All About AI: Joshua Young of Kalypso

We’re observing a big disconnect (in both our own fashion-only data and in wider benchmarks from other sources) between the increasing capability and reliability of AI, in both a general sense and in discrete applications, and the lack of trust that end users place in its outputs. What do you think trust in AI looks like, and what’s the right rubric for measuring it?

At its best, fashion is human, physical, emotional, and cultural. Being worn on the body, it can create performance, beauty, identity, or all three at once. It can be the Gucci Fashion Show taking over Times Square, with the noise, traffic, billboards, celebrity, commerce, and chaos of New York moving around it. Fashion can be precise and out of control at the same time. That contrast is part of the magic. The outfits are important, but so is the world they enter, the people wearing them, and the meaning a brand creates around them.

That is why AI-generated fashion can feel uncomfortable. When people see synthetic models or products that feel assembled by algorithms, they can sense the absence of human intent. It may be technically impressive, but it can still feel emotionally lacking.

But I do not think that is the most useful way to frame AI’s role in fashion.

The bigger opportunity is not to ask AI to become the designer or the taste-maker. AI may inform design decisions, but its greater value is in the things machines have always been good at: making sense of complexity, finding patterns, reducing friction, orchestrating work, and helping people make better decisions.

A product must move from strategy to execution. Along the way, teams make hundreds of decisions about what to create, what to buy, how much to make, when to commit, and how to bring it to market. Those decisions are rarely made with perfect information. They happen across functions, systems, regions, suppliers, and channels. That is where AI becomes useful, not because it replaces judgment, but because it gives judgment better context.

Trust should not be measured by whether people believe the AI. Trust should be measured by whether AI improves confidence in a specific decision. Can teams understand the signal? Can they see the assumptions? Can they act on the recommendation without losing accountability, brand intent, or product integrity?

The goal is not to manufacture taste. The goal is better human decisions.

The most prominent use cases for generative AI are at the beginning and the end of the product funnel: in ideation and visualization early on in the lifecycle, and then in buyer-facing content creation afterwards. At least in part, that prominence is down to the maturity of the underlying models; image generation and editing models have come a long way in just a couple of years, and if a company wants to explode the number of ideas they can create, or wants to slash its photography and videography budgets, the tools are clearly there to make that happen.

But is that destined to remain the big-ticket application of AI? Or do you see something more profound coming?

In product creation, there is a useful concept called “Best Available Image.” At any point from concept to consumer, teams need the best available representation of a product to support decision-making.

Historically, that usually meant one of two things: a 2D sketch or a photograph. A sketch is easy to make, but low fidelity. A photograph is high fidelity, but difficult to create because it depends on physical product availability. More recently, 3D has created a middle ground. It is much more accurate than a sketch and less dependent on physical samples, but still relatively difficult to create at scale.

Generative AI image creation has disrupted this thinking. It is very easy to create and can be high visual quality, even if not always completely accurate to the final product. But from an intent perspective, it has been immensely valuable in allowing teams to approve product concepts at speed and will continue to become more product accurate as the technology improves.

That said, a high-quality image is not the same thing as an accurate representation of the manufactured product. Generative AI can create a strong statement of intent, but it does not necessarily tell you how or if the product can be made. And images by themselves have limited value if they are not connected to data or decision-making systems.

This is where many companies miss the point. Digital images only create better value than physical samples when they are connected to a data-driven, visual decision-making environment. Otherwise, teams are still manually copying and pasting images into offline tools like Miro, PowerPoint, or Adobe workflows. The image may be digital, but the decision process is still manual.

So yes, generative AI can produce beautiful images. But by itself, it does not solve line and assortment planning, forecasting, product development, manufacturing, material approvals, retail planning, or the many other business-critical and high-friction parts of the product process.

That leads to the more important question: where does fashion actually lose speed, confidence, and agility? For most leaders, image generation does not top that list.

If that’s true, and the milestone opportunity in front of fashion is responsiveness to market demand, what does it look like to practically achieve that? There’s a huge volume and variety of market and consumer signals out there, as well as a long-established and entrenched art of forecasting. How do you see AI helping brands to better parse and act on external data, quickly? And what impact do you think this has on the way the typical product calendar operates?

That’s a great question. Reducing friction and creating agility requires several things. AI is certainly one of them. Companies need the ability to aggregate and process large amounts of real-time data very quickly, but they also need to orchestrate more complex workflows across teams, systems, suppliers, and channels.

But technology is only useful if the process can take advantage of it. Or maybe we should look at it the other way around: what would your ideal process look like if you were not constrained by your current technology?

Could forecast decisions be made much closer to market, reducing the risk of significant overproduction and underproduction? Could merchandisers build a more consumer-relevant line with faster access to better market data? Could supply chain orchestration help ensure the right materials are in the right place, at the right time, in the right quantities? Could brands better read market sentiment, consumer behavior, and retail performance? Finally, could decisions move through the enterprise without being slowed down by systems, meetings, files, and functions?

This is not about creating a fast fashion company, abandoning innovation, or chasing the market. It is about making better decisions, faster, at the right time, and with more intentionality. That does not mean reacting to the market. It means planning to make decisions when they are most likely to be accurate, based on the best available data.

In some cases, that may have very little to do with AI. It may mean moving material innovation and approval into a seasonless process instead of keeping it inside the product creation critical path. It may mean starting the season with a tested and approved color palette. It may mean moving some manufacturing closer to market and then flowing product based on consumer signals.

But we’re here to talk about AI, so I would start with one of the things most companies struggle with: creating the line plan and assortment structure. Tools such as Stylumia and Heuritech can help teams make sense of historical sales, consumer signals, ecommerce data, social imagery, competitive activity, and real-time market movement. This does not mean AI creates your product line for you. It means a merchandiser can make better decisions based on stronger evidence, brand strategy, and human taste.

The next question is commitment. Most companies still make quantity, pricing, and inventory decisions far in advance of the market. But agility is about making the right product, at the right time, in the right quantities. Planning to hold some open-to-buy closer to market, then using tools such as Stylumia or Invent.ai to support demand sensing, buy-depth decisions, localized assortment decisions, and allocation is a better model than locking everything too early.

But that only works if the supply chain can support it. This is where platforms such as Inspectorio and TrusTrace become relevant. They do not magically make the supply chain agile, but they help create better visibility into supplier risk, quality, compliance, traceability, and production readiness. That visibility is critical if brands want to preposition greige, delay dye decisions, multi-source closer to market, manage tariff exposure, or flow product based on consumer signals.

Finally, how much do brands really know about their consumer? How is a product landing in the marketplace and shaping consumer sentiment? What regional opportunities are being missed because a company did not see a cultural movement early enough? How are consumers responding to the retail experience, marketing, fit, price, or the product itself? Tools such as Corto and MakerSights use AI as the listening system for the market.

Not just what sold, but what people are saying, searching, returning, reviewing, sharing, and buying. Not just demand after the fact, but the signals that help brands understand what is gaining traction, what is losing heat, and what deserves action.

That is the practical opportunity. AI can help brands sense more signals, make decisions closer to the moment of truth, and move those decisions through the enterprise with less delay. But the calendar and process must change with it. If they are still built around locking every decision far in advance, then AI becomes another insight tool trapped inside an old operating model.

Thinking about the specific disciplines that make up some of that journey we’ve just talked about – line planning, manufacturing commitments, inventory planning and allocation – these are usually done by dedicated teams using task-specific tools, and the output they create is what gets seen by other people. Nobody in design is thinking about channel allocation, for example.

It feels like AI might be about to change that, by extending the user community for those kinds of functions to be “basically everyone,” in the sense that deciding what to make, what to make more (or less) of, and what not to make at all is a multi-disciplinary challenge, so giving a wider cohort of people the ability to query some of those signals feels like it might be the right approach. Do you see it that way, or is the change to those functions more likely to stay within the well-worn lanes we already have?

I do not think AI means everyone suddenly becomes a planner, merchant, designer, or supply chain expert. Each of these subject matter areas require specific expertise, accountability, and understanding of trade-offs. What AI can change is the context available when those decisions are made.

A designer does not need to own product development, but they should be able to understand if a design choice creates risk for material, cost, margin, or supply chain execution. A merchant does not need to make technical design decisions, but they should be able to see where the line is becoming too complex, repetitive, or disconnected from consumer signals. A planner does not need to dictate product taste, but they should be able to understand where consumer sentiment, brand heat, or marketplace behavior may challenge the forecast.

This is where the opportunity lies; not everyone having to be involved in every decision, but in having broader context to make the right decision within their function.

The risk is that companies mistake broader access to data for broader decision rights. That would create more volatility, not more agility. AI should not turn every function into a committee but should help each function make its own decisions with better visibility into the consequences.

In smaller companies, where specialized roles may not exist, AI may help teams validate decisions outside their core expertise. But for large brands, the goal should be different: keep ownership clear, make the decision environment richer, and help teams understand how their choices affect the rest of the system.

So no, I do not think the future is “basically everyone” making every decision. The better future is experts staying in their lane, with a clearer view of the road around them.

AI should not blur accountability. It should make accountability better informed.

AI is obviously a catalyst for this conversation, but this has the telltale feeling of all being as much of a cultural change as a technological one – if not more so. And it’s also inevitable that the use cases for AI are going to need to be rolled into existing transformation strategies that encompass digital twins, automation ambitions, production orchestration and distribution platforms, and plenty more. If a fashion brands want to actually take what AI can do, and then build an agile go-to-market engine on top of it, where should it start and what does the sequencing of work look like thereafter.

We’ve all been in situations where systems technically meet design requirements, but are actively user-hostile. At the same time, they often operate as disconnected islands. Or worse, teams are using off-the-shelf tools for work they were never intended to support.

Coming back to the Gucci example, that New York fashion show had a vast array of components: lighting, sound, crowd control, hair and makeup, models, and the fashion itself. Yet it came together as an intricately choreographed performance that unfolded with ease, delighted the viewer, and integrated seamlessly with its surroundings. All the roles, as different as they were, operated as one.

If only fashion technology worked as smoothly.

For true agility, there are several unlocks. The first is usability. Our industry is largely populated by creative thinkers who need processes and systems that are intuitive, visual, and easy to use. Unfortunately, many of the tools used in fashion come from industries whose primary focus is engineering or technology. Closing that usability gap, with real focus on persona design and user experience, is critical if teams are expected to work faster and make better decisions.

The second unlock is integration. Many fashion systems are disconnected by default. Adobe and Microsoft tools are useful, but they were not designed to operate as connected product decision environments. Other systems have been built without enough understanding of shared decision-making or the value of a single source of truth.

The impact goes far beyond usability. Disconnected systems lower the quality of decisions, create lag, and introduce more opportunities for error. Copying and pasting from PLM to Excel to SharePoint to Miro to Adobe Illustrator is neither effective nor efficient, yet this is still the reality at many companies.

Solutions such as VibeIQ, Trasix, and Centric Visual Boards point toward a better model: visual, data-connected decision-making environments where content can remain in its system of record while being surfaced where, when, and how each function needs it. Done well, this creates visibility vertically and horizontally. Leaders can see how decisions roll up. Downstream teams can see what has been decided. Functions can work from the same context without forcing everything into the same tool.

That creates the agility, accountability, governance, and accuracy required to operate at modern cultural speed.

The third unlock is orchestration. Even with better integration, many process steps still must be moved forward manually. Teams often lack visibility into what should happen next, who needs to act, which exception requires attention, or which decision can move forward based on rules, timing, and triggers.

This is where agentic AI and orchestrated AI solutions become useful. Not because every decision should be automated, but because many non-value-added manual steps do not require human judgment. If the work is routing, checking, flagging, summarizing, updating, monitoring, or reminding, AI should be able to help move that work forward.

Inspectorio is a useful example of this type of shift in the production chain. Work that previously required significant manual effort to monitor, submit, validate, and document quality, compliance, traceability, and supplier data can be more intelligently managed. That gives human experts more time to focus on the issues that require physical investigation, context, and judgment: a material supplier with repeated wastewater compliance issues or a quality team investigating outsole delamination.

The same principle applies in design. We constantly hear about the time designers spend on repetitive documentation tasks instead of designing. That is not a good use of creative talent. AI should help remove the administrative drag around expert work, not dilute the expertise itself.

So where should companies start? We suggest looking first at the decision or workflow creating the most friction.

Identify the decision that needs to improve, such as line adoption, sample approval, supplier readiness, allocation, or market response. Then define the process that would be ideal if the current technology constraints were removed. After that, connect the systems and data needed to support the decision. Then layer in AI, automation, and orchestration where they remove manual work, improve visibility, or help teams act faster, while keeping usability in view.

This is not a small productivity issue. APQC found that knowledge workers lose roughly a quarter of the workweek to productivity drains, including internal communication, searching for information, unnecessary meetings, workarounds, duplicated effort, and finding the right experts. In fashion, that drag shows up as slower decisions, weaker handoffs, and less time for functional experts to do the work that requires their judgment.

If a brand wants to build an agile go-to-market engine, it should not start by asking where AI can be inserted. It should start by asking where expert time is being consumed by work that does not require expertise. That is where usability, integration, orchestration, and AI can create real value.

Last year, we asked technology executives whether they predicted AI would become more obvious, as a primary interface paradigm, or whether it would disappear into the background in the way that vital but invisible platforms like AWS and Azure have done. This year’s data suggests that the answer is both: fashion professionals that use AI interact with it as both the engine and the steering wheel. But we also see that, for a lot of companies, both kinds of applications are still largely in the scoping or refined-pilot stage. What do you see being the trigger for deeper adoption and wider roll-out?

The answer is not more AI or different kinds of AI. It is better human-centered design.

Humans like shiny objects. When exciting new technologies drop, we pick them up and then start searching for problems we can solve with them. That approach rarely works because it avoids the harder discussion: what is the root cause, how do people actually work, and what should the ideal process look like?

In Kalypso’s experience, many technology teams think the solution to a business problem is a technology implementation, and success is judged by adoption. In reality, business problems are best solved by human-centered process change supported by the required capabilities, and success is judged by value delivery. I am using the word “capabilities” on purpose, because not every capability is technology-related. The minute people start talking about this or that technology as the solution to a problem, they tend to over-index on the technology itself. The solution is usually some combination of behavior change, process change, calendar change, data improvement, and technology.

So perhaps we should update the question. Instead of asking how AI moves from pilots to scaled adoption, we should ask: how can AI capabilities support scaled business process change as part of a holistic enterprise operating model?

There are three main components to this.

The first is the one we have already discussed: start with a modern, human-centered, agile operating model and identify the capabilities required to support it. AI will be much more successful when it supports a strong process rather than operating as a point-solution fix.

The second requirement is data. To operate with agility, accuracy, and speed, companies need real-time, accurate, and available data across the business. From concept to consumer, across Discover, Create, Make, and Move, teams need clean, integrated, and timely access to data. We have all heard the phrase “garbage in, garbage out,” and nowhere is that more true than when ensuring AI output quality.

The third requirement is a rethink of technology architecture. The old approach was to focus on large monolithic ERP, PLM, and DAM systems that tried to serve many needs at the same time. The future looks more composable.

In the same way that a fashion show is supported by many expert specialists rather than a monolithic group of fashion generalists, we should be looking at a similar approach to building fashion technology capabilities. That also aligns with how AI is likely to evolve: more specialized, more contextual, and trained or tuned around specific domains of work.

The future of enterprise systems may look less like one giant platform and more like a data-rich backbone surrounded by modular applications, AI agents, and function-specific insight engines. The goal is to give the human knowledge worker exactly the information they need, when they need it, to make the best decision possible.

There is another advantage. As applications move to the cloud and AI becomes more specialized, technology can improve more continuously. Teams should not have to wait years for the next monolithic transformation program to get better tools. A composable architecture allows companies to upgrade capabilities more easily and keep the best tool available for the work being done.

That is the composable future.

Ironically, the future looks more human, not less. Technology becomes more useful when it better supports human decision-making and productivity. To get there, companies need to resist the temptation of shiny objects and focus instead on usability, process, data, and operating model change.

At the end of the day, fashion is emotional, physical, and personal. AI should reflect that reality, not fight against it.