Technical Debt in the AI Coding Era - By Antonio Bianco

In Technical Debt in the AI Coding Era, Antonio Bianco, VP of Engineering at StubHub, explores how AI-assisted development is accelerating code production and with it, a wave of new technical debt. The talk covers how tools like AI code assistants, while boosting productivity, often introduce hidden costs: bloated code, security vulnerabilities, and architectural misalignment. Antonio also outlines strategic approaches for engineering leaders to mitigate these effects through better governance, developer education, quality-first workflows, and ongoing measurement.

1. Technical Debt
in the AI Coding Era
The risk of unchecked AI-accelerated development
and its long-term consequences
Antonio Bianco - August 28th, 2025 - CTO Talks - Orange County, CA

2. Key Insight
Never seen so much technical debt being created in
such a short period
"
"
— Kin Lane

API evangelist with 35 years of technology experience

kinlane.com

3. Technical Debt
Metaphors
Financial Debt

●Borrow speed
●Compound interest
●Intentional loan
●Bankruptcy
City Pollution

●Introduced slowly
●Affects your neighbors
●Hard to measure
●Prevent vs. remediate
Wood Shop

●Wood chips, misplaced tools
●Unpleasant work env
●Slower output, lower quality
●Clean up as you go

4. Fowler Quadrant
Reckless
Deliberate:
●Intentional
●Disregard consequences
●No plans to repay

Inadvertent:
●Unaware of accrual
●High impact on quality
and cost

Reckless
Deliberate
Prudent
Deliberate
Reckless
Inadvertent
Prudent
Inadvertent
Prudent
Deliberate:
●Purposeful
●Trade-offs
●Backlog and cleanup plans

Inadvertent:
●Manifests later
●Concern about impact
●Investment plans

5. AI coding
assistants
are here!

6. AI Coding Tools
Market Overview
Capabilities
Code completion & generation
troubleshooting, debugging
code reviews,
integrated (IDEs, repos, etc.)
Popular Products
Gemini CLI
Claude Code
GitHub Copilot
OpenAI ChatGPT
Amazon Q

Cursor
Windsurf
Cline
Lovable
ReplIt

84%
Software engineers regularly
use AI coding assistants
68%
Software engineers think AI is
boosting their productivity

7. Coding Productivity
Benefits
●+56% coding speed
●+21% task completion
●+98% more PR merges
https://www.faros.ai/blog/ai-software-engineering

8. But…

10. AI Productivity Paradox
Insights
●+154% PR sizes
●+91% PR review time
●-19% dev task speed
●+7.2% instability
https://www.faros.ai/blog/ai-software-engineering

12. AI-coding introduces a
significant amount of
tech debt. How?

13. Major Debt Accrual
1Code Bloat & Duplication
AI tools violate DRY principles generating redundant
code blocks instead of reusing existing functions

14. Bloated Code
Efficiency Problems
Verbose Implementation
AI prioritizes working solutions over optimal ones creating
unnecessarily complex and verbose code that increases
cognitive load for developers
Maintenance Burden
More code lines create maintenance overhead diverting
resources from innovation to upkeep

15. Code Duplication
8 x
Increase in large duplicate
code blocks
https://www.gitclear.com/ai_assistant_code_quality_2025_research

16. 8%
6%
4%
2%
2023202220212020
Percent of Commits with Dup Code
2024

17. Major Debt Accrual
1Code Bloat & Duplication
AI tools violate DRY principles generating redundant
code blocks instead of reusing existing functions
2Security Vulnerability
45% of AI-generated code contains security flaws
including SQL injection and cross-site scripting

18. Security Analysis
vulnerabilities in AI-generated code
Language/Type Vulnerability Rate Critical Issues Risk Level Mitigation Priority
Overall AI Code 45% Multiple High Immediate
Java Language 70% Buffer overflows Critical Urgent
Cross-site scripting 86% XSS attacks Critical Urgent
Log injection 88% Data exposure Critical Urgent
https://www.veracode.com/blog/genai-code-security-report/

19. Major Debt Accrual
1Code Bloat & Duplication
AI tools violate DRY principles generating redundant
code blocks instead of reusing existing functions
2Security Vulnerability
45% of AI-generated code contains security flaws
including SQL injection and cross-site scripting
3Architectural Misalignment
Limited context windows prevent AI from
understanding system architecture and integration
requirements

20. Architectural Misalignment
Two Reasons:
1.Documentation does not
match production architecture

2.AI is not great with large
context (local optimal)

21. Long-Term Consequences
40%
IT budget consumed by
maintenance activities
Organizations with high technical debt allocate
substantial resources to maintenance over
innovation
2T
HFS Research estimates Global 2000 companies
face unprecedented technical debt burden
requiring immediate action
Trillion dollars in
accumulated technical debt

22. As an exec, how can I
prevent & mitigate?

23. Recipe
keep AI tech debt
under control

24. 1. Have a strategy

25. Don’t be this leader

26. 1. Have a strategy





➔Structural approach
➔Cultural shift
➔Leadership commitment
➔Communicate

27. 2. Education

28. I’ve watched brilliant young programmers become
increasingly dependent on AI suggestions. When
the tools fail or provide suboptimal solutions, they
struggle to think independently.
"
"
— James Wilson

Veteran Software Architect at Microsoft

29. Skill Crisis
Skill Atrophy
➢Developers gradually lose core coding abilities
through over-reliance on AI assistance
➢Teams develop superficial understanding of code
➢Developer role is changing
Accountability
➢Developers blame AI for outages rather than taking
ownership responsibility
➢Traditional accountability chains become blurred
with AI code generation

30. 2. Education





2. Education
➔Guidance & best practices
➔Prompt & Context Engineering
➔Understand AI behaviour
➔Treat AI code as draft
➔Security threats

31. 3. Quality First

33. 3. Quality First




3. Quality First
➔Automated testing
➔Static analysis (AI?)
➔Security scans
➔AI auto-fixing

34. 4. Measurement

36. 4. Measurement






4. Measurement
➔Code health
➔Reliability
➔Coding vs. Review times
➔AI vs. human-generated
➔Maintenance effort (debug, refactoring)
➔Surveys

37. Tech Debt @ Google
Types identified by engs:
Q: Top 3 types of debt
hindering you?
●Migrations
●Documentation
●Testing
●Code quality
●Dead code
●Dependencies
●Team expertise
●Process
Tech Debt Management
Form TD coalition to
drive change: metrics,
tooling, best practices
Framework:
●Identify
●Measure
●Prioritize
●Resolve
●Control
https://ieeexplore.ieee.org/document/10109339
Measurement
●Automated checks
●Self-reported
●Hindrance surveys
●Process maturity
Impact
●30% less hindered
by technical debt
and complexity

38. Thanks for Watching
Continue the Journey
The challenge of managing AI-generated technical debt requires ongoing vigilance,
strategic thinking, and commitment to software excellence. We can harness AI's
transformative potential while preserving the quality, security, and sustainability
that defines great software engineering.

39. Strategic Framework for Mitigation
Governance Approach
Organizations must adopt comprehensive strategies to harness
AI benefits while mitigating technical debt accumulation
Clear Guidelines Quality Gates Measurement Education

40. Mitigation Strategies
Human-AI Collaboration
Reimagine development workflows with
effective human-AI pair programming
approaches
Quality Assurance
Strengthen testing pipelines and automated
security validation for AI-generated code
Continuous Learning
Cultivate accountability culture with ongoing
education and skill development programs

41. Implementation Roadmap Analytics
Strategic Implementation Framework
Successful AI debt mitigation requires phased implementation with clear metrics, governance structures, and continuous monitoring
to balance productivity gains with long-term software quality and organizational sustainability.
Phase 1
Governance Setup
Guidelines and
Training
Phase 2
Quality Gates
Enhanced
Pipelines
Phase 3
Monitoring
Continuous
Assessment

42. Balanced AI Integration Strategy
Strategic Balance
Harness transformative AI
potential while safeguarding
software quality
Success Factors
Organizations achieving successful AI integration balance immediate productivity gains with
long-term software health through robust governance frameworks, enhanced quality assurance
pipelines, and continuous developer education programs that maintain human expertise while
leveraging AI capabilities
Quality First
Prioritize software quality over quantity of code
generated
Human Oversight
Maintain developer expertise
and accountability in AI-assisted
workflows

43. Maintenance Overhead Impact
Resource Allocation
Technical debt forces disproportionate IT
budget allocation toward maintenance
instead of innovation initiatives
Developers spend time fixing
AI-generated bugs
Code cleanup dominates
development workloads
increasingly
Strategic projects delayed by
maintenance burden
Innovation Drag
Organizations lose competitive advantage
as resources shift from value-generating
activities to debt management
New feature development slows
significantly over time
Strategic initiatives postponed for
technical cleanup efforts
Market responsiveness decreases
due to maintenance overhead


Development Cycles
Long-term maintenance burden creates
cascading effects on development velocity
and team productivity
Defect remediation becomes
primary developer activity
Quality assurance efforts multiply
exponentially
Release cycles extend due to
stability concerns

44. Risk of Vibe Coding
Skill Atrophy
Developers gradually lose core coding abilities through over-reliance on AI assistance
Expertise Illusion
Teams develop superficial understanding of
code they supposedly wrote but cannot
explain or debug effectively when problems
arise
Junior developers use AI as crutch performing
at high levels initially but lacking foundational
experience to identify vulnerabilities
Professional Risk
Pipeline of deeply skilled engineers becomes
compromised for future senior roles

45. Key Takeaways and Recommendations
Strategic Imperatives
Engineering leadership must recognize that AI coding tools
present both unprecedented opportunities and significant
risks. The rapid accumulation of technical debt through
AI-generated code demands immediate attention to
governance frameworks, quality assurance processes, and
developer skill preservation. Organizations that proactively
address these challenges will realize sustainable benefits
while maintaining software quality and security.
Strategic
Governance Required
Quality
Gates Essential