By Tom Whitfield

A year from now, the produce industry will no longer debate whether to adopt artificial intelligence; it will be measuring how AI already changed its operations, Bennett Innovation Lab co-director Elliot Grant told attendees at the 2026 Global Produce and Floral Show in Orlando, Florida, on Thursday.

Where does produce sit on AI today?

Grant framed the sector as behind the curve but not disengaged. "I think it's fair to say not everyone in the produce industry is integrated with AI yet," Grant said. "They're not using it yet, and I don't think that's a problem." He projects "a very steady but substantial adoption of AI" by the time the same audience reconvenes next year.

The industry is already running live deployments in three operational areas:

• Visual inspection services
• Quality control
• Yield forecasting

Grant said adoption will accelerate fastest where monitoring replaces human walk-throughs. Today's grower may scout a field once or twice a week; continuous monitoring — soil moisture sensors, satellite imagery (high-resolution orbital imagery of crop acreage), drone flights and in-field robots — turns scouting into a 24-hour operation personalized to each operation. The agronomist relationship shifts from one or two in-person visits a year to constant digital contact.

What does the three-stage adoption model look like?

Grant broke AI rollout into three stages, each with different margin and labor implications for distributors, packers and growers:

• Augmentation: AI supports human tasks such as drafting emails or analyzing photos.
• Replacement: AI fully takes over defined workflows, including invoice management and supplier onboarding.
• Transformation: workflows themselves are redesigned around AI capability.

"These are paradigm shifts in the way we do agriculture," Grant said. "We're not going to get there in one leap, but I really see there's a progression." The third stage, he argued, is "where all the power lies" — the layer at which gross-margin impact compounds rather than simple offsets.

Who inside the company actually drives the rollout?

Grant pushed back on the assumption that transformation must be a board-level initiative. A CEO who grants permission is necessary but insufficient. The daily users sit in middle and entry roles. "You give permission to people to try and experiment with small things, and it bubbles up from the bottom," he said.

That bottom-up pattern has direct implications for produce distributors managing high SKU counts (stock-keeping units, or distinct product lines) and tight inventory turns (the number of times per year that average stock sells through). Employees closest to shrink (unsold or spoiled product), quality claims and supplier paperwork are best positioned to identify what should be automated first.

Which jobs does AI threaten?

Grant does not expect AI to displace harvest labor or field judgment. He does expect pressure on white-collar tasks built around retrieval, summarization and repetition. The categories he flagged:

• Information lookup
• Document summarization
• Repetitive administrative tasks

He drew the line at work grounded in proprietary relationships, weather judgment, genetics and consumer behavior. "The produce business is based on proprietary data, proprietary relationships, dealing with things like the weather and genetics, the weirdness of consumer behavior and flavor," Grant said. "These things are not reducible to algorithms."

What does this change for produce distributors?

Two risk vectors stand out for wholesale and category buyers. First, guardrails: regulation has not kept pace with the technology, and Grant placed the danger in the absence of controls, not in AI itself. "AI can be safe if the industry chooses to use it safely," he said. AI agents (autonomous software programs that act on a user's behalf) carry extra exposure if granted root code access (unrestricted system-level permissions) or left to interact with customers without supervision.

Second, channel power: AI shopping agents may start rerouting consumer demand without notice. Grant singled out Muse, a multi-function AI agent that handles shopping and other tasks for individual users, as a near-term variable. Its recommendation engine could move produce volume overnight. "Probably this time next year there will be some folks saying, 'I have no idea why my sales of mushrooms suddenly shot up,'" Grant said, citing a hypothetical driven by an algorithm rather than merchandising. He urged the industry to become the authoritative source Muse and similar agents consult — pre-empting any mushroom-style spike, or dip, with clear, machine-readable guidance on nutrition and safety.

The produce sector's near-term test is whether it can install guardrails and authoritative data feeds faster than consumer-facing AI agents start setting purchase recommendations on the industry's behalf.