I co-founded Passionfruit because I watched quality and compliance teams in food and beverage spend the majority of their time on work that should not require their expertise. Answering the same customer questionnaire for the fifteenth time. Searching for a certification that lives in someone's sent folder. Rebuilding an answer that was already approved six months ago.

That problem has not changed. What has changed is the maturity of the tools available to solve it, and the speed at which food companies are adopting them.

What has actually changed since 2025

A year ago, most conversations about AI agents in food quality were still largely theoretical. The vision was clear. The implementations were early.

That has shifted. According to Deloitte, 70% of food manufacturers increased investment in compliance technology in 2024. Gartner projects that by 2028, one third of enterprise software applications will include agentic AI, up from less than 1% in 2024. The AI in food safety and quality control market was valued at $2.7 billion in 2024 and is projected to reach $13.7 billion by 2029, growing at 30.9% annually.

These are not projections about potential. They reflect decisions already being made.

What I notice at Passionfruit is that the conversation with quality teams has shifted too. A year ago, the question was often "is AI ready for this?" Today, it is more often "how do we implement this without disrupting what is already working?"

That is a more useful question, and it is the one I want to address here.

What are AI agents?

The term gets used loosely so it is worth being precise. An AI agent is not a search tool or a chatbot. It is a system that can take a goal, break it into steps, execute those steps using available tools and data, and deliver an output, without a human managing each step in between.

The distinction that matters for quality teams in food is the difference between a system that answers questions and a system that completes work. A chatbot tells you where a document is. An agent finds the document, checks whether it is current, maps it to the relevant question in a customer questionnaire, and drafts the response. The human reviews and approves. The agent does the retrieval and assembly.

That is a meaningful difference in a workflow where most of the time is currently spent on exactly that retrieval and assembly work.

What this means for quality teams specifically

The honest answer is that it changes what the job looks like more than it changes whether the job exists.

Quality Managers in food are not primarily valuable because they can find a certification document or copy an answer from last year's submission. They are valuable because they understand what the answer should say, whether it reflects current practice, and what the consequences are if it does not. That judgment does not go away. What AI removes is the hours spent before that judgment is even needed.

The teams I see getting the most from AI agents are not the ones trying to automate their way out of compliance work. They are the ones who have freed up enough time to actually improve their systems, close gaps before audits surface them, and respond to customers faster than their competitors.

If you are a Quality Manager thinking about what that infrastructure looks like in practice, this article on building a supplier response system covers exactly that.

On data privacy and responsible AI

This comes up in almost every conversation we have with quality teams, and it should.

The concern is legitimate. Questionnaire responses contain commercially sensitive information. Policies, supplier relationships, process details. The wrong AI implementation creates exposure rather than reducing it.

At Passionfruit, our infrastructure is hosted in Europe on AWS Frankfurt, GDPR-compliant, and zero data sharing beyond the client environment. Every answer the system produces includes a confidence score and a source reference so the team can see exactly where the output came from and verify it before anything goes out. We call this responsible AI: the system supports the decision, the qualified person makes it.

The question to ask any AI tool in this space is not just what it can do but where your data goes and what happens when the output is wrong. Those are not edge cases. They are the baseline requirements for using AI in a regulated industry.

What AI agents are actually doing for quality teams today

An AI agent is not a chatbot. It does not wait for a question. It executes a workflow: takes in information, makes decisions based on rules and context, and produces an output, often without requiring a human at every step.

For quality and compliance teams in food, the workflows where this is genuinely useful right now are specific.

Answering customer questionnaires. A customer sends a 60-question qualification questionnaire. An AI agent scans your existing documentation, maps each question to a relevant approved answer or source document, and produces a draft response. Your team reviews, adjusts where needed, and approves. What used to take two days takes two hours. The agent does not decide what is true. It finds what you have already decided is true and organises it.

Managing certification and document expiry. Most quality teams track certification expiry dates in a spreadsheet that someone remembers to check before a questionnaire arrives, if they remember. An AI agent monitors expiry dates, flags upcoming renewals with enough lead time to act, and in some cases initiates the update workflow automatically. This is not complex AI. It is reliable process automation that removes a category of avoidable risk.

Preparing for audits. Audit preparation is one of the most time-intensive tasks a quality team faces. An AI agent can compile the required documentation against a known standard like BRCGS or FSSC 22000, identify gaps between what exists and what the standard requires, and flag what needs attention before the auditor arrives. The FDA deployed agentic AI for exactly this kind of multi-step compliance workflow in December 2025, with more than 70% of staff using the tool voluntarily within months.

Maintaining answer consistency across customers. When the same underlying question is answered differently in two separate submissions, it creates problems that surface during audits or customer reviews. An AI agent working from a shared answer library ensures that the allergen management answer sent to retailer A in March is consistent with what was sent to retailer B in October, because both drew from the same approved source.

What AI agents are not doing

This matters as much as what they are doing.

AI agents are not making food safety decisions. The judgment about whether an allergen risk is acceptable, whether a supplier meets approval criteria, or whether a corrective action is sufficient still requires a qualified person. An agent can surface the relevant information and flag inconsistencies. The decision belongs to the team.

AI agents are not replacing institutional knowledge. They make it accessible. The knowledge of how to answer a specific customer's questionnaire, which document supports which claim, who approved what and when, that knowledge was always in the business. What AI does is give it structure and make it retrievable by anyone on the team, not just the person who built it.

And AI agents are only as good as the data they work from. An agent that draws on outdated policies, lapsed certifications, or unreviewed answers produces faster wrong answers, not better ones. The foundation of any AI-supported quality workflow is still a well-maintained answer library and a document layer that reflects current practice.

What I think the next twelve months look like

The category is forming quickly. Food companies that have been cautious about AI adoption are moving faster now, partly because the tools are more mature, partly because the regulatory environment is creating more demand for structured, traceable compliance data.

CSRD, EUDR, and increasingly detailed retailer requirements mean the volume of compliance data that food suppliers need to manage and share is growing. Manual processes that worked at lower volumes are creating bottlenecks at higher ones. That pressure is driving adoption faster than I expected a year ago.

What I am watching is whether the adoption stays at the document automation layer, faster questionnaire responses, better document management, or whether it moves into more proactive risk management. The predictive analytics use case, where AI identifies quality risks before they become non-conformances, is real but still early in food. I think it gets more traction in 2026 than most teams expect.

At Passionfruit, we are building for the teams doing this work today. The quality managers handling 200 questionnaires a year. The ESG leads trying to coordinate a coherent EcoVadis submission across four departments. The technical teams managing specifications across fifteen retail customers. That is where the daily friction is, and that is where AI agents deliver the clearest value right now.

The vision in my April 2025 post has not changed. The timeline has compressed.

Book a demo here to see how Passionfruit uses AI to handle the questionnaire and compliance workflow for food and beverage teams.‍