This article is sponsored by Unframe and was written, edited, and published in alignment with our Emerj sponsored content guidelines. Learn more about our thought leadership and content creation services on our Emerj Media Services page.

Retail is facing an AI deployment bottleneck. While organizations increasingly experiment with AI, relatively few have succeeded in scaling it across the enterprise.

Carnegie Mellon University’s Software Engineering Institute, working with Accenture, reports that only 8% of organizations have achieved enterprise-scale AI deployment maturity. A U.S. Census Bureau working paper similarly found that among firms using AI, 57% deploy it in three or fewer of the 15 business functions tracked by the survey, suggesting that AI adoption often remains limited in scope rather than embedded across enterprise operations.

In retail, the challenge is compounded by comparatively low adoption rates and fragmented operating environments. The U.S. Census Bureau’s Business Trends and Outlook Survey discovered that roughly 14% of retail-trade businesses reported current AI use in May 2026, below the 19.8% average across all businesses.

MIT Technology Review reports that leading retailers are increasingly shifting from standalone AI pilots toward integrated systems that support decisions across inventory management, fulfillment, customer engagement, and operational planning. For retailers operating across multiple functions, teams, and technology environments, the challenge is often not building AI models but deploying them consistently in ways that create measurable business value.

In a series on why enterprise AI deployment has become a defining bottleneck and competitive battleground for modern retail, Larissa Schneider, Co-Founder and COO at Unframe, and Chris Slovak, Global Field CTO at Unframe, join Emerj on the AI in Business Podcast to discuss the operational barriers that stall AI initiatives and the approaches organizations are using to move from pilot projects to production systems at scale.

This article examines the decisive operational insights from these discussions, that determine whether enterprise AI ever reaches production in modern retail:

• Unified AI program design for enterprise‑wide alignment: Establish a single architectural and governance hub that stops teams from building isolated pilots and ensures every solution snaps into one coherent enterprise AI system.
• Modular AI components for rapid enterprise deployment: Assemble AI workflows from reusable components so teams can launch and iterate solutions in weeks instead of rebuilding bespoke systems that stall and require constant re‑architecture.
• Distributed reasoning across retail systems for immediate value: Enable agentic systems to pull context directly from ERP, CRM, POS, and warehouse tools so AI can deliver accurate, real‑time outputs without waiting for multi‑year data‑centralization projects.
• Reusable Enterprise Context for Faster AI Deployment: Build a reusable layer of business knowledge, system relationships, and operational context so every AI deployment accelerates the next.

Unified AI Program Design for Enterprise‑Wide Alignment

Listen to the full episodes below:

Episode 1:  AI Deployment at Retail Speed – with Larissa Schneider of Unframe

Guest: Larissa Schneider, Co-Founder and COO of Unframe

Expertise: Enterprise AI; Go-to-Market; Product Marketing; Enterprise Technology

Brief Recognition: Larissa Schneider is Co-Founder and COO of Unframe, where she leads efforts to help enterprises deploy AI use cases. She previously held go-to-market and marketing leadership roles at Nutanix and Noname Security, including leading global marketing for Frame before and during its acquisition by Nutanix. She holds a master’s degree in International Marketing Management with Distinction from Hult International Business School.

Larissa Schneider describes organizations where AI experimentation is happening faster than leaders can see, track, or evaluate. Employees across departments are building small automations, agents, and POCs to eliminate manual work — copying data between systems, stitching together workflows, or testing low‑code tools on their own. These efforts are well‑intentioned, but without a central program, they create a landscape in which AI activity is happening everywhere, and leadership cannot determine how it fits together or whether it can be trusted.

She explains the dynamic:

“You end up with these little islands of projects — trials and POCs running across every part of the organization, which naturally happens. And it’s great; we love people who are eager to try new technology and innovate their work. But how does all of that fit together? That’s the question we’re asking. Because to have reliable, governable, secure AI that truly moves the needle for the business, everything has to work in tandem — everything has to work together.”

–  Larissa Schneider, Co-Founder and COO of Unframe

For Larissa, governable AI means more than policy documents. Leaders need operational visibility into how AI workflows operate, who owns them, what data they use, and how outputs are evaluated.

• What AI is being built
• Which systems it touches
• What data flows through it
• Who owns the workflow
• How output quality is measured

Without that visibility, pilots stall not because the model is weak, but because the enterprise cannot trust or scale what it cannot see.

Larissa’s guidance for C‑suite leaders forms a clear structural playbook:

• Create one AI program that everything plugs into: A single architectural center prevents teams from building isolated tools that cannot be evaluated or deployed beyond their department.
• Define output‑quality benchmarks before scaling: Employees adopt AI only when outputs are consistent. Quality criteria must be shared across all workflows so leaders can measure reliability.
• Create visibility into workflows, decisions, and data movement: Leaders need to know where data originates, what actions agents take, and who is accountable for the outputs.
• Require every new use case to align with the same backbone: This ensures that AI activity across the enterprise is coherent, measurable, and compatible with future initiatives.

Larissa describes this as a structural shift — AI fails when it is scattered. It scales when leaders create one place where all AI activity fits together. A unified program gives executives the visibility, trust, and architectural stability required to move from experimentation to enterprise‑wide deployment.

Modular AI Components for Rapid Enterprise Deployment

Larissa breaks down the component‑level insight that underpins a modular approach:

“What we noticed very quickly is that AI use cases can come from all different parts of an organization, from all different industries and verticals and teams, but the underlying components that you need to put these AI use cases into practice are actually very similar. And those can range from very basic UI components and dashboards to enterprise connectors to your SAP and Salesforce instances and your data lakes. They can be things like reasoning and AI audibility, it really depends on the use case. But you would be surprised to see how much similarity we can see between inventory planning in retail and automated claims processing in insurance or lease abstraction in commercial real estate.”

–  Larissa Schneider, Co-Founder and COO of Unframe

Larissa’s central argument is that enterprise AI deployment becomes difficult when organizations treat every use case as a custom engineering project. While inventory planning, claims processing, and lease abstraction may appear unrelated, she notes that many rely on the same underlying connectors, interfaces, reasoning layers, governance mechanisms, and workflow components. As she puts it, “We open our box of Legos, we have hundreds of them at this stage, we take out the few dozens or the hundreds that we need to put together that use case.” The opportunity is not to build every solution from scratch but to assemble solutions from reusable building blocks.

Once leaders recognize that, deployment stops being a reinvention exercise and becomes an assembly exercise. The work shifts from building to configuring.

Here’s how Larissa translates that into action for executives:

• Build once, reuse everywhere: Components like SAP/Salesforce connectors, reasoning modules, and UI elements should serve dozens of workflows, not one.
• Let the architecture carry the complexity: The modular system should handle the heavy lifting so each new use case requires only light configuration.
• Treat integrations as shared infrastructure: Core system connectors shouldn’t be rebuilt for every department — they should be standardized and inherited.
• Customize only the final layer: The unique logic of a workflow is the only part that should require bespoke engineering.
• Scale horizontally, not from scratch: Once the components exist, new use cases become incremental rather than transformational.

Larissa argues that most enterprise AI efforts are not about writing code. The larger challenge is defining workflows, outputs, user requirements, and business value before any implementation begins.

From Larissa’s experience, modularity is what makes enterprise AI deployable at speed. When teams assemble solutions from reusable components, timelines compress, integration risk drops, and every new workflow inherits the stability of the ones already in production.

Distributed Reasoning Across Retail Systems for Immediate Value

Episode 2:  The Predictive Model Reshaping Retail Operations at Scale – with Chris Slovak of Unframe

Guest: Chris Slovak, Global Field CTO at Unframe

Expertise: Artificial Intelligence; Data Infrastructure; Go-to-Market Strategy; Solutions Consulting

Brief Recognition: Chris Slovak is Global Field CTO and Head of AI Architects at Unframe, helping enterprises achieve business outcomes with AI. He previously spent nearly eight years at Tealium, where he led global solutions consulting, helped grow the business to 20x ARR across three financing rounds, supported global expansion to eight international offices, and secured six patents. He also co-founded Challenger Interactive, which developed patented AI technology for gaming.

Chris Slovak describes retail environments in which core operational truths are scattered across ERP, CRM, POS, and warehouse systems. Order state, supply status, inventory accuracy, and customer context each live partly in different tools. Retailers often try to solve this fragmentation by centralizing everything first — spending years aggregating, cleaning, and modeling data before any workflow can move into production. Chris is of the opinion that this approach stalls deployment long before AI can deliver value.

He argues that treating data centralization as a prerequisite for AI has become one of the largest deployment bottlenecks in retail.

In his interview he explains that agentic systems can operate directly on distributed systems, pulling the specific pieces of context they need in real time rather than waiting for a unified schema or a completed data‑warehouse migration:

“Agents and agentic systems in particular don’t necessarily need one source of aggregated data truth. They can reason like a human can across multiple systems, so long as there’s context and semantic linking. You have ERPs and CRM, and then you have your point of sale systems, and they probably all semi-talk, but the truth is the state of order, supply, inventory probably to some extent live a little bit in each.”

— Chris Slovak, Global Field CTO at Unframe

Centralizing everything first also disrupts workflows. Teams must change how they work, adopt new tools, and reorganize processes to satisfy a data‑architecture ideal. Chris notes that this becomes a bridge to nowhere — systems evolve, businesses evolve, and the centralized model never reaches a stable end state. Allowing agents to pull data where it lies and how it lies avoids that disruption and removes the adoption risk.

Chris points out the following actionable realities:

• Retail truth lives across ERP, CRM, POS, and warehouse systems — no single system contains the full picture.
• Humans already combine these partial truths manually; agents can replicate that reasoning in real time.
• Multi‑year aggregation projects delay value and block deployment.
• Using data in place avoids workflow disruption and accelerates outcomes.
• Agents can pull only the components they need from each system, eliminating the need for a unified schema.

The structural conclusion drawn from Chris’s remarks is that AI becomes immediately deployable when agents pull context directly from ERP, CRM, POS, and warehouse systems. This bypasses the centralization bottleneck and enables real‑time inventory, supply, and promotion decisions on the architecture retailers already have.

Reusable Enterprise Context for Faster AI Deployment

Chris Slovak describes a recurring pattern he sees in enterprise AI initiatives: organizations spend months or years perfecting data foundations before solving a single business problem. Teams focus on data migrations, new models, and centralized architectures, assuming that AI can only deliver value once the entire environment is rebuilt. In practice, this delays deployment and prevents organizations from learning which use cases actually matter.

Chris asserts that AI deployments should begin with a tangible business problem and expand through short implementation cycles that generate measurable value. Rather than waiting for a perfect environment, organizations should build solutions using the data, systems, and workflows they already have. Each deployment then contributes new business knowledge, system connections, and operational context that future initiatives can reuse.

He explains:

“The concept that the core context is going to evolve, that has to be core to your design decisions, if my first use case gives me exposure to 60% of the major business entities and tools that I use today, use case number two is already 60% of the way there.”

— Chris Slovak, Global Field CTO at Unframe

The implication for leaders is that the speed of AI deployment compounds over time. Each successful implementation creates reusable context that reduces the effort required for the next one.

Chris highlights several practical principles:

• Start with a business problem, not a transformation program: Focus on a specific operational challenge where AI can create measurable value quickly.
• Deliver value in short cycles: Smaller deployments create opportunities to learn, adjust, and improve without committing to multi-year bets.
• Treat enterprise context as a reusable asset: Data relationships, business rules, workflows, and system connections created for one use case can accelerate future deployments.
• Build for continuous evolution: Models, integrations, and business requirements will change. Architectures should assume change rather than resist it.
• Allow knowledge to compound across use cases: Every deployment should make the next deployment easier, faster, and more informed.

From Chris’s perspective, the real advantage does not come from completing a single AI project. It comes from creating a growing layer of reusable business context that shortens implementation timelines and increases the value of every subsequent deployment. Organizations that build this foundation can move from isolated AI projects to a repeatable system for enterprise-scale transformation.

Together, Larissa and Chris describe deployment as a compound process. Larissa focuses on the organizational and architectural foundations that make AI governable, while Chris explains how agents can begin delivering value immediately using distributed data and reusable context.