Retailers are investing in AI. Marketers have to make it matter.
AI is everywhere in retail, but adoption and impact aren't the same thing. A recent Berkeley Research Group report says more than eight in ten retailers have integrated AI to a moderate or large extent. The takeaway for marketers: the budget is there-proof of business value still isn't.
Where AI is used right now
- Marketing: 70%
- IT and digital functions: 62%
- Digital commerce: 56%
- Merchandising strategy and pricing: 54%
Where retailers plan to expand next
- Planning and product flow: 40%
- Corporate operations: 38%
- Supply chain and sourcing: 36%
- Distribution and logistics: 32%
Adoption ≠ outcomes
Tools like ChatGPT and Copilot can crank out product descriptions and ad copy. That saves time. But the report questions whether these wins are moving core metrics for retailers. Busywork is easy to automate; business results are not.
The marketer's scoreboard
Set AI projects against measurable outcomes. If you can't tie it to one of these, pause and reframe.
- Average order value (AOV)
- Conversion rate and checkout completion
- Revenue and contribution margin
- Customer retention, repeat purchase rate, and LTV
- Customer acquisition cost (CAC) and ROAS
- Inventory turnover and markdown rate (for promo and pricing work)
- Labor efficiency (content, campaign ops, analytics)
Practical playbook: from idea to impact
- Define the business problem in one sentence (e.g., "Lift email revenue per recipient by 12% in Q1").
- Map data and guardrails: sources, consent, brand voice, disclaimers, human review steps.
- Pick the smallest useful solution: prototype with off-the-shelf models before custom builds.
- Pilot with a clean control group and a clear primary metric. Keep pilots under six weeks.
- Calculate ROI (benefit - cost) and decision rules to scale, iterate, or kill.
- Document workflow changes and train teams; automation without adoption stalls.
Real examples to watch
- Sam's Club uses an AI-driven Scan & Go app to validate purchases and has a store in Grapevine, TX operating without traditional checkout lanes.
- Levi Strauss & Co. partnered with Microsoft on a complex agentic framework spanning IT, HR, and operations.
- Walmart launched an AI framework built on four "super agents": customer-facing Sparky, partner agent Marty for suppliers/sellers/advertisers, a store associate agent, and a developer agent-with more specialized subagents planned.
- Target is applying generative AI for merchant ideation via Target Trend Brain and to evaluate vendors for its Target Plus marketplace.
High-value marketing use cases in retail
- Content at scale: product descriptions, titles, alt text, email subject lines-with brand voice checks.
- Personalized offers: audience clustering and next-best action for on-site promos and CRM.
- On-site search and recommendations: intent parsing, semantic matching, cold-start solutions.
- Pricing and promo comms: auto-generate copy variants synced to price changes and inventory.
- Media optimization: creative iteration based on performance patterns, budget pacing prompts.
- Customer care assist: summarization, suggested replies, and escalation tagging to feed insights back into marketing.
Quick experiments you can ship in 30 days
- Subject line generator + bandit testing to lift open and click rates; measure incremental revenue per recipient.
- LLM-assisted PDP copy refresh on the bottom 20% of converting SKUs; track conversion and returns.
- Semantic search synonyms for top-miss queries; monitor zero-result rate and search-led revenue.
- Creative variant generation for paid social with automated UTM discipline; compare ROAS vs. human-only sets.
- RAG knowledge base for marketplace vendor FAQs; cut response time and measure case deflection.
Governance that keeps you out of trouble
- Bias and brand safety checks on generated content.
- Clear human-in-the-loop points for approvals.
- Data privacy and vendor contracts that limit model training on your data.
- Runbooks for model drift, outages, and rollback.
If you need a starting point, review the NIST AI Risk Management Framework and benchmark your current controls.
Budget asks that get approved
- Attach projects to revenue or cost lines the CFO already cares about.
- Show a tight payback window (e.g., three months) with specific KPIs and a time-boxed pilot.
- Present a rollout plan: data, people, and process changes-not just the tool.
The signal from the Berkeley Research Group report is clear: AI spend is high, but impact is uneven. Marketers who tie use cases to P&L metrics-and run disciplined pilots-will win budget and compounding results. Everyone else will add tools and see little change.
Resources
- Berkeley Research Group: thinkbrg.com
- Build your skills: AI Certification for Marketing Specialists and Top AI tools for copywriting
Your membership also unlocks:
