CIOs: Your AI Tech Stack Needs a New Look

Data management, AI applications, and risk mitigation are key parts of an AI tech stack.

1. Data management

Centralized and decentralized sources: Connect structured data from data lakes with unstructured data like emails, recordings and documents to allow you to derive insights from a broader range of data sources, enhancing decision-making processes. The ability to leverage unstructured data is crucial, as it represents an estimated 70%-90% of all enterprise data. 

Data accessibility and security: AI models must be trained on high-quality, relevant data. Implement sophisticated management practices, like setting appropriate access controls and ensuring compliance with data privacy regulations, to ensure data is both accessible and secure.

2. AI applications

Embedded AI: To quickly adopt AI without significant upfront investment in development, use AI features integrated into existing enterprise applications. As more vendors embed AI into their offerings, organizations can benefit from enhanced functionalities and improved efficiencies in their existing workflows.

Built AI: Developing AI solutions in-house offers greater customization and control. Proprietary AI models can also provide a competitive edge in the market.

Bring-your-own AI: Leverage AI applications procured by individual departments while ensuring that initiatives align with overall strategic goals and comply with governance frameworks.

Pinpoint high-impact AI opportunities with Gartner’s AI Use Case Insights for IT Leaders. Discover, evaluate, and prioritize AI opportunities to accelerate IT transformation and demonstrate value to the business.

3. TRiSM Layer

This layer acts as a safeguard, ensuring that AI deployments are secure and trustworthy. TRiSM technologies can prevent unauthorized access to sensitive data and filter out inappropriate or non-compliant AI outputs.

These TRiSM technologies also provide ongoing oversight and governance of AI outputs. Continuous monitoring allows you to detect and address issues promptly, maintaining the integrity, reliability and value of AI applications.

It’s the combination of three layers that defines an enterprise’s AI tech sandwich. Every organization’s will be unique. As a starting point, consider three:

Vendor-packaged sandwich: Ideal for midsize enterprises with limited AI resources, the vendor-package sandwich relies on embedded AI from software upgrades and bring-your-own AI capabilities. It includes a thin layer of TRiSM technologies to ensure vendor accountability.

TRiSM-rich sandwich: Well-suited for public-sector organizations or those in highly regulated industries, this sandwich features a robust TRiSM layer and centralized data management. To maintain compliance, it limits bring-your-own AI.

Deluxe sandwich: Designed for large enterprises with extensive AI ambitions, the deluxe sandwich combines built, embedded, and bring-your-own AI capabilities. It leverages comprehensive TRiSM technologies and diverse data sources.

By focusing on these components and archetypes, organizations can build a robust AI tech sandwich that supports their AI strategy and broader strategic objectives and drives innovation. As AI technologies continue to evolve, maintaining a flexible and scalable tech stack will be key to staying competitive.

The AI tech sandwich is a valuable communication and planning tool. It helps:

• Governance teams understand what’s ahead — and what they will need to govern.

• IT organization members get a better sense of how AI will be present in their areas.

• Senior business leaders ascertain the technology and governance elements required to execute AI.

Consider the tech sandwich as a conceptual framework that can fuel the creation and testing of plans, such as:

• An initial concept to start, test or finalize a formal technology stack or reference architecture.

• Determining how the democratization of AI should take place with centralized IT and data science team support.

• Anticipating the costs and resources required to execute AI in the enterprise.