top of page

Jeet Poptani
Enterprise AI Advisory · AumentoAI

AI Doesn't Work in a Silo: The Digital Foundations Most Leaders Overlook

Jeet Poptani
5 days ago
3 min read

Most boardroom conversations about AI start with the model: which one, which vendor, how clever it is. In my experience delivering ERP and AI programmes, that is rarely where success or failure is decided. AI is not a standalone product you bolt on. It sits on top of data, applications, integrations and processes, and it is only as useful as the foundation underneath it.

Put simply, an AI initiative that cannot reach clean data, cannot trigger a workflow and cannot write back into the systems people use every day will stay a demo.

A shared vocabulary for business and technology

Leaders do not need to become technologists, but they do need a common language with their technology teams. Every digital estate does a small number of things:

  • Compute. Processing information and running business logic.

  • Interact. Web and mobile screens through which people work.

  • Connect and exchange. Linking systems, devices and partners, and moving data between them.

  • Store and capture. Holding records and collecting transactions, sensor readings and user input.

  • Automate. Executing workflows with less manual effort.

  • Govern. Controlling access, securing identity and keeping data usable.

AI draws value from how well these capabilities are joined up with real business processes, not from any one of them on its own.

Six questions to ask about any technology

Before approving any AI or digital investment, I encourage executives to work through six lenses:

  • Purpose. Why do we need it, and what value will it create?

  • Understanding. What does it actually do, and how mature is it?

  • Enterprise fit. How will it integrate with our existing systems, data and APIs?

  • Readiness. Which skills, owners and governance are required?

  • Risk. What are the privacy, security, compliance and change-management issues?

  • Impact. Who is affected, how will work change, and how will we measure success?

A simple habit pays off here: keep a technology portfolio that records purpose, owners, integrations, maturity, risks and value metrics for each tool. It turns scattered experiments into a managed estate.

Architecture is shifting under our feet

Enterprise architecture has moved from mainframes to client-server, web, cloud, microservices and event-driven designs. Agentic AI is the next shift, and it changes some long-held assumptions:

  • Components that reason, rather than only execute pre-written logic.

  • Agents that discover each other's capabilities instead of relying on fixed API contracts.

  • New protocols for agents to call tools and talk to one another.

  • Memory, short and long term, becoming a design concern in its own right.

  • Event-driven, loosely coupled design mattering even more than before.

From digitisation to autonomy

A useful way to locate your organisation is a five-step progression: digitisation of records, automation of workflows, connectivity through cloud, APIs and IoT, intelligence through analytics and machine learning, and finally autonomy, where processes adapt under human supervision. You cannot leap to step five without the first four in place.

AI cannot be implemented effectively in silos. Its value depends on data, connected systems and the workflows it is allowed to change.

What this means for your next AI decision

Before funding the next pilot, ask a plainer question: are our data, systems and processes ready to let AI act? If the honest answer is "not yet", the best AI investment you can make this quarter may be in the foundations.

Key terms

  • AI readiness: how prepared an organisation's data, systems, processes, skills and governance are to adopt AI at scale.

  • Digital foundations: the data, applications, integrations, cloud and security capabilities that AI depends on.

  • Enterprise architecture: the blueprint of how an organisation's systems, data and processes fit together.

  • API (application programming interface): a standard way for software systems to exchange data and trigger actions.

  • Microservices: small, independent services that each perform one business function and connect through APIs.

  • Event-driven architecture: systems that react in real time to business events, such as a new order or a status change.

  • Model Context Protocol (MCP) and agent-to-agent (A2A): emerging protocols that let AI agents call tools and work with other agents.

  • Technology portfolio: a managed register of each technology's purpose, owners, integrations, risks and value metrics.

Related searches: is my company ready for AI; AI readiness assessment; data foundations for AI; why AI fails in silos; AI architecture for leaders; preparing an organisation for AI

Comments


bottom of page