AI Pilot Failures: The Architecture Problem
AI
Generative AI
Agentic AI
Enterprise Architecture
Data Integration
API Security
Certinia
Platform Native
8
Architectural Reality
Media Hype
6/10
Real Impact
8/10
What is the Viqus Verdict?
We evaluate each news story based on its real impact versus its media hype to offer a clear and objective perspective.
AI Analysis:
While AI hype is still considerable, Malhotra’s grounded analysis forces a pragmatic recognition that architecture is the key determinant of AI success, which will ultimately have a significant long-term impact.
Article Summary
The prevailing excitement surrounding Generative and Agentic AI is rapidly fading as organizations grapple with the reality of underwhelming pilot programs. Certinia’s Raju Malhotra identifies the root cause: a fundamentally flawed architecture. Most businesses operate with a "Franken-stack" of disparate point solutions—CRM, project management, ERP—connected by brittle APIs. This architecture creates a ‘context gap’ – the AI agent lacks a complete, real-time view of the business, leading to inaccurate and confidently presented answers. The problem isn’t the intelligence of the AI itself, but the inability to access a single source of truth. Malhotra highlights that in fragmented environments, the AI agent might see the signed contract but not the resource shortage or revenue targets, resulting in ‘confident, plausible-sounding wrong answers’. Furthermore, this fragmentation creates a significant security risk, exposing sensitive data via numerous API connections. The solution, he argues, is a platform-native architecture, typically built on a common data model like Salesforce, that ensures agents have access to a unified, trusted view of the business. This approach eliminates translation layers, reduces latency, and strengthens security by consolidating data within a single system. Malhotra stresses that addressing the architectural problem is crucial before investing in AI, emphasizing that “fix the architecture, then curate the context.”Key Points
- The primary reason for AI pilot failures is not the AI models themselves, but a fragmented and disconnected enterprise architecture.
- AI agents require a single, unified source of truth – a platform-native architecture – to access real-time data and context.
- A fragmented architecture exposes organizations to significant security risks through numerous API connections.