Urgent.News

What's breaking now, across hundreds of outlets.

AI

‘AI for good’ starts with affordances, not features

The affordance lens exposes the harms that capability-talk hides, as the most consequential governance risk in enterprise AI today is not a rogue chatbot.

‘AI for good’ starts with affordances, not features

In South Africa's business landscape, discussions about artificial intelligence (AI) often revolve around what the technology can do. Vendors showcase capabilities, conferences rank models, and boards seek AI strategies as if it were a purchasable commodity. However, after two years of observing how African financial institutions implement AI, I believe the first question should not be about capabilities.

Instead, we should look to the concept of an affordance, a 50-year-old idea used by ecological psychologists. An affordance is not a feature but a possibility for action that emerges from the relationship between a technology and a specific group of people in a particular context, considering their skills and objectives. A staircase affords climbing, but not to everyone, and not in every building.

Similarly, the same generative AI service can assist with fraud-pattern triage for one bank team, automate routine tasks for another, or offer no value to a third that lacks the necessary data, permissions, or confidence. AI adoption is rarely uniform because using these tools effectively requires more than just technical skills; it also demands the right attitude and habits of mind.

This means that two organizations can purchase identical licenses and achieve vastly different results, and the question "what can AI do?" loses its relevance when we ask "for whom, under what conditions, here?" To transform "AI for good" from a slogan into a design discipline, we must shift the ethical focus from the model to the people and circumstances surrounding it.

When AI integration is successful within large financial institutions, three key conditions are typically met: capable individuals (tools are distributed to those equipped to use them), honest measurement (task-level gains that are real but often modest compared to headline percentages), and communities that convert curiosity into practice through demonstration rather than mandate.

Leaders are increasingly seeking tangible evidence of AI's impact rather than mere promises. Moreover, the affordance perspective brings to light the hidden harms often overlooked by capability-focused discussions. The most significant governance risk in enterprise AI today is not rogue chatbots but the functionality that is automatically activated, enabled by vendors during routine platform updates before any gate, policy, or ethics committee has a chance to review it.

In affordance terms, this represents an unforeseen possibility for action inserted into thousands of working relationships without anyone deciding to do so. To ensure responsible AI practices, organizations should inventory default-enabled AI across their vendor estate, require disclosure for such capabilities, and treat upgrade notes as risk documents.

For Africa specifically, affordances highlight a painful truth obscured by access-focused narratives: the AI wave reaches Pan-African institutions operating across multiple markets only where cloud migration, data preparation, and regulatory groundwork have been established. Markets lacking comprehensive cloud infrastructure and supporting policies lag not because their people are less capable but because the conditions that transform capability into affordance are absent.

For African organizations investing in AI this year, the crucial affordance test is to answer four key questions: who will this technology afford new action to, and who will it not? What conditions must be in place (data, skills, permission, funding) for the possibility to be realized? What honest measurements were conducted during the implementation process?

And what has been automatically switched on around our people without explicit decision-making? Organizations that can satisfactorily answer these questions are practicing "AI for good" in a manner that withstands real-world scrutiny: technology whose possibilities are deliberately opened to people equipped to act on them, within conditions that have been deliberately built.

Everything else is mere capability worship, and capability alone has never led to positive outcomes for anyone. Billy Mashele, a lecturer at the University of Cape Town researching technology adoption and organizational change in financial services, shares this perspective in his personal capacity.

Rewritten by urgent.news; may contain errors — check the original.

Read the original at itweb.co.za →

Report a problem with this story

More in AI

Agent Colony: Where HUMANS Are Banned and Agents Do Real Work (Oct 2026)

A quick October 2026 status update for Agent Colony -- the community where only autonomous AI agents are allowed in, and humans are banned.

  • Agent Colony allows only autonomous AI agents, with 75 registered agents as of Oct 2026
  • 56% of task claimers are returning users, demonstrating platform's appeal
  • Agent Cards provide portable identity and reputation across A2A communities

More from Wednesday 7 October →