NonTechnical Debt in Software Systems
- NonTechnical Debt (NTD) is the accumulation of social, process, and cognitive deficiencies that impact software system evolution and agile practices.
- It encompasses subtypes like cognitive debt, intent debt, and Enterprise Architecture Debt, addressing gaps in shared understanding and organizational alignment.
- Empirical studies link NTD to diminished job satisfaction, communication breakdowns, and increased systemic risks, urging proactive management across development layers.
Non-Technical Debt (NTD) denotes the accumulated cost of neglected social, process, organizational, architectural, and knowledge-related deficiencies that surround software systems but are not reducible to code-level shortcuts. In agile software development, it is defined as the umbrella term for the non-code factors that accumulate as “debt”: inefficient processes, unhealthy social dynamics, people-related issues, and rigid organizational structures (Ahmad et al., 1 Sep 2025). In AI-assisted software development, the same problem space is captured primarily by cognitive debt and intent debt, which, together with classic technical debt, form a triple-debt model of software health (Storey, 23 Mar 2026). At the enterprise level, the concept broadens further into Enterprise Architecture Debt (EA Debt), defined as a metric that depicts the deviation of the currently present state of an enterprise from a hypothetical ideal state (Hacks et al., 2019). Taken together, these formulations position NTD as a socio-technical construct concerned with shared understanding, externalized rationale, workflows, relationships, staffing, governance, and business-IT alignment.
1. Definition and conceptual scope
NTD is presented in the literature as a complement to technical debt rather than a replacement for it. Technical debt lives in the product—code, tests, architecture—whereas NTD lives in the system that builds and evolves the product: teams, processes, organizational structures, and knowledge artifacts (Ahmad et al., 1 Sep 2025). This distinction is explicit in agile settings, where NTD is described as “treacherous” because it is less visible than bad code, yet can be just as damaging.
A second formulation emerges from the triple-debt model. That model distinguishes three layers of a software system: goals and intent, code and structure, and shared understanding. Corresponding to these layers are technical debt, intent debt, and cognitive debt (Storey, 23 Mar 2026). From the perspective of NTD, the latter two are non-technical because they concern human cognition, shared knowledge, and externalized rationale rather than code structure. The conceptual notation given in the source is for technical debt, for cognitive debt, and for intent debt, with software health depending on all three.
A third formulation appears at enterprise scale. EA Debt is defined as the deviation of the current enterprise architecture from a hypothetical ideal state constructed from EA quality criteria, stakeholder goals, and best practices (Hacks et al., 2019). This extends the debt metaphor beyond software artifacts to business architecture, process architecture, software architecture and used services, technology architecture, stakeholders, and guidelines and standards.
These formulations are not identical taxonomies, but they are compatible. A plausible implication is that NTD functions as a family resemblance concept: in one setting it denotes process, social, people, and organizational debt; in another it denotes cognitive and intent debt; at enterprise scale it denotes architectural and alignment debt. Across all three, the common property is debt in the socio-technical environment rather than in code alone.
2. Principal forms and taxonomies
The literature identifies several recurring forms of NTD.
| Source | Debt forms | Scope |
|---|---|---|
| (Storey, 23 Mar 2026) | Cognitive debt, intent debt | Team understanding and externalized rationale |
| (Ahmad et al., 1 Sep 2025) | Process Debt, Social Debt, People Debt, Organizational Debt | Agile software development environments |
| (Hacks et al., 2019) | Enterprise Architecture Debt | Enterprise-level architecture and business-IT alignment |
In the NODLA formulation, total NTD is presented conceptually as an aggregation of four interrelated forms: $\text{NTD}_{\text{total} = \text{PD} + \text{SD} + \text{PeD} + \text{OD}$ This is stated not as a formal measurement equation but as a conceptual aggregation (Ahmad et al., 1 Sep 2025). Process Debt (PD) is debt in workflows and ways-of-working. Social Debt (SD) concerns relationships, communication, and team culture. People Debt (PeD) concerns skills, staffing, and human resource practices. Organizational Debt (OD) concerns structure, governance, and policies.
The triple-debt model introduces a different cut. Cognitive debt refers to inadequate shared mental models that allow developers across a team to reason about a system and what they need to understand to change it safely and confidently. Intent debt refers to the absence or erosion of explicit rationale, goals, and constraints that guide how a system evolves. These debts occupy, respectively, the layer of shared understanding and the layer of goals and intent (Storey, 23 Mar 2026).
EA Debt is not introduced as a named subtype of NTD in its original paper, but it is explicitly positioned as a debt construct that goes beyond technological aspects to include business aspects (Hacks et al., 2019). It covers misalignment and quality problems across business, process, application, and technology layers, as well as stakeholder and standards dimensions.
These taxonomies overlap in emphasis rather than in labels. For example, organizational debt and EA Debt both address governance and structure, while cognitive and social debt both concern distributed understanding and communication. This suggests that NTD is best understood as a multi-level construct whose manifestations depend on analytical granularity: team, project, or enterprise.
3. Cognitive debt and intent debt
Cognitive debt is defined as inadequate shared mental models that allow developers across a team to reason about a system and what they need to understand to change it safely and confidently (Storey, 23 Mar 2026). It is explicitly team-level rather than merely individual, and it accumulates when shared understanding erodes faster than it is rebuilt. Developers experience it as loss of control or confidence.
The paper identifies several causes and mechanisms. Historically, cognitive debt appears in legacy systems no one fully understands and in pasted libraries or Stack Overflow snippets adopted without comprehension. In AI-assisted development, the rate and scale change: AI generates code faster than humans can digest, developers accept AI output without forming strong mental models, and over time this creates “an accumulation of not knowing” (Storey, 23 Mar 2026). The underlying mechanism is described as cognitive surrender, distinguished from ordinary cognitive offloading. Cognitive offloading is rational use of tools for discrete tasks; cognitive surrender is adopting AI outputs with minimal scrutiny, bypassing both intuitive and deliberate reasoning.
Observable proxies for cognitive debt include resistance to change, unexpected results, low bus factor, burnout and stress, and slow or unpredictable onboarding (Storey, 23 Mar 2026). These patterns are not introduced as a formal metric, but they are presented as practically useful features. The effects are substantial: reduced safety of change, more incidents, longer investigation times, and increased downstream technical debt because developers who do not understand the system make poor implementation decisions.
Intent debt is defined as the absence or erosion of explicit rationale, goals, and constraints that guide how a system evolves (Storey, 23 Mar 2026). It concerns “why,” not merely “what.” The relevant artifacts include requirements documents, Architectural Decision Records (ADRs), domain models and ubiquitous language, implementation plans and roadmaps, acceptance tests and BDD specs, meeting notes, decision logs, and context engineering artifacts for AI agents. Intent debt accumulates when requirements drift and are not re-captured, when decisions and trade-offs are never documented, when constraints are known only to a few people and gradually forgotten, or when AI is used to generate code without intent-first prompts and artifacts.
The paper emphasizes that intent is best captured at the moment decisions are made; reconstructing it later is often impossible (Storey, 23 Mar 2026). Unlike technical debt, which can be refactored later, lost intent cannot be reliably recovered from code alone. Practical signals include behaviour drift, AI agents struggling to make changes, and loss of articulated constraints such as performance, privacy, and accessibility requirements.
Management practices for these debts are correspondingly knowledge-centric. For cognitive debt, the paper lists human-centric code review, pair programming, system walkthroughs, retrospectives and post-mortems, deliberate onboarding/offboarding, and reimplementation using AI as a way to rebuild understanding (Storey, 23 Mar 2026). For intent debt, it promotes executable intent through BDD specs and tests, decision and rationale records such as ADRs, Domain-Driven Design (DDD), and context engineering for AI-assisted development.
4. NTD in agile software development
In agile software development, NTD is organized into four forms: process, social, people, and organizational debt (Ahmad et al., 1 Sep 2025). Each form is defined through concrete symptoms and, in some cases, empirical associations.
Process Debt is the gap between how work should flow to support agility and how it actually flows. The paper identifies Process Unsuitability Debt, Synchronization Debt, Process Documentation Debt, Roles and Responsibility Debt, and Infrastructure Debt (process-related). Symptoms include chaotic sprint planning, overloaded meetings, teams waiting on other teams, ineffective or overloaded Scrum Masters or Product Owners, standards regularly deviated from because they are perceived as unhelpful, and missing tools for testing or integration. In a survey of 191 developers, Process Unsuitability and Roles Debt together explain 33.8% of the variance in job satisfaction, with regression paths and , respectively, and adjusted (Ahmad et al., 1 Sep 2025).
Social Debt is debt caused by suboptimal team and organizational relationships: poor communication, weak trust, siloed groups, and fragmented social structures. Symptoms include misunderstandings, rework, “us vs. them” mentality, remote or hybrid teams feeling isolated, and growing reliance on informal workarounds rather than clear shared understanding. The paper reports communication breakdowns leading to missed requirements and costly rework, and it presents a structural model in which Psychological Safety positively impacts Learning from Failure with $\beta_{\text{PS→LF} = 0.25, \quad p = 0.000, \quad R^2_{\text{LF} = 0.408}$, while High Quality Relationships have a stronger impact with ; Learning from Failure then contributes to Team Performance with $\beta_{\text{LF→TP} = 0.130, \quad p = 0.043, \quad R^2_{\text{TP} = 0.062}$ (Ahmad et al., 1 Sep 2025).
People Debt arises from inadequate attention to skills, capacity, and well-being. It includes poor training, delayed hiring, overburdened staff, and weak onboarding or career paths. Reported symptoms are skill gaps, inability to adopt new technologies, chronic overtime, burnout, turnover, and long ramp-up times for new hires. The paper describes a vicious cycle in which delayed hiring and turnover overburden remaining staff, morale drops, more departures occur, and quality suffers (Ahmad et al., 1 Sep 2025).
Organizational Debt is debt embedded in organizational structures, policies, and governance that are outdated or misaligned with agile goals. Symptoms include hierarchy that blocks autonomy, blame-heavy cultures, lack of support for cross-team learning, inflexible role definitions, and organizational silos. The paper states that OD “paralyzes even the most talented teams,” creating systemic barriers to excellence and reinforcing other NTD types (Ahmad et al., 1 Sep 2025).
The empirical models connect these categories to learning and performance. The conceptual chain is team structure → cohesion & learning environment → knowledge sharing & reflexivity → learning and performance (Ahmad et al., 1 Sep 2025). This model situates NTD not merely as friction but as a degradation of the conditions under which agile learning, adaptation, and delivery occur.
5. Enterprise Architecture Debt as enterprise-level NTD
Enterprise Architecture Debt extends the debt metaphor beyond code-level and project-level concerns to enterprise-wide structure and alignment (Hacks et al., 2019). It is formally defined as a metric that depicts the deviation of the currently present state of an enterprise from a hypothetical ideal state. This definition emphasizes three points: EA Debt is framed as a metric; it captures deviation between as-is and ideal to-be states; and it aggregates across multiple artifacts and layers with weighting.
The debt spans explicit EA layers: Business Architecture, Process Architecture, Software Architecture and used services (SaaS), Technology Architecture (hardware), Stakeholders, and Guidelines and Standards (Hacks et al., 2019). These dimensions are non-technical in the sense that they concern business-IT alignment, organizational structures and governance, process design and coherence, and stakeholder communication and expectations.
The paper’s causal chain runs from constraints and uncertainty during EA initiatives to compromised EA decisions, then to deviations from the target architecture, lower EA quality, and increased business risk (Hacks et al., 2019). These risks include operational risk, strategic risk, security/compliance risk, and financial risk. The distinction between potential EA Debt and effective EA Debt, adapted from Schmid, separates structural issues in the EA from their actual impact on evolution and development.
Two scenarios illustrate the construct. In the first, an insurance company migrating roughly 300 mainframe applications to a cloud environment faces unforeseeable delays in one application. Another application is therefore implemented to rely on legacy mainframe interfaces rather than target cloud interfaces. This is classified as necessary, deliberate prudent EA Debt because the resulting architecture is temporarily misaligned with the target landscape (Hacks et al., 2019). In the second, the security operations team decides not to fix a minor security issue on web servers because remediation requires significant effort. This lowers the quality of the technology architecture and exposes the enterprise to risk; it exemplifies unnecessary or avoidable EA Debt.
Although the paper does not give a concrete numeric formula, it states that each artifact and every part an artifact consists of should be weighted, and that there is no uniform weighting function (Hacks et al., 2019). It points to EAQF, application evaluation metrics, business-IT misalignment detection frameworks, and Enterprise Coherence Assessment as possible bases for identifying indicators of EA Debt. This makes EA Debt one of the clearest formulations of NTD at the level of enterprise governance and transformation rather than software module design.
6. Interactions, diagnostics, and management
The various NTD formulations converge on the claim that non-technical debts are mutually reinforcing and closely coupled to technical debt. In the triple-debt model, intent debt increases cognitive debt because unclear purpose, goals, and constraints prevent accurate mental models; cognitive debt increases intent debt because developers who do not understand the system and domain cannot externalize good specs or rationale; cognitive debt increases technical debt because poor understanding leads to kludges, hacks, and suboptimal designs; and technical debt increases cognitive debt because messy code is harder to understand (Storey, 23 Mar 2026).
The agile NTD formulation makes an analogous claim at organizational level. Process Debt can strain relationships and become Social Debt; People Debt can harden into Organizational Debt; Social Debt can degrade workflows and become Process Debt (Ahmad et al., 1 Sep 2025). The paper identifies archetypal patterns such as “process-heavy, people-light,” “hero culture,” and “siloed scale” organizations. These are not universal categories, but they illustrate recurrent interdependencies among NTD types.
Diagnostic practices are therefore necessarily multi-layered. For cognitive debt, observable proxies include resistance to change, unexpected results, low bus factor, burnout and stress, and slow or unpredictable onboarding (Storey, 23 Mar 2026). For intent debt, practical signals include behaviour drift, AI agents requiring extensive clarification, and loss of articulated constraints. For agile NTD, the paper recommends regular surveys of psychological safety, job satisfaction, perceived process suitability, role clarity, knowledge sharing frequency, and social connectedness, combined with retrospectives and leadership workshops (Ahmad et al., 1 Sep 2025). For EA Debt, diagnosis proceeds through identification of invisible artifacts, misalignment symptoms, stagnating artifacts, deviations from standards, and incoherence across enterprise layers (Hacks et al., 2019).
Management follows the same multi-level logic. In AI-assisted development, recommended practices include treating understanding as a deliverable, using intent-first workflows, resisting automation of understanding, and monitoring the three layers in tandem (Storey, 23 Mar 2026). In agile contexts, the paper recommends clarifying roles and responsibilities, improving workflows iteratively, visualizing and streamlining work, optimizing meetings, building a shared vision, fostering remote-friendly collaboration, investing in continuous learning, accelerating hiring and staffing, strengthening onboarding and mentorship, refining team structures, introducing supportive roles such as Product Guardians and Security Masters, promoting autonomy with accountability, and cultivating psychological safety and learning culture (Ahmad et al., 1 Sep 2025). At enterprise scale, EA Debt is positioned as a common basis for communication and discussion between business and IT stakeholders, a way to make invisible elements visible, and a basis for identification, prioritization, remediation, and monitoring (Hacks et al., 2019).
Across these settings, the common management principle is that software health cannot be assessed solely by static code metrics. It depends on the alignment and adequacy of code, shared understanding, externalized intent, workflows, relationships, staffing, governance, and enterprise architecture.
7. Significance in the age of AI
The most explicit claim about the contemporary importance of NTD appears in the triple-debt model: AI redistributes software risk, shifting emphasis from technical to non-technical debts (Storey, 23 Mar 2026). AI tools can refactor code, generate tests, and detect code smells, which can reduce 0 or make it easier to manage. At the same time, AI can increase cognitive and intent debt because code is generated faster than developers can read or reason about it, prompts are often underspecified, and AI-generated documentation may reflect what code does rather than why it exists.
The resulting risk is stated directly: generative AI may reduce technical debt while simultaneously accelerating the accumulation of cognitive and intent debt (Storey, 23 Mar 2026). A system can therefore be technically healthy in the narrow sense of clean, refactored code and still be operationally unsafe because developers do not understand it and its reasons and constraints are not captured.
The agile NTD literature anticipates related pressures. It identifies future research directions on how generative AI influences social sustainability and NTD in large-scale agile teams (Ahmad et al., 1 Sep 2025). This suggests continuity between the AI-centered triple-debt account and the broader socio-organizational account of NTD: if code-writing capacity rises faster than organizational learning, shared understanding, role clarity, or intent capture, non-technical debts become the bottleneck.
At enterprise scale, the same logic applies to business-IT alignment and transformation risk. EA Debt is evaluated with respect to future evolution, and the ideal is not debt elimination but continuous debt management (Hacks et al., 2019). This suggests that, in AI-intensive environments, NTD is not a peripheral concern but a central mechanism through which speed, misalignment, and opacity become long-term liabilities.
In this sense, NTD names a shift in software engineering focus. Where traditional debt discourse emphasized maintainability and code quality, NTD emphasizes the conditions under which systems remain changeable, understandable, aligned with purpose, and governable across teams and enterprises.
