AI Grocery Fulfillment: Lessons from Riesbeck’s Move

3% is the number most grocery teams should have pinned above their desks right now.

That’s the often-quoted industry average growth rate for grocery e-commerce sales. If your online orders are growing slower than that, you’re losing share. If they’re growing faster, you’re probably feeling the strain—more substitutions, longer pick times, frustrated store teams, and customers who quietly stop ordering.

That tension is why Riesbeck’s Food Markets—an employee-owned, 17-store grocer across Ohio and West Virginia—just deployed eGrowcery for online shopping and fulfillment. The announcement reads like a standard platform upgrade. I think it’s more interesting than that: it’s a signal that regional grocers are done treating e-commerce as a side project and are starting to operationalize it like a core business.

This post is part of our “AI in Retail and E-Commerce” series. While the press release focuses on “speed” and “convenience,” the real story is how modern fulfillment platforms (often with AI under the hood) make omnichannel grocery profitable—or at least not painfully unprofitable.

Why grocery e-commerce wins or loses in the back room

Answer first: Grocery e-commerce success is mostly determined by fulfillment execution, not the website.

Yes, your digital storefront matters. But in grocery, the customer experience is shaped by what happens after the “Place order” button:

• Was the item in stock when the picker arrived?
• Did substitutes make sense?
• Did the order arrive at the right temperature?
• Did pickup happen in under five minutes?

If you miss on those, customers don’t complain loudly. They just switch back to in-store—or switch stores.

Riesbeck’s CEO, Brian Riesbeck, described the common pattern: their prior e-commerce system helped pre-pandemic, but shopper expectations changed quickly. That’s polite language for a harsher reality: pandemic-era e-commerce proved demand; post-pandemic e-commerce proved operations.

The cost problem nobody wants to say out loud

Answer first: Picking and packing can erase your margin unless you control labor and substitutions.

A typical in-store pick workflow introduces hidden costs:

• Pickers walking inefficient routes
• Re-picks when items are out of stock
• Manual exceptions (refunds, substitutions, customer calls)
• Congestion at staging and pickup

When platforms advertise “unified commerce,” they’re really promising something more practical: fewer exceptions, faster picks, and less time wasted.

What Riesbeck’s eGrowcery deployment suggests about “AI fulfillment”

Answer first: The most valuable AI in grocery fulfillment is the kind that quietly reduces decisions and exceptions.

The announcement highlights online shopping, pickup, and delivery, plus a stated goal to exceed the 3% growth benchmark. It also includes a vendor claim: retailers upgrading to eGrowcery’s unified commerce platform can see up to 23% sales increases “right out of the gate.” Treat vendor numbers carefully—but the direction is credible if a retailer fixes conversion killers like bad availability, clunky substitutions, and slow fulfillment.

Here’s where AI commonly shows up in platforms like this (even when it’s not spelled out explicitly):

1) Demand forecasting that feeds real inventory decisions

Answer first: Forecasting matters only if it changes replenishment and allocation.

AI demand forecasting can improve accuracy at the item-store-day level, especially when it accounts for:

• Seasonality (hello, late-December holiday shops)
• Local events and weather patterns
• Promotion lift and cannibalization
• Delivery-slot constraints

In grocery, better forecasting pays off quickly because in-stock rate is conversion rate. If a shopper builds a basket and hits three out-of-stocks, you don’t just lose those items—you risk losing the whole order.

2) Smarter substitutions that protect trust

Answer first: Substitutions are a loyalty moment, not a logistics detail.

Customers judge substitutes like a personal decision: “Did you get what I meant?”

AI-driven substitution logic can use signals like:

• Brand affinity and past purchases
• Dietary preferences (gluten-free, low-sodium)
• Price sensitivity (avoid trading up without permission)
• Basket context (taco night vs. school lunches)

A solid substitution experience reduces refunds, reduces customer service contacts, and increases repeat purchase. That’s a direct line to e-commerce growth.

3) Pick-path optimization that buys back labor hours

Answer first: The fastest way to improve fulfillment is to reduce picker walking time.

Even modest improvements in pick-path logic can produce noticeable gains:

• Faster picks per hour
• Less congestion in high-traffic aisles
• More predictable staging and handoff

This is one of those areas where “AI” can be overhyped. You don’t always need fancy models—sometimes you need clean store maps, smart batching, and practical rules. But when platforms combine historical order patterns with store layouts, the result is simple: more orders per labor hour.

4) Slotting and capacity decisions that keep promises

Answer first: If you can’t fulfill on time, don’t sell the slot.

Delivery and pickup slots are a contract. AI can help retailers avoid overcommitting by forecasting:

• Expected order volume per time window
• Average pick duration by basket size
• Staffing coverage and shrink-wrapped constraints

The best customer experience is boring: the slot is available, the order is ready, and nobody had to “check in” three times.

Unified commerce is really about one thing: fewer conflicting truths

Answer first: Unified commerce works when inventory, pricing, promos, and customer data match across channels.

Most companies get this wrong by focusing on the “front end” first. They polish the app, add a nicer search bar, maybe launch a loyalty revamp. Then customers order items that store teams can’t find, promos don’t apply correctly, and pickup associates spend their shift apologizing.

A unified commerce approach aims to align:

• Item availability (what customers see vs. what stores have)
• Pricing and promotions (no surprises at checkout)
• Customer profiles (preferences, order history, loyalty status)
• Operational workflows (picking, staging, substitutions, handoff)

For grocers, this alignment is the difference between “e-commerce growth” and “e-commerce chaos.”

What regional grocers (including in Ireland) can learn from this move

Answer first: You don’t need 500 stores to benefit from AI-enabled fulfillment—but you do need operational discipline.

Our “AI in Retail and E-Commerce” series often highlights personalization, pricing optimization, and customer behavior analysis. Grocery adds a twist: your personalization and pricing are only as good as your ability to fulfill the promise.

For retailers in Ireland—and really any market where labor is tight and customers are price-aware—Riesbeck’s approach maps to a practical playbook:

Start with the moments that create churn

Answer first: Fix the top three churn drivers before chasing fancy features.

In most grocery operations, those drivers are:

1. Out-of-stocks and weak substitutes
2. Missed pickup readiness times
3. Poor freshness/temperature control during staging

Measure these weekly. If your platform can’t expose them cleanly, that’s already a problem.

Treat store teams as product users

Answer first: If store teams hate the workflow, customers will feel it.

Riesbeck’s CEO called out “early feedback from our store teams already positive.” That detail matters. Store teams are the ones living with:

• Pick lists
• Exceptions
• Customer handoff

When they have better tools, they move faster and make fewer mistakes. When they don’t, they improvise—and improvisation doesn’t scale.

Make AI accountable to business metrics

Answer first: AI should be judged on labor, availability, and repeat rate—not demos.

If you’re evaluating an AI-enabled fulfillment platform, anchor on outcomes like:

• Pick rate (items per hour)
• Order accuracy (including correct substitutes)
• On-time pickup/delivery
• Refund and cancellation rate
• Repeat purchase rate (30/60/90 days)

If a vendor can’t explain how their system improves at least two of those, you’re buying vibes.

A practical 30-60-90 day rollout plan for AI fulfillment

Answer first: The fastest wins come from reducing exceptions, then optimizing labor, then personalizing.

Here’s a rollout sequence I’ve found works better than the usual “launch everything at once” approach.

First 30 days: stabilize the basics

• Clean up item data (units, weights, images, substitutions)
• Set substitution rules customers can control
• Standardize staging zones and cold-chain handling
• Track daily exception reasons (out-of-stock, damaged, not found)

Goal: fewer surprises.

Next 60 days: improve throughput

• Implement batching and pick-path improvements
• Align staffing plans to order patterns by daypart
• Add capacity controls to prevent oversold time slots

Goal: more orders per labor hour.

By 90 days: use customer behavior signals

• Personalize substitutes and recommendations based on purchase history
• Segment customers by sensitivity (price vs. premium vs. convenience)
• Use insights to plan promos that don’t break fulfillment capacity

Goal: higher repeat purchase and bigger baskets.

What I’d watch next at Riesbeck’s

Answer first: The real test is whether growth comes with better unit economics.

It’s easy to grow e-commerce by discounting fees, offering aggressive promos, or expanding delivery zones. The harder—and more valuable—win is growth with controlled cost-to-serve.

If the eGrowcery deployment delivers on speed and convenience, a few leading indicators should show up quickly:

• Higher conversion due to better in-stock visibility
• Lower refund rates due to smarter substitutions
• Faster pickup handoffs because orders are staged correctly
• Improved associate satisfaction because workflows are simpler

If you’re planning your 2026 roadmap right now, this is the mindset shift to make: AI in grocery is less about flashy features and more about running a tighter operation across every channel.

You can’t “market” your way out of late orders and bad substitutes.

Ready to pressure-test your fulfillment stack?

If you’re evaluating an e-commerce and fulfillment platform—or trying to get more from the one you already have—start by mapping where exceptions are created (and who has to fix them). That single exercise usually reveals why growth stalls.

In the next post in our AI in Retail and E-Commerce series, we’ll dig into how customer behavior analysis and pricing optimization should adapt when fulfillment capacity is the limiting factor.

What’s the one part of your online grocery experience that still feels manual or chaotic—and what would it be worth to make it boring?