Enterprises embracing AI-driven code generation and agents without proper governance may inadvertently create a new wave of legacy system challenges, says Appian’s VP Gregg Aldana.
- AI code generation speeds development but needs governance for compliance.
- Shadow AI causes accountability gaps and unmanaged technical debt.
- Hybrid human-AI workflows are key to balancing innovation and control.
What happened
Appian’s Gregg Aldana highlights a growing crisis where enterprises are rapidly adopting AI-enabled coding tools and agents without sufficient oversight. Dubbed ‘vibe coding,’ this practice involves prompting AI to generate software code quickly for prototypes and components. While accelerating development, it lacks the structured governance required for secure and auditable enterprise applications.
Why it matters
Regulated industries rely on systems that enforce business rules, maintain audit trails, and ensure security under stringent compliance regimes. AI-generated code that bypasses these controls risks replicating the very legacy problems companies have struggled for decades to resolve, undermining modernization efforts.
Shadow AI complicates accountability as employers cannot track who authorized or assessed AI agents’ actions. Without deterministic governance—defining agent roles, data access, escalation paths, and outcome documentation—organizations face unpredictable results and hidden costs that could balloon into significant technical debt.
What to watch next
Appian recommends integrating AI-generated code within a ‘deterministic harness’ that incorporates policy guardrails, strict change management, and human-in-the-loop oversight. Platforms that support hybrid workflows—where AI proposes and humans decide—are already in use at institutions like the University of Southern Florida and Acclaim Autism, demonstrating this balanced approach in procurement and care matching.
Enterprises should monitor developments in AI governance frameworks and invest in controlled adoption models. Creating clear policies defining AI agent boundaries and accountability mechanisms will be crucial to preventing runaway shadow AI practices and safeguarding regulatory compliance while capturing AI’s productivity benefits.