that a single evaluator might miss. After gathering the ratings, the system also incorporates narrative comments, allowing nuances such as mentorship style or collaborative spirit to be quantified through natural‑language processing techniques that translate textual insights into comparable scores.
Beyond the peer‑derived performance metric, Wonder adds a proprietary measure called “value above replacement” (VAR). VAR estimates how much harder it would be to fill a given role with someone of equivalent seniority and skill set, essentially quantifying the irreplaceability of an employee. By combining the VAR figure with the performance score, the AI calculates a promotion likelihood that Lore describes as “very objective.” The dual‑metric approach attempts to balance short‑term results with longer‑term strategic value, ensuring that high‑impact contributors who may not yet have flashy metrics are still recognized.
The algorithm does not stop at a yes/no promotion call; it also recommends an optimal tenure in the current position before moving up. This timing guidance aims to prevent premature advancement that could leave skill gaps or overwhelm newly promoted staff. While human managers retain the ability to override the model’s suggestion—provided they can present a compelling argument that the algorithm overlooked a critical factor—Lore notes that such overrides are becoming rarer. Each disagreement feeds back into the training data, allowing the model to learn from managerial expertise and gradually reduce the need for manual intervention.
Lore’s reliance on quantitative methods is not a new experiment; it is a thread that runs through his entire entrepreneurial journey. His uncle Joe recalled that even as a teenager Marc would arbitrage horse‑racing bets, spreading small wagers across multiple horses to improve his odds rather than placing all his hope on a single favorite. That mindset of playing the probabilities has persisted, influencing how Lore evaluates everything from product pricing to talent decisions. At Wonder, the same principle applies: decisions are reduced to percentages, and the goal is always to tilt the odds in the organization’s favor.
To make hierarchy instantly readable, Wonder adopts a visual ranking system inspired by taekwondo belts. Positions are color‑coded from white and yellow for entry‑level roles through brown and black for senior leadership. A quick glance at the organizational chart reveals where the company is concentrating its most experienced talent and how many junior staff report to each leader. This transparency not only aids Lore in spotting imbalances but also gives employees a clear visual map of potential career pathways within the firm.
Compensation openness complements the visual hierarchy. Wonder maintains a transparent pay structure that lets any team member see what others at the same level or in comparable roles earn. By removing secrecy around salaries, the company aims to curb perceptions of unfairness and to encourage constructive conversations about market value, skill development, and performance expectations. The openness also serves as a deterrent against discriminatory pay practices, because discrepancies become visible and can be addressed promptly.
Wonder’s growth trajectory has been nothing short of meteoric, providing a robust testing ground for its AI‑driven HR tools. In the summer of 2026 the startup secured more than $650 million in fresh capital, pushing its post‑money valuation to roughly $9 billion. Cumulative funding since its 2018 founding now approaches $3 billion. The firm operates 135 food halls spread across ten East Coast states, and Lore told Fortune exclusively that Wonder is “ready and prepared to go public early next year.” This scale means that any HR innovation must be able to handle thousands of employees while maintaining consistency and fairness.
Automation extends beyond the front office into Wonder’s kitchens, where the company has deployed an automated bowl‑making system capable of producing up to 500 bowls per hour. By contrast, a human worker typically manages no more than 45 bowls in the same timeframe. The robotic line not only boosts throughput but also reduces variability in portion size and preparation time, contributing to a more predictable customer experience. Lore highlighted this example at the Fortune Brainstorm Tech conference in June as evidence that Wonder’s commitment to automation permeates every layer of the operation.
The intersection of AI‑guided promotions and large‑scale automation raises important questions for the broader HR technology market. On one hand, data‑driven decision‑making can help mitigate unconscious bias that often creeps into managerial assessments, potentially improving outcomes for women and under‑represented minorities. On the other hand, reliance on historical data risks encoding existing inequalities if the training set reflects past discriminatory patterns. Companies must therefore invest in continual bias audits, diversify their data sources, and maintain human oversight to ensure that algorithmic recommendations serve as a tool for equity rather than a reinforcement of the status quo.
For leaders considering similar AI‑enabled promotion frameworks, several practical insights emerge from Wonder’s experience. First, the quality and breadth of input data are paramount; peer reviews must be structured, anonymous, and sufficiently numerous to yield reliable signals. Second, a hybrid governance model—where algorithms provide recommendations but humans retain veto power—helps capture edge cases while still benefiting from statistical consistency. Third, transparent communication about how scores are calculated builds trust and reduces fear of opaque “black‑box” decisions. Finally, integrating the promotion system with learning and development platforms ensures that employees receive actionable feedback on how to improve their VAR and performance scores.
To harness the promise of AI in talent management while avoiding its pitfalls, organizations should adopt a staged, evidence‑based approach. Begin with a pilot program in a single department or geography, collecting both quantitative metrics and qualitative sentiment from participants. Conduct regular fairness audits that compare promotion rates across demographic groups and adjust model weights if disparate impacts emerge. Invest in training managers to interpret AI outputs and to construct constructive override arguments that are grounded in observable criteria. Finally, treat the algorithm as a living asset: continuously feed it new outcome data, retrain it periodically, and keep a transparent changelog so that everyone understands how the system evolves.