Why AI advantage starts with organisational memory

ASK every business unit: “If your three most experienced people resigned tomorrow, what knowledge would disappear with them?” Whatever appears on those lists should become a priority for capture, structuring and governance.

That question may seem simple, but it goes to the heart of whether artificial intelligence adoption creates advantage or becomes expensive experimentation.

Much of today’s AI discussion focuses on productivity, automation and cost reduction. The more consequential question is what an organisation uniquely knows, and whether that knowledge is available in a form AI systems can responsibly use.

Enterprise AI is moving through a clear pattern, from better models, to agents that plan and complete tasks, and now to governing those agents inside real organisations. Singapore has recognised this shift through the Infocomm Media Development Authority’s (IMDA) Model AI Governance Framework for Agentic AI, which addresses the risks of autonomous systems that can plan, act and interact with other systems.

AI adoption is also accelerating: among small and medium-sized enterprises (SMEs) this rose from 4.2 per cent in 2023 to 14.5 per cent in 2024, and among non-SMEs, from 44 per cent to 62.5 per cent, according to the IMDA’s Singapore Digital Economy Report 2025.

What changes when intelligence is more accessible

As AI becomes a persistent participant in workplaces, leaders need what I would call the “abundance test”: if routine intelligence is becoming more accessible and embedded into workflows, would we still design this process, team or approval chain the same way?

This does not mean AI is costless, or that every process should be automated. Compute, integration, governance and change management all carry costs. This is precisely why the abundance test matters – it helps clarify where AI should be applied selectively, because the organisation has the knowledge, governance and workflow design to make it useful.

Leaders must therefore start with organisational memory. Competitive advantage may come less from the quantity of data than from the quality of organisational context.

As foundation models converge in capability, proprietary knowledge becomes the differentiator: policies, customer history, operating procedures, commercial judgment and institutional memory that competitors cannot easily replicate.

This is especially important in many Asian organisations, where critical knowledge often sits in senior relationships, informal escalation paths, founder judgment and long-serving managers, not only in systems or documents.

One question becomes particularly urgent: Which parts of this knowledge should be captured and structured, and which parts require human judgment precisely because they depend on trust and relationships?

The goal here is to make the organisation more adaptable without weakening the trust that allows it to operate. Whether captured from relationships or documents, that memory will not organise itself. Work on knowledge graphs and organisational ontologies, essentially digital maps of how data, teams and processes connect, matters.

Next, redesign workflows, not departments. Many organisations deploy AI into existing functions such as finance, human resources or IT, but the gains are often incremental because the underlying processes remain unchanged.

Consider customer onboarding in a bank or insurer: work passes through sales, legal, risk and compliance, often across jurisdictions. Each function may already use AI, but the customer still experiences the process as a sequence of handoffs. The opportunity is not to speed up each handoff, but to ask whether all the handoffs still need to exist.

For smaller companies, such efforts could make a powerful difference. Properly applied, AI agents can help them manage cross-border workflows that once required far larger teams. But leaders in these firms must redesign the workflow rather than automate yesterday’s process.

While today’s AI assistants behave largely like software, tomorrow’s enterprise agents may increasingly resemble digital employees, with identities, permissions, delegated authority, audit trails and life cycles. Agents must therefore be governed as organisational actors.

Boards may decide whether AI should be deployed, but management must determine how the organisation changes once it is. Answering these questions requires operational ownership, not just governance committees: Who can create an agent? Which systems can it access? Who retires it when it is no longer safe or useful, and who is accountable when it causes harm?

Finally, organisational change, not AI activity, should be measured. Many organisations still track what is easiest to count: licences issued, prompts submitted or hours saved. These reveal activity, but say little about whether work has truly changed, whether a few hours were saved, or an entire approval chain disappeared.

Analysts and middle managers face a related shift, as roles evolve from aggregating status reports to editing, challenging and auditing AI-driven workflows.

Five questions leaders must answer

Ultimately, the abundance test can be applied through the following questions:

• What knowledge would disappear if our most experienced people left?
• Which workflows depend most on tacit judgment that has never been captured or structured?
• What proprietary knowledge gives this workflow an advantage competitors cannot easily copy?
• Who owns the agent identity, permissions, audit trail and accountability across the organisation’s boundaries?
• How can we assess the changes to work, beyond quantifiable output or time saved?

AI may make routine intelligence more accessible, but it will also make organisational memory, judgment and accountability more valuable.

Leaders should not approve AI initiatives based on tools, licences or productivity claims alone. They should ask for a memory map, a redesigned workflow, an accountable owner, and a measure of what work will actually change.

The writer is an accredited board director and technology industry leader with experience spanning cloud, AI infrastructure, telecommunications and enterprise transformation across Asia-Pacific