Ulta’s $200M bet: Why supply chain pain should shape your roadmap
In today's issue:
Ulta Beauty’s $200M+ supply chain overhaul
Battle-tested frameworks from Kroger's enterprise AI transformation
How Jack in the Box is scaling AI from pilots to enterprise transformation
1. Ulta Beauty’s $200M+ supply chain overhaul
NASCES 2025 Fireside Chat, Session: Transform Your Supply Chain to Accelerate Growth: Ulta’s Bold Overhaul (Nov. 10, 2025)
For: Retail transformation leaders, supply chain executives, CDOs implementing automation, omnichannel operators, specialty retail VPs, warehouse operations leaders
Background: Ulta Beauty went global, opening stores in Mexico, acquiring SpaceNK in the UK and Ireland, and expanding into the Middle East. But none of that would be possible without the massive supply chain transformation they started in 2020. VP of Supply Chain Strategy and Systems Heather Lane led the overhaul of all 7 US distribution centers, implementing autonomous mobile robots and automated storage systems while reducing training time from 10 hours to 45 minutes. Here's the framework Ulta used to transform their supply chain without losing their team.
TLDR:
Start with associate pain points, not technology: Ulta's Chief Supply Chain Officer visited all 7 DCs and heard consistent complaints about heavy push carts and manual processes, which became the blueprint for automation investments.
Build for 30 years, not for today: Ulta's decision to invest in-house rather than use 3PLs was driven by understanding their 5-7 year enterprise strategy and knowing they needed owned capacity for international expansion.
The framework many retailers miss:
Start with people pain, not technology specs. Ulta’s Chief Supply Chain Officer didn't start with vendor pitches or automation ROI models. She went to all seven distribution centers and listened. The feedback was consistent: associates were pushing heavy carts, processes were too manual, tasks were physically demanding.
Most retailers start transformation with a technology mandate from executives. Ulta started with frontline complaints and built the business case around solving those specific problems. When you anchor transformation to associate pain points, you get automatic buy-in because the team sees you're solving their problems, not imposing corporate initiatives.
Make build-vs-buy decisions based on your 30-year vision, not 3-year ROI. Ulta faced the classic question: should we build our eighth distribution center or use a 3PL to scale?
But the calculation isn't just about cost per unit. It's about strategic flexibility. Ulta knew they were going international. They knew they were expanding into wellness categories. They needed owned capacity that could adapt to future needs, not just handle today's volume.
Most retailers optimize for short-term cost efficiency. Ulta optimized for long-term strategic control. They could capitalize investments in their own network, build the exact automation they needed, and maintain the culture of continuous improvement.
The key insight: when you're this close to your enterprise strategy and know what's coming 5-7 years out, you can make different infrastructure decisions than competitors focused on quarterly metrics.
Use formal qualifications to eliminate training excuses. One of Ulta's biggest change management innovations was implementing formal qualification processes with signed documentation.
Before anyone moves into a new role post-automation, they complete training modules on iPads on the warehouse floor. Then they work alongside a master (someone with total mastery of the role) who must also sign off. Both the associate and the master sign a qualification document stating the person is ready.
"When they sign off on that, you cannot use the excuse you were not trained."
Lane calls these "be-able-to's": specific capabilities you must demonstrate to succeed with the new technology. The qualification isn't designed for people to fail; it's designed to ensure success. If something's missing during qualification, you get more training.
This eliminated the most common transformation complaint: "The system went live and people said you didn't train us." With signed qualifications, that excuse is off the table.
What to do about this:
→ Run a "pain point listening tour" before selecting automation vendors. Schedule visits to your distribution centers or stores. Ask frontline teams: What's physically hard? What's frustrating? What takes too long? Build your automation business case around solving these specific problems, not generic efficiency gains. This creates instant buy-in and gives you concrete metrics to show ROI.
→ Map your 5-7 year enterprise strategy before making build-vs-buy decisions. Convene a session with your enterprise strategy team and supply chain leadership. What categories are you expanding into? What geographies? What volume growth? If your growth requires strategic flexibility and control, in-house investment may beat 3PL economics even if unit costs look higher today.
→ Implement formal qualification with signed "be-able-to" documents. Create a list of specific capabilities required for each role post-transformation (e.g., "able to operate robot interface,"). Build training modules that directly address each capability. Require both the trainee and a master to sign off that qualification is complete before moving into production roles.
2. Battle-tested frameworks from Kroger's enterprise AI transformation
Zoolatech webinars, Session: Unlocking Big Retail Revenue with AI (Nov. 13, 2025)
For: Chief Digital Officers, CDTOs, VPs of Data & Analytics, retail technology strategists, enterprise transformation leaders, heads of AI/ML
Background: Todd James spent years as Chief Data and Technology Officer at 84.51°, Kroger's retail media and insights subsidiary, where he was accountable for enterprise-wide AI strategy. During his tenure, he achieved a 14% reduction in technical team growth projections through generative AI tools while maintaining code quality. Now, as Founder and CEO of Aurora Insights, he helps enterprise boards and executive teams scale AI.
TLDR:
AI success requires operational discipline, not just data and scientists. Having great data and data scientists means nothing if your organization can't actually implement and act on AI recommendations faster than competitors.
Go after quick wins first, then build differentiation. Use off-the-shelf AI tools immediately for operations efficiency and customer personalization, then invest in proprietary AI where you have unique data assets and competitive moats.
The framework many retailers miss:
Operationalization beats data and science every time. James identifies 3 factors that determine AI success:
unique data assets,
access to data scientists, and
operational discipline.
Most retailers obsess over the first two and ignore the third.
"If I have the data and the science, do I have the operational discipline and capability within my firm to act upon it more than my competitors?"
Here's why this matters: you can have the world's best pricing algorithm, but if your merchandising team takes two weeks to implement changes while your competitor does it in two hours, you lose. You can have perfect personalization models, but if your marketing team can't activate them across channels, the science is worthless.
This is why he tells retailers to look at their current operational cadence first. Are you good at testing and learning? Do you have clear decision rights? Can you move fast on data-driven recommendations? If the answer is no, AI won't fix that – it will expose it.
Go after quick wins with off-the-shelf tools, then build proprietary moats. Retailers face a strategic tension: should we use off-the-shelf AI tools or build proprietary capabilities?
James's answer: do both, but sequence them correctly.
"Win the first race, get your data positioned, get your organization ready, and start using off-the-shelf tools. The sooner you do that, the sooner you're going to see operational and economic benefits."
Start with areas where off-the-shelf AI delivers immediate value:
customer segmentation
marketing automation
operations efficiency
demand forecasting
These tools are increasingly accessible and don't require building models from scratch.
Where James invested in proprietary AI: areas where Kroger had unique data assets that competitors couldn't replicate.
Three years ago, data scientists were the scarcest resource. Now, James says data is the bigger differentiator. "You can increasingly find tools and other capabilities making AI more accessible. You can rent data scientists if you don't have them."
The strategic question: where do you have proprietary data that creates competitive advantage? Invest there. Everywhere else, use off-the-shelf solutions and focus on operational excellence.
Start using AI in marketing – it's the most fertile ground. When James advises retail CTOs and heads of data on where to start, he consistently recommends marketing.
"There's generally a lot of data there and a math-aware workforce that wants to get better."
Marketing teams already use data for decisions. They understand A/B testing. They measure conversion rates and CAC. They're mathematically literate and hungry for better targeting.
This is the opposite of starting with operations or supply chain, where you're often fighting cultural resistance to data-driven decision making. Marketing teams pull you in; operations teams need convincing.
Address the workforce story early, or it will kill your transformation. The biggest blocker to AI transformation isn't technology. It's fear.
"Never too early to have a story on what this means for your technical workforce and what it means for the business workforce."
James saw this play out repeatedly: technical teams worried AI would replace their jobs. Business teams worried they'd lose decision-making authority to algorithms. Both fears derailed projects when not addressed head-on.
His advice: get out in front with transparent communication about how AI augments rather than replaces people. For technical teams, show how AI tools make them more productive. For business teams, emphasize that AI provides better insights, but humans make final decisions.
What to do about this:
→ Audit your organization's operational discipline before scaling AI. Assess your current execution speed: How long does it take to go from insight to action? Can you run weekly experiments? Do teams act on data recommendations within 24 hours? If not, spend six months improving operational cadence before investing heavily in AI.
→ Start your AI program in marketing with personalization or campaign optimization. Identify 2-3 quick wins in marketing: customer segmentation, email personalization, digital ad targeting, or promotion optimization. Partner with a marketing VP who wants better results. Use this success story to build momentum for expansion into operations and merchandising.
→ Map your unique data assets and decide where to build vs. buy. Convene your data and strategy teams. Identify: What data do we have that competitors don't? Where does this data create pricing power, customer loyalty, or operational advantage? Invest in proprietary AI only where you have data moats. For everything else, use off-the-shelf solutions and focus on implementation speed.
3. How Jack in the Box is scaling AI from pilots to enterprise transformation
Reimagine with AI podcast by Sigmoid, Episode: AI in Fast Food: How Jack in the Box Is Transforming Personalization & Supply Chain (Nov. 12, 2025)
For: CDOs, CTOs, retail technology leaders, data and analytics executives, QSR operators, omnichannel strategists, digital transformation teams
Background: Paul Bruffett built the machine learning engine powering millions of Starbucks customer experiences daily, then led global retail AI at Accenture. Now as VP of Data & Analytics at Jack in the Box, he's wrestling with the same challenge every retail DX leader faces: digital-first experiences generate massive data volumes, but technical debt moves faster than your ability to extract insights. Here's his playbook for scaling AI from pilots to enterprise transformation while maintaining the human hospitality that drives loyalty.
TLDR:
Technical debt accumulates faster than AI innovation: digital experiences generate exponentially more customer data than legacy POS systems, creating an "appetite for insights" that outpaces infrastructure capacity without disciplined prioritization.
The next AI wave isn't automation, it's decision aids through reinforcement learning and simulation that help businesses explore trade-offs and quantify uncertainty rather than simply forecasting single outcomes.
The framework many retailers miss:
Build for scale by managing technical debt, not just adding features. Bruffett identifies the core challenge: "Digital-first experiences generate massive amounts of data." Retailers are used to handling transactions from POS systems. But digital experiences – first-party apps and third-party delivery – generate exponentially more customer information.
This isn't new for data teams, but it's "become turbocharged with both the variety and velocity of data." The framework: You can't build fast enough if you're carrying technical debt.
Most retailers think about AI as adding new capabilities. Bruffett thinks about it as managing the tension between innovation and infrastructure. Before you deploy another AI pilot, ask: Do we have the data foundation to scale this if it works? If not, you're creating future technical debt while your team is still paying down the last wave.
Use generative AI to synthesize siloed knowledge, not just automate tasks. Bruffett sees the biggest near-term opportunity in knowledge synthesis. Jack in the Box has "rich quantitative qualitative research – surveys about why people engage with the brand, how they engage. It's survey-based data, handcrafted, curated. But we also have digital signals from people transacting in our apps."
These two pools live in silos: "One are PDFs and PowerPoints locked away in the customer insights repository. The other is in our data warehouse." Generative AI's strength: "Contextualizing both of those. That knowledge synthesis, information retrieval… that's what they really excel at."
The strategic principle: Most companies treat AI as a replacement for human work. Bruffett treats it as a bridge between structured and unstructured knowledge.
Your biggest AI opportunity might not be automating existing processes—it's unlocking insights trapped in formats (research reports, customer service transcripts, competitive analyses) that traditional analytics can't process.
Define success metrics before launching pilots. "A lot of companies experiment with pilots, and I don't think they have a clear definition for success. If you launch an AI pilot tomorrow, how could you prove it actually worked? If you can't answer that, you're not ready to scale."
This separates AI pilots from AI transformation. Everyone can launch pilots. Almost no one defines clear success metrics upfront. The result: endless experimentation without learning or scaling.
Bruffett's test: Before approving any AI initiative, the business must articulate:
What metric will change?
By how much?
Within what timeframe?
What would cause us to kill this project?
If you can't answer these questions, you're experimenting for experimentation's sake, not building toward enterprise AI transformation.
Balance automation with human hospitality. On in-store personalization, Bruffett notes some competitors experimented with dynamic pricing, "which got a negative reaction." Starbucks Korea uses license plate scanners to automatically start your order when you drive up, but "a lot of countries wouldn't be very excited about that."
His framework: "The basics of the industry – quality, menu innovations, clean stores, timely food, you get what you ordered – have remained the same for a long time and probably continue to do so." Technology should "surprise and delight" but only after meeting basic expectations. "If you don't get the basics right, focusing on these super gimmicky things sometimes feels very pointless."
Most retailers deploy AI to create differentiation. Bruffett deploys AI to enhance fundamentals. "Meet the customer's expectations first. Then surprise and delight them. That's where technology comes in, but we need to make sure we have everything in order first."
What to do about this:
→ Audit your technical debt before launching new AI pilots. Create a two-column list: (1) AI capabilities your business wants, (2) Infrastructure gaps blocking scale. If column 2 is longer than column 1, pause new pilots and invest in data foundation (cloud migration, data quality, integration architecture).
→ Start with knowledge synthesis, not task automation. Identify three knowledge silos in your organization (customer research, competitive intelligence, operations documentation). Build a generative AI proof-of-concept that synthesizes insights across these sources. This delivers immediate value while building organizational comfort with AI.
→ Create a "pilot success framework" before approving any AI initiative. Require teams to define: (1) Target metric and baseline, (2) Success threshold (X% improvement), (3) Timeline (30/60/90 days), (4) Kill criteria (if we don't see Y by day Z, we stop). Make this a gate for all AI spending.
→ Map customer journey fundamentals vs. differentiation opportunities. List your top 10 customer experience pain points. Mark which are "table stakes" (must work consistently) vs. "delight factors" (create differentiation). Deploy AI first on table stakes to eliminate friction, then on delight factors once fundamentals are solid.
Disclaimer
This newsletter is for informational purposes only and summarizes public sources and podcast discussions at a high level. It is not legal, financial, tax, security, or implementation advice, and it does not endorse any product, vendor, or approach. Retail environments, laws, and technologies change quickly; details may be incomplete or out of date. Always validate requirements, security, data protection, labor, and accessibility implications for your organization, and consult qualified advisors before making decisions or changes. All trademarks and brands are the property of their respective owners.
