13 signals in the race to real-time retail: insights from theCUBE + NYSE Wired
Retail automation has entered its AI era. Every checkout is a live load test, and small delays turn into big revenue losses when millions hit "buy" at once.
During theCUBE + NYSE Wired: AI & Retail Trailblazers interview series, retail technology leaders unpacked what actually works at holiday scale. Here are 13 signals - and what General, Operations, IT, and Development teams can do next.
1) Data hygiene defines the wins
AWS posted roughly $33B in Q3 2025 revenue, up 20% year over year - a reminder that retail now runs on hyperscale infrastructure. But scale only pays off if your data foundation is clean and connected. As Justin Honaman of AWS put it, "AI is not the job of IT… AI is part of the business."
- Consolidate core entities (products, customers, locations, inventory) into a shared data model.
- Set SLAs for data freshness and lineage. No clean data, no reliable automation.
2) Agentic AI beats dashboard sprawl
Quantum Metric sees a shift from passive reporting to real-time intervention. If conversion drops, agentic systems don't wait for someone to read a chart - they investigate and propose fixes in the moment.
- Define top "break glass" events and let agents triage root causes.
- Alert on impact, not noise: revenue at risk, funnel breakpoints, time-to-fix.
3) Small wins supercharge margins
Zebra Technologies focuses on targeted workflows where a 2%-3% labor improvement matters - think shelf management. In grocery, those few points can flip the economics.
- Pick one high-frequency task and measure unit economics before and after.
- Automate the boring work first: counts, gaps, substitutions, pick-paths.
4) Move fast or fall behind
Many retailers stall in proof-of-concept purgatory. Celonis notes the blocker is usually execution clarity - leaders can't see how work actually moves across fragmented systems.
- Map real process flows (not the slideware version). Kill dead-end pilots.
- Prioritize by time-to-value and measurable P&L impact.
5) Orchestrate bits and atoms
GreyOrange connects planning and execution so a local signal triggers both in-store tasks and upstream replenishment. That's how you keep promises during peak traffic.
- Use one orchestration layer for people, robots, and sensors.
- Treat every signal as an action starter, not a report entry.
6) Agents are the new shelf space
Mirakl calls this Commerce 3.0: agent-driven buying becomes a sales channel of its own. It demands live pricing, availability, and assortment data wherever the shopper is.
- Expose clean product, price, and stock data to agent ecosystems.
- Prioritize interoperability over fancy features.
The National Retail Federation has been tracking these shifts in buyer expectations. See NRF for broader industry context.
7) Readiness gap: few are truly prepared
Dataiku sees a small group moving decisively, while most take incremental steps. The difference isn't tools - it's trust and process governance.
- Adopt an AI risk framework and enforce model lifecycle controls.
- Measure bias, drift, and decision quality like uptime and latency.
For reference frameworks, explore the NIST AI Risk Management Framework.
8) The funnel is conversational, not linear
DaVinci Commerce points out that shoppers want comparison and confidence, not a straight line from ad to cart. AI now handles browsing, questions, and shortlisting - then hands off to checkout.
- Build chat-first experiences that can reason over your catalog and policies.
- Instrument every step so you can see where confidence drops.
9) Trust wins the sale
Getbee shows how live video with real people beats generic chat for high-consideration buys. Those conversations also create qualitative data you can reuse for future guidance.
- Blend human and AI: let agents draft, let people reassure.
- Turn chat and video transcripts into reusable answers and playbooks.
10) Decision intelligence: from answers to advice
Aily Labs treats decisioning as a product: connect data, recommend actions, track profit impact. Inventory allocation and seasonal forecasting get better fast when advice loops back into learning.
- Ship decisions on mobile where managers live. Log "accept/override" to learn.
- Tie every recommendation to P&L fields: margin, sell-through, waste, stockouts.
11) Cameras are operational tools
Verkada turns in-store video into metrics: traffic, engagement, and alerts that trigger action. It's loss prevention, staffing, and conversion optimization rolled into workflows.
- Set clear policies for privacy, access, and retention.
- Integrate alerts with tasking systems so issues get resolved, not observed.
12) Personalization must be instant
Klaviyo notes the shift from curiosity to urgency. Customers expect brands to recognize them across channels and respond within minutes, not days.
- Unify customer profiles and events. Kill channel silos.
- Stand up always-on agents for service, win-back, and replenishment flows.
13) Discovery and conversion are merging
Similarweb observes that AI compressed the path from intent to checkout. In 2025, shoppers got comfortable using AI for research; in 2026, they complete purchases through those tools.
- Optimize for "buy in the moment" - fewer clicks, instant financing, live stock.
- Treat latency as a revenue leak. Measure it like you measure cart size.
What to do next
- Run a 30-day data audit: freshness, lineage, duplication, and access controls.
- Pilot one agent in production for a single workflow (e.g., out-of-stock triage). Set a clear success metric and a rollback plan.
- Implement an orchestration layer that assigns tasks to people, bots, and robots from the same event stream.
- Define SLOs for CX: page speed, search response, payment authorization, and post-purchase updates.
- Stand up AI governance: model inventory, risk scoring, monitoring, and human-in-the-loop checkpoints.
- Upskill cross-functional teams so Operations, IT, and Dev speak the same "agent + data" language.
If you're building team capability around agents, automation, or data-driven decisioning, explore practical upskilling paths here:
AI courses by job role and AI Automation Certification.
This series from theCUBE's livestream studio makes one thing obvious: retail's winners are merging clean data, agentic systems, and tight orchestration - then proving ROI in weeks, not quarters.
Your membership also unlocks:
