Imagine waking up to find your AI agent has already ordered groceries, planned meals, and scheduled grocery deliveries based on your health, preferences, and local trends. It also books your travel, manages check-ups and medications, and even makes investments toward your financial goals. Agentic artificial intelligence systems (Agentic AI) are autonomous systems that require limited supervision and are guided by their defined goals. Agentic AI systems do more than just follow instructions; they make independent decisions aligned with their programmed goals, autonomously and often without oversight. This is not science fiction; businesses and individuals alike are increasingly adopting AI agents to carry out administrative tasks. These agents can create multimedia reports by coordinating research, writing, and design; manage supply chains by predicting demand and balancing supply; and optimise manufacturing from raw materials to final assembly.

The rise of agentic AI raises critical questions about the capacity of organisations to anticipate and manage risks from autonomous decisions. For instance, how can organisations properly assess and monitor agentic AI systems as complexity grows from linking a few models to interacting with multiple AI agents internally and externally? Who bears responsibility for evaluating risks, making go/no-go decisions, and intervening in real time when human oversight of system outputs is limited? How can staff be trained and supported to continually and responsibly manage, use, and respond to these highly complex AI systems? Beyond this, their growing influence prompts an even broader question: where might the expansion of agentic AI ultimately take us?

Using the cultural resonance of food and food systems, The Memory of Spice highlights the impact of agentic AI on food sovereignty, cultural expression and the flattening of difference. In the story, set in a near-future Bangalore, agentic AI runs much of economic, social, and cultural life, providing convenience to consumers and optimising their lives for sustained productivity. The agentic AI systems of 2030 are designed to optimise everything from restaurant supply chains to farming, gradually standardising what people eat. The drive for efficiency and speed in practice has led to greater standardisation, resulting in cultural homogenisation. Restaurant owners in Bangalore adopt IDLI-GPT, a system that optimises recipes for mass appeal and profit, to retain their commercial viability in an increasingly competitive market. IDLI-GPT helps them create mathematically perfect versions of dishes that taste good enough and cost less to make. For example, a biryani recipe that appeals to the majority becomes the template for every biryani served across the city.

While restaurants design many recipes for mass appeal even today, the difference with agentic AI systems driving food systems lies in its scale and the loss of cultural nuance. Dishes become iterations of iterations produced by machines with no sense of flavour, heritage, or local identity. As a result, restaurants no longer offer distinctive or memorable experiences; everything is standardised, serving primarily to provide basic sustenance and keep people productive, rather than celebrating culinary creativity or local culture.

The IDLI-GPT agentic AI system only sources the most profitable ingredients to maximise efficiency, creating massive demand for specific crops and making others commercially unviable for most farmers. Following the trajectory of today’s projections of agentic AI’s impact on potential job displacement and sectoral transformation in India, farmers are pushed out of production, with most food now lab-grown. However, some have chosen to continue cultivating their crops to maintain local practices, creating alternative paths within this highly standardised food system — a level of agency almost chipped away completely in urban Bengaluru. We are given a peek into their world as the protagonist, Anna, seeks these alternative farmers to obtain a “discontinued ingredient” for her grandmother’s original Sambhar-rice recipe.

In The Memory of Spice, the agentic AI systems' infrastructure, knowledge, and compliance requirements are programmed for a limited number of standard recipes. The entire system is premised on a prosaic understanding of food. Food carries more than nutrition; it carries memory. A dish holds centuries of adaptation: trade routes that brought spices, migration patterns that mixed cooking techniques, and local ingredients that shaped flavours over generations. When AI systems flatten this diversity into data points, they lose the context that makes food meaningful. In the Memory of Spice, this erasure isn't just personal — it’s cultural. Communities lose autonomy over their food traditions and are forced to consume them through corporatised platforms that decide which memories are worth preserving. Young people grow up never knowing what they're missing, while older generations find their knowledge suddenly irrelevant or dated.

Triggered by Anna’s traditional cooking, i.e. decision-making outside the grain, and unable to recognise her recipe and ingredients, the agentic AI system crashes and deletes the city's entire culinary database in a tragic glitch — a catastrophe not unlike what we’re seeing with these systems today. As a result, the hyper-efficient, standardised food system collapses.

We have become comfortable allowing algorithms to decide what we buy through optimisations and ads, and we risk treating the errors that follow as inevitable instead of working to anticipate and pre-empt them. The growing trust in these systems also means we’re not paying attention to the unintended consequences of these models failing or falling into the same trap humans do. For instance, companies could institute dark patterns that manipulate AI agents’ predictable behaviour, leading to serious repercussions for the humans that deploy them.

Because of how narrowly they’re programmed to execute specific tasks, agentic AI systems may not consider elements outside their goals viable for decision-making. The value of cultural heritage, local knowledge, and the satisfaction of creating something by trial and error, is dismissed despite holding significance to communities. Stories like ‘The River's Ledger’ demonstrate how systems that seem reliable can overlook crucial human context, erasing practices, relationships, and knowledge that aren’t easily quantifiable.

The Memory of Spice is not really about food; it's about what happens when we optimise away human agency and nuance in the name of efficiency. When we delegate decisions to AI systems, we forego the right to make independent choices and engage in the cognitive, social, and cultural processes that make those choices possible and worthwhile. The story ends with hope, but it's hard-won. It's a reminder that the future isn't inevitable; it will be the result of our choices today about how much agency we're willing to delegate and what we're determined to preserve. Ultimately, The Memory of Spice is a reminder to:

• Question what Agentic AI optimises for: Efficiency and profit maximisation are not neutral goals. They promote specific outcomes such as profits, while marginalising others, such as diversity. We must be explicit about whose values agentic AI systems are built to serve and which values should remain under human oversight. In addition to questioning what Agentic AI optimises, it is essential to note that not every decision needs to be optimised. Some areas of life benefit from inefficiency, variation, creativity and human judgment.
• Recognise that convenience-driven adoption at scale has hidden costs: The story shows how market pressures and overwork make agentic AI feel like a saviour. For instance, restaurant owners adopt it to survive thin margins, and busy consumers embrace it for convenience. But this adoption happens faster than we can understand its long-term impacts. While evaluating technology, we need better mechanisms to assess immediate benefits and the cultural and social costs that compound over the years.
• Technology could amplify cultural homgenisation: The AI in The Memory of Spice optimises for mass appeal and customer preferences. Agentic systems risk creating feedback loops where they optimise for artificially narrowed preferences, making diversity seem less desirable. Without deliberate correction, AI may entrench existing exclusionary practices and amplify them.