The maturity gap

Organizational design and UX readiness in the age of AI

Abstract: AI deployment showcases design maturity, or the lack of it, in product processes. For many companies, the results are confronting. Underinvestment in design infrastructure, research capability and cross-functional governance creates clear risks. After all, AI amplifies design decisions at scale. As a result, their consequences are faster, wider and harder to reverse than in any previous technology cycle.

This article takes that diagnosis seriously. At the same time, it argues that the gap can be measured, structured and closed. To do so, the article:

1. introduces a three-part framework for identifying UX debt, model debt and organizational debt;
2. maps the structural conditions that distinguish organizations where design has institutional authority from those where its role remains advisory;
3. argues for a specific organizational model: design as an autonomous discipline with a defined collaborative architecture, able to deliver human-centered value at scale.

A critical dimension of this argument is the capacity to design for care-based trust and human agency at once. When organizations fail to govern both, the result is not merely a suboptimal product. Instead, they build systems that simulate genuine relationship while denying its essential condition. Finally, the McKinsey Design Index and Forrester economic impact data anchor the investment case for organizations that act now.

What AI reveals and why that is an opportunity

AI exposes problems that already existed

AI deployment does not create organizational design problems. Instead, it reveals them, at speed and at scale. Many organizations struggle with AI product quality, weak adoption and a persistent gap between what their systems do and what users need. However, in most documented cases, the cause is not model limitations. These companies struggle because design, as an institutional practice, never had the authority or scope to shape what those systems were built to do.

A window of opportunity for leadership

For leadership, the productive reframe is this: AI has made the cost of design immaturity visible. Moreover, it has done so at a moment when fixing it is still manageable. Companies that treat this as diagnostic information, rather than a verdict, can close the gap now. In this window, the investment required is bounded and the competitive return is meaningful. On the other hand, those that treat it as a fixed condition accumulate consequences as fast as their AI deployment accelerates.

What defines a design-mature organization

In a design-mature organization, human experience has institutional authority to constrain technical and business decisions. In other words, it does not merely inform them after the fact. That authority is structural, and it covers the full scope of design in AI systems. This includes the interface and the visual language, but also the communicative register and the behavioral architecture. It also covers the balance between care and human agency, and decisions about what the system will and will not do for the user. For the first time, AI deployment makes that authority measurable in product outcomes, at a scale that commands executive attention.

Three forms of organizational debt

Organizations deploying AI without mature design infrastructure tend to accumulate three forms of debt. They are distinct, yet interlocking. In addition, they require different interventions and carry different recovery timelines. Even so, naming all three already puts a company ahead, because diagnostic precision is the precondition for acting on any of them.

UX debt

UX debt is the accumulated deficit of unresolved user experience decisions. It includes every interaction pattern adopted for velocity rather than user need. Beyond that, it covers every research finding acknowledged and then deprioritized, as well as every friction point deferred to an iteration that never arrived.

In AI-enabled products, this debt becomes structurally dangerous. That is because probabilistic outputs place interpretive load on users who must decide based on variable results. When the interaction design is incoherent, that cognitive burden accumulates. Then the lack of trust that follows grows rapidly and silently. On the other hand, UX debt is the most immediately addressable of the three. It is visible in user research, recoverable through focused investment and generates measurable return in adoption and retention within a reasonable product cycle.

Model debt

Model debt is the gap between what AI systems are optimized to achieve and what users genuinely need from them. It emerges when the metrics driving model training are proxies for user value rather than direct measures of it. For example, engagement substitutes for genuine value, session length for meaningful use and confidence scores for contextual accuracy.

Without research embedded in development, AI systems are designed for a user who was never actually observed (Auernhammer, 2020). Thus, model debt shows up in high accuracy scores paired with declining engagement. Similarly, it appears in technically correct outputs paired with low adoption. Addressing it requires redesigning the feedback loops between user research and model optimization.

Organizational debt

Organizational debt is the most structurally significant. After all, it is the condition that allows the other two to accumulate unchecked. This debt refers to the institutional arrangements that prevent design from operating with the authority its scope requires. Among them are design teams reporting into product management rather than holding independent standing. There are also research findings filtered through prioritization layers without design literacy. Finally, design is often absent from the rooms where decisions about automation, data use and system behavior are made.

When design has no institutional standing in those decisions, the distribution of agency between human and system defaults to whoever does (Shneiderman, 2022). As a result, systems may perform their stated function while denying users the participatory conditions that make genuine trust sustainable.

A practical example in insurance

These patterns are recognizable in engagements. At one insurance client, for example, the stated problem was low research maturity. There were no standardized processes and no shared repositories. In addition, work was done ad hoc, and teams kept repeating studies that already existed.

Therefore, the response addressed the structures first, so the technology layer could follow. Initially, seventeen interviews across internal teams diagnosed where the effort was actually going. Next came templates, method guides, recruitment and compliance processes, and a single repository. The result was centralized knowledge with less duplicated work. Teams became able to run their own research, and insights reached leadership in a consistent format. Only afterwards was AI capability layered on that foundation.

Early interventions: where to begin

To start reducing these debts, five early actions make a difference:

• Audit where design currently participates in AI system decisions and map the gaps explicitly.
• Establish research findings as documented inputs with traceable relationships to product decisions.
• Give design a standing seat in the specification process for AI behavior, defaults and automation levels.
• Build design literacy at the leadership level, enough to make resource decisions with informed judgment.
• Define one measurable proxy for trust quality in current product data and track it quarterly.

What organizations miss when design is visual only

The invisible channels of communication

One of the most costly blind spots in immature design practices is treating design as a visual discipline. When they limit design to the graphical interface, organizations overlook the communicative work done by every other channel. Yet those are exactly the channels through which their AI products interact with users.

A conversational AI communicates through its words. However, it also communicates through its latency, the rhythm of its responses and the vocabulary it selects. Likewise, its tolerance for user ambiguity and the way it frames what it does not know send signals. Each of these channels shapes user behavior independently of the semantic content. Kress’s multimodal communication theory explains this structure: every mode of communication carries distinct meaning potential (Kress, 2010). Therefore, coherence across modes is a design responsibility. Organizations that have never structured this responsibility cannot deliver it by hiring more engineers or adding post-hoc design reviews.

The cost of communicative incoherence

In practice, the consequence appears as a specific pattern in AI product data. Systems that are technically capable and visually polished, but communicatively incoherent, lose engagement depth. In addition, task scope shrinks and verification of outputs increases. Gradually, users also withdraw the contextual detail that makes AI interaction genuinely valuable.

Recognizing the multimodal scope of design is, therefore, one of the earliest and highest-return structural changes. This is because it reframes what design governance covers. At the same time, it makes previously invisible decisions visible without requiring significant new investment.

The authority gap, the uncanny valley risk and structural maturity

Design as informant or as constraint

The distinction between design as informant and design as institutional constraint is the most consequential organizational decision in AI-enabled product development. An organization can have substantial design teams, mature visual systems and genuine investment in user research. Even so, it may still treat design as advisory to the decisions that actually determine how systems behave.

The relational uncanny valley

This authority gap produces a risk that goes beyond product quality. Sometimes design can shape the interface but not the behavioral architecture that positions the user. In that case, it lacks the standing to govern care-based trust and human agency at once. Consequently, organizations are liable to produce AI systems that simulate genuine relationship while denying its essential condition.

Researchers studying mind perception and social AI describe this as the relational equivalent of the uncanny valley (Gray & Wegner, 2012; Mori, 1970/2012). These systems are close enough to genuine care to attract attachment. However, they lack the reciprocal human agency that makes the relationship real. For the product, the consequence is the structural trust failure documented in AI adoption research. Users form attachment and then encounter a failure that feels disproportionately personal. After all, the relationship they believed they had was not what they were experiencing.

Design authority is the organizational condition that prevents this. That is because designing care and agency together, rather than trading them against each other, requires design standing.

The stages of design maturity

The progression toward design maturity is not binary. On the contrary, organizations move along a spectrum with recognizable markers. At the earliest stage, design has advisory status and limited research integration. By contrast, at an intermediate stage, design has documented authority over interaction behavior and participates in AI system specifications. Finally, at a mature stage, human experience works as a hard constraint in the development process, including the balance between care and agency. At this stage, design holds standing equivalent to engineering and product management in consequential decisions.

The McKinsey Design Index identifies four dimensions of design maturity (McKinsey & Company, 2018). According to the study, top performers do not stand out by team size or research investment alone. Rather, they differ because senior leadership treats design decisions as business decisions. That recognition is the structural condition from which the progression follows.

A discipline in motion: autonomy, reach and collaboration

Design that moves with the market

There is a dimension of design maturity that goes beyond the debt framework. It is particularly relevant for large consulting and technology organizations serving B2B markets. Indeed, design is not a discipline that reaches a steady state. Its legitimacy depends on continuous responsiveness to shifting consumer behavior, market context and technology. At scale, a design function that cannot move with the market has already fallen behind, regardless of headcount or tooling.

A cross-functional reach

In multi-practice organizations, design naturally occupies the layer where human need and consumer behavior live. That layer, in turn, cuts across all practice areas. Service Design and Design Thinking inform how organizations transform their operating models. Product Design, meanwhile, determines how those models create value for end users. Growth strategies, in addition, require an understanding of adoption, trust and behavioral motivation that is inherently design-informed.

These disciplines do not belong to a single practice area. Their value comes precisely from their reach across development, engineering, infrastructure, security, cloud, data and analytics, industry knowledge and market positioning. For this reason, design needs its own departmental autonomy. This is not because it supersedes other disciplines, but because its contribution is cross-functional by nature. Thus, confining it to a single reporting line severs the connections through which it creates value (Verganti et al., 2020).

Autonomy with a clear collaborative model

In this context, design autonomy means design has its own epistemic standing. In other words, it has the authority to define what counts as a human experience problem and what research is required to answer it. It also holds human experience considerations as constraints that engineering, data and infrastructure must work within. This includes the care-agency balance that determines whether trust is sustainable.

For this autonomy to work, the collaborative model must be clear. Each discipline owns its domain of authority. Moreover, it contributes to decisions in other areas through defined engagement protocols rather than ad hoc negotiation.

What design owns and leads

• Human experience quality across all touchpoints
• Behavioral and multimodal communication standards
• Research methodology and user insight authority
• Service Design and Design Thinking facilitation
• Product Design and growth experience strategy
• Human experience constraints in AI system design

Where design collaborates with adjacent disciplines

• Engineering: interaction feasibility and system behavior
• Data and analytics: experience metrics and feedback loops
• Infrastructure and security: experience within technical bounds
• Cloud: scalability of human-centered product architectures
• Industry knowledge: domain context for experience design
• Market placement: positioning grounded in human value

Maturity can be built

This model describes contribution protocols: where design leads, where it collaborates and where decisions are made together. Organizations that have made this explicit report shorter decision cycles and less rework. They also see higher cross-functional confidence in design’s contributions.

Moreover, maturity of this kind can be built rather than hired. Internally, over forty percent of people now use AI tooling regularly, and that number keeps growing. This use is spread across five regions, with capability rising at a comparable rate in each. The mechanism was not a central review function. In one region, for example, it was an ambassador structure that grew from three squads to about ten. There, the design discipline sat inside each squad rather than assessing output from outside. Where capability has already diffused, governance has to grow into the places where the work happens.

The investment case and the recovery profile

The financial evidence

The financial evidence for design maturity as a business discipline is now substantial. Over five years, companies in the top quartile of design maturity outperformed industry benchmarks by 32 percentage points in revenue growth. In total returns to shareholders, the gap was 56 percentage points (McKinsey & Company, 2018).

Similarly, a Forrester Consulting economic impact study of enterprise design practice found a three-year ROI of 301 percent. That return came mainly from reduced development rework and faster delivery. In addition, design and alignment time fell by 75 percent, and design defects were cut by half (Forrester Consulting, 2018).

The hidden cost of change requests

That mechanism is also visible in our own delivery record. On engagements of eight to ten sprints with too little time for design research with users, we observe an average delay of two to four sprints. Afterwards, the change requests that follow add roughly twenty percent to the total project price.

However, this cost does not appear as an overrun. It arrives as a series of individually reasonable change requests, each justified by something learned late. As a result, the project absorbs it as ordinary scope evolution. An organization that has never added these requests up has never seen the figure. That is why the same underlying decision keeps being made.

Return within 12 to 18 months

These figures describe the destination. By contrast, the recovery profile for organizations starting from a design-adjacent position is encouraging. The return on early structural changes materializes within standard product cycles. Within 12 to 18 months of consistent application, exponential effects become visible in adoption data, support costs and rework reduction. In other words, organizations do not need full design maturity before the investment begins generating return.

Genuine trust and sustainability

Today, the sustainability dimension of this argument is also supported by data beyond design investment indices. Organizations that design for genuine trust combine care-based orientation with preserved human agency. As a result, they earn customer loyalty that lasts over time. That loyalty is rooted in genuine perceived value, not exit friction or engineered dependency.

Research on the financial distinction between genuine and coerced retention documents this contrast (Reichheld, 2003; Edelman, 2024). It shows up in Net Promoter Score data, long-term retention curves and brand trust metrics. In the first case, these metrics are recoverable; in the second, they are not. For organizations investing in AI at scale, therefore, design maturity determines whether AI produces sustainable commercial value or systematic trust debt. That is why the two investments must scale together.

References

• Auernhammer, J. (2020). Human-centered AI: The role of human-centered design research in AI development. DRS 2020.
• Edelman. (2024). Edelman Trust Barometer. Edelman.
• Forrester Consulting. (2018). The Total Economic Impact™ of IBM’s Design Thinking Practice. Study commissioned by IBM.
• Gray, K., & Wegner, D. M. (2012). Feeling robots and human zombies: Mind perception and the uncanny valley. Cognition, 125(1), 125–130.
• Kress, G. (2010). Multimodality: A Social Semiotic Approach to Contemporary Communication. Routledge.
• McKinsey & Company / McKinsey Design. (2018). The business value of design. McKinsey Quarterly.
• Mori, M. (1970/2012). The uncanny valley (K. F. MacDorman & N. Kageki, Trans.). IEEE Robotics & Automation Magazine, 19(2), 98–100.
• Reichheld, F. F. (2003). The one number you need to grow. Harvard Business Review, 81(12), 46–54.
• Shneiderman, B. (2022). Human-Centered AI. Oxford University Press.
• Verganti, R., Vendraminelli, L., & Iansiti, M. (2020). Innovation and design in the age of artificial intelligence. Journal of Product Innovation Management, 37(3), 212–227.