When more capability meant more people
For most of retail history, pricing capability scaled much like a horse-drawn carriage: if you wanted to carry more or move faster, you needed more horses. For retailers, those horses were pricing analysts. Expanding coverage, maintaining rules, and fine-tuning prices all required more people. The combustion engine changed that equation by putting the power inside the machine. Pricing technology is now doing something similar, making capability less dependent on the size of the team behind it.
But the way pricing software has traditionally been implemented for large and complex retail organizations has not made that transition as fully as the technology itself. Vendors have added AI, and the models genuinely produce better signals. Yet better signals do not necessarily mean less work. In many cases, the effort has simply shifted elsewhere. Instead of a retailer needing more analysts, it needs vendor-side consultants, data scientists, implementation teams, and custom development to configure the platform around the way its business actually operates. The people may be different, but the dependency on people remains.
That is what changes when more of the capability sits within the technology itself. The measure of a powerful pricing platform is no longer just the sophistication of its models, but how much a retailer can achieve without having to scale the team around them.
Dehner, one of Europe’s leading garden and pet retailers, runs more than half of its online turnover through Quicklizard with just one internal team member overseeing it. This is not a small pilot or a narrow use case. It is pricing across the majority of a large retailer’s e-commerce business, managed by a small team because much of the complexity and manual work that once required more people is handled by the platform itself.
Why fit has always been expensive
Pricing software for large and complex retail organizations has traditionally offered two ways to buy, each with its own compromise.
The standardized product deploys faster but forces the retailer into a fixed structure, meaning the retailer has to adapt its strategy to the software rather than the other way around. When the two do not fit, it is usually the strategy that gets simplified. Commercial knowledge built over years gets flattened into what the platform can represent, and the retailer ends up working around a system that was supposed to work around the business.
The alternative is a heavily customized implementation. It can reflect the business much more closely, but that flexibility comes at a cost: years of consulting, bespoke development, and specialist support. Every requirement the base product did not anticipate becomes something that needs to be designed, built, and maintained, whether by the vendor’s professional services arm or an external systems integrator. And the cost does not end at go-live. As the business evolves, new requirements often mean returning to the same cycle of specification, development, and implementation.
So while the two approaches look different, they create the same fundamental trade-off: either the business adapts to the software, or the software is adapted to the business at significant cost.
The underlying problem is that customization has traditionally been treated as a project. It gets scoped, staffed, and invoiced, and because each change requires another project, meaningful customization becomes something retailers can only afford to do periodically.
Quicklizard changes that model by building customization into the software itself, rather than treating it as a separate project around it. That fundamentally changes the economics of customization for large and complex retail organizations.
Customization built into the architecture
Every software and data science module in Quicklizard is built with the expectation that it will need to adapt to the retailer using it. Extension points, customer-specific logic, model training, and data structures are accounted for from the start, rather than retrofitted for each account. That is the difference between a platform designed to accommodate a retailer’s requirements and one that needs to be reworked every time those requirements fall outside the standard product.
In practice, retailers can write their own pricing logic directly into the platform. That logic runs alongside Quicklizard’s models, rules, guardrails, channels, and approval processes rather than being added as a separate layer around the system. Half of Quicklizard’s customers already work this way, with customer-written logic executing 6.5 million times a day. This is not an edge case reserved for the most technical customers. It is a normal part of how the platform is used.
The same principle applies to the data science layer. Price Elasticity 2.0, Demand Forecasting, the Competitor Sensitivity Index (CSI™), Article Segmentation, and Clearance & Markdown are not fixed black boxes shipped identically to every account. Each is designed to be trained and tuned using the individual retailer’s data and combined according to the outcomes the business is trying to achieve. Channel Tree and Tiered Pricing then extend that flexibility across online, marketplace, regional, and physical-store operations, allowing different parts of the business to follow different pricing approaches without requiring separate systems.
The same design determines how quickly a retailer can change course once it is live. When a pricing strategy needs to change, the change is made directly in the platform and takes effect in the next pricing cycle. There is no release, no support ticket, and no consulting project, so pricing logic keeps pace with the business instead of waiting behind a development queue.
When the business expands, the platform expands with it rather than requiring another implementation. One in six customers has already grown beyond its first use case, running rules-based, hybrid, and fully AI-driven pricing side by side for different parts of the business on the same platform, without implementing a second solution.
There is a direct commercial consequence to this approach. Quicklizard does not depend on consulting revenue to rebuild or heavily adapt the platform for each customer, nor does customization need to be handed off to external partners. The ability to adapt is already part of the architecture, reducing the implementation effort and cost associated with traditional, highly customized pricing software deployments.
What agents add before and after go-live
Agents and large language models sit on top of that modular foundation, extending what retailers can do with it. Instead of only asking what a price should be, the platform can also examine what the retailer is trying to achieve, whether its current strategy is producing that outcome, and where the underlying logic could be improved.
Pricing agents explain why a specific product received a specific price, gathering the relevant history, strategy, and supporting evidence into an explanation that a person can follow. Our Strategy Challenger Agent examines the strategy itself, identifying where pricing rules repeatedly constrain recommendations or produce results that conflict with commercial objectives, and proposing improvements for human review. The category Optimizer takes the decision beyond an individual product and looks across the category, considering substitution, complementarity, and cannibalization while the retailer sets goals around revenue, margin, sell-through, or price image. Quicklizard’s MCP connection makes the same intelligence available to other authorized AI assistants and workflows while preserving the explanations, limits, and controls that make pricing decisions accountable.
These capabilities are grounded in continuously refreshed market and pricing data. Quicklizard processes 2.5 billion competitor prices a month, nearly a thousand every second, refreshes more than 400,000 product-level elasticity estimates each week, and prices over 30 million SKUs daily across 60 countries.
Agents also play a less visible role before the retailer ever goes live. AI can handle much of the data-onboarding work, helping interpret the retailer’s feeds, structures, and assortment logic and map them into the platform. Data onboarding has historically consumed significant specialist time in complex pricing software implementations. Automating much of that work helps explain why implementation timelines can be compressed so dramatically.
Why weeks instead of years is not corner-cutting
Quicklizard customers have reached live pricing in a matter of weeks, even across large assortments and multiple markets. One retailer with more than 15,000 active products went live within five weeks, while a global consumer-electronics brand operating across ten European markets did so within seven.
Timelines like these naturally raise a question: what had to be sacrificed to move that quickly? The answer is not the retailer’s requirements, but two categories of work that traditional implementations often create in the first place. Legacy implementations spend significant time adapting a static architecture to requirements it was not designed for and building new modules for specifications the base product never anticipated. When the architecture is designed for customization from the outset, much of that work disappears. AI-assisted onboarding then compresses much of what remains, further reducing the time and specialist effort required to go live.
That leaves more time for the work that genuinely requires people. The main implementation effort becomes working with the retailer’s own stakeholders: aligning commercial teams, agreeing on guardrails and approval thresholds, and managing the change in how pricing decisions are made. The modular architecture creates room to focus on those decisions rather than consuming the implementation with technical adaptation.
Automation grows with trust
Going live quickly does not mean handing control over to AI on day one. Retailers increase automation at their own pace, because adaptability without accountability is not deployable at the scale and complexity of large retail organizations. Glass Box AI makes every decision visible, explainable, and traceable. Guardrails define what the system is permitted to do, while tiered approvals keep people involved wherever the business wants them. The Audit Log provides a record of every change and who made it, supporting accountability across the pricing process. These controls operate within Quicklizard’s SOC 2-compliant environment. Recommendations are also built only from product, category, channel, demand, competition, and inventory data, never from an individual shopper’s identity.
The result is a level of automation that retailers are willing to put into production. In the last quarter, Quicklizard made a quarter of a billion pricing decisions, 99% of them automatically within each retailer’s own guardrails, while fewer than one in 5,000 was overridden by a person. That level of autonomy is reached gradually rather than assumed from day one. Retailers can begin with rules and human approvals, then give the models greater autonomy as confidence grows.
That progression is visible across individual retailers as well. Camping World increased the share of its catalog on AI pricing from roughly a third to nearly 90% in three months. Dehner expanded its AI-priced online assortment from 8-9% to 46-54%, while achieving a 5% conversion uplift. Elkjøp has now automated approximately 75% of its pricing catalog using Quicklizard’s competitor intelligence, segmentation, and forecasting.
The impact extends beyond automation itself. Customers have achieved up to 15% revenue growth, an 11% profit increase, and a 300% improvement in pricing-team productivity.
Traditional pricing platforms often still require more people to create more capability, whether those people sit inside the retailer, with the software vendor, or at an external integrator. Quicklizard is built around a different assumption: more of that capability should live inside the platform itself. That is what makes deep customization for large and complex retail organizations possible without the lengthy, resource-intensive implementation traditionally associated with highly customized pricing software. It is not the same model made faster. It is a different architecture.
