What Investec Knows That Most Boards Have Not Yet Asked

Client value, not competitive pressure, is the only legitimate starting point for an AI strategy.

Sign up for my Substack daily AI newsletter here.

See my AI Training course portfolio for corporate Business Leaders here.

Article link: https://open.substack.com/pub/johanosteyn/p/what-investec-knows-that-most-boards?r=73gqa&utm_campaign=post&utm_medium=web&showWelcomeOnShare=true

Follow me on LinkedIn: https://www.linkedin.com/in/johanosteyn/

Graeme Lockley did not make headlines by announcing a major AI investment. He made headlines by declining to make one. As CIO of Investec Specialist Bank, Lockley told TechCentral in May 2026 that his institution has no intention of putting AI at the centre of its strategy. “We say human first. We are a human-first organisation.” While Capitec has deployed AI tools to nearly 5,000 employees and runs agentic AI inside its credit processing operations, and while Discovery Bank is spending at a pace that produced an operating loss of R155 million in six months building its Vitality AI division, Investec is doing something that looks, on the surface, like restraint. It is not restraint. It is strategy.

The distinction matters enormously, and it is one that most boards navigating their own AI decisions have not yet made.

CONTEXT AND BACKGROUND

South Africa’s banking sector has become, in 2026, one of the most instructive laboratories for AI strategy in the world, precisely because its major institutions have arrived at genuinely different positions from a shared starting point of competitive pressure, regulatory scrutiny, and customer expectation. The divergence is not the product of different levels of technical capability or different levels of access to AI infrastructure. It is the product of different answers to a prior question: what do our clients actually pay us for, and does AI serve that value or substitute for it?

Capitec's answer is unambiguous. The bank serves 23 million retail clients who expect speed, availability, consistency, and low cost. AI serves all of those things directly. Capitec has put AI tools in the hands of almost 5,000 employees and is running an agentic AI system inside its business banking credit processing operation, with its strategy focused on serving more clients with more products without proportional headcount growth.

Discovery’s answer is built on a different asset. The group draws behavioural and health data from more than 40 million Vitality members across more than 40 markets, and its AI strategy is premised on the argument that the data position created by that network is genuinely defensible and genuinely translatable into AI outputs that competitors cannot easily replicate. The R155 million operating loss recorded specifically within Discovery’s Vitality AI initiatives for the six months to end-December 2025 is not waste. It is the cost of building an infrastructure whose value is contingent on the data moat holding.

Investec’s answer is different from both. Its competitive position rests not on scale, not on data volume, but on the quality of the human relationships through which it serves high-net-worth individuals and specialist corporate clients who require bespoke judgment, deep contextual understanding, and the kind of accountability that only a named human being can provide. Lockley is explicit: “We don’t have the size and scale of FNB and Standard Bank and all of those. And so the way we differentiate is how we interact with our clients.” Critically, human-first does not mean AI-absent. Investec deploys AI aggressively in its backend security and fraud prevention operations, where AI-driven models have blocked fraudulent beneficiaries, prevented scam payments, and averted significant potential client losses — precisely the domains where AI serves the client relationship rather than replacing it.

INSIGHT AND ANALYSIS

The insight that makes the Investec position strategically coherent, rather than merely contrarian, is that AI cannot replicate what Investec’s clients are actually paying for. This is not a claim about AI’s limitations in general. It is a claim about a specific value proposition in a specific market segment. When a high-net-worth client engages a private banker, they are not buying access to analytical processing capability. They are buying judgment, trust, accountability, and the assurance that the person advising them has both the expertise and the professional obligation to act in their interest. Those are human qualities. They cannot be automated without ceasing to be the thing the client is buying.

Lockley’s warning about the savant trap is the sharpest observation in the TechCentral piece. “If you think of AI as the super smart savant, you are going to make some terrible choices. But if you see it as a tool that helps you in a decision-making process, then it is very, very useful.” This is not technophobia. It is a CIO’s precise assessment of where AI adoption goes wrong in organisations that have answered the wrong question first. The wrong question is “how can we use AI?” The right question is “what are our clients paying us for, and does AI serve that or replace it?” Organisations that answer the first question without having answered the second are not building an AI strategy. They are building an AI installation, and the two things are not the same.

I have previously written about the specific risks of AI in financial services client interactions, and the regulatory and trust dimensions of replacing human judgment with automated guidance in contexts where the client has a reasonable expectation of accountability. The Investec position extends that argument into the strategic domain: not only is there a regulatory and trust risk in replacing human judgment with AI in client-facing financial services, there is a strategic risk in doing so when the human judgment is the product being sold.

IMPLICATIONS

The lesson the Investec-Capitec-Discovery contrast offers to boards across industries and geographies is not that any of these strategies is universally correct. It is that all three are coherent, and they are coherent because each one is derived from a clear and honest account of what the organisation’s clients value and why they chose it. That account is the missing first step in the majority of AI strategy conversations happening in boardrooms today.

The pattern of AI strategy failure is well-documented. Organisations adopt AI tools because their competitors have, because a vendor's presentation was compelling, because a board member read a concerning statistic about being left behind, or because the technology team has been given an AI mandate without a business mandate to guide it. The result is deployment without purpose, governance without direction, and eventually the discovery that the tools produced outputs that nobody needed and disrupted processes that were working. Gartner has predicted that at least 30 per cent of generative AI projects will be abandoned after the proof-of-concept phase due to poor data quality, inadequate risk controls, escalating costs, or unclear business value — and that figure is not primarily a technology problem. It is a strategy problem.

For boards navigating AI strategy decisions, the Investec example offers a practical discipline. Before any AI investment is approved, three questions should be answered with specificity rather than aspiration. The first is what value your clients are paying you to create and whether it is human-generated or process-generated in nature. The second is whether AI serves that value directly, as it does at Capitec, or substitutes for it, which is the risk Investec is managing. The third is whether the AI advantage you are seeking is genuinely defensible, as Discovery’s data moat is intended to be, or whether you are investing in capability that any competitor with access to the same model can replicate in six months. Answering those questions honestly will produce different answers for different organisations. The point is not to arrive at a universal position. It is to arrive at a position that is grounded in something more durable than competitive anxiety.

Lockley’s comparison of current AI trust to the early days of Google search is also worth carrying into boardrooms beyond banking. Nobody fully trusted Google in 1999, but it was useful, and trust accumulated over time as its outputs proved reliable in specific contexts. The organisations deploying AI most effectively today are those that have been precise about which contexts have earned that trust within their own operations, and which have not. Investec has made that assessment and concluded that the high-stakes, relationship-intensive context of private banking is not one where AI has yet earned the authority that an AI-first strategy would effectively grant it. That is a judgment, not a prejudice, and it is one that every board should be making explicitly rather than deferring to competitive pressure.

CLOSING TAKEAWAY

The question that the Investec story puts back on the boardroom table is one that should have been asked before the first AI vendor was invited to present. Not “what can AI do?” or “what are our competitors doing with AI?” but “what do our clients pay us for, and does AI serve that or threaten it?” That question has no universal answer. At Capitec, the answer leads to AI tools in 5,000 employees’ hands and agentic systems processing credit applications. At Discovery, it leads to a nine-figure investment in a data-driven AI advantage. At Investec, it leads to a human-first declaration that is as strategically precise as any deployment decision its rivals have made.

The board that cannot answer the client value question clearly has no legitimate basis for an AI strategy. The board that can answer it clearly may find, like Investec, that the most strategically sophisticated position available to it is not the one that makes the most headlines.

Johan Steyn is a prominent AI thought leader, speaker, and author with a deep understanding of artificial intelligence’s impact on business and society. He is passionate about ethical AI development and its role in shaping a better future. Find out more about Johan’s work at https://www.aiforbusiness.net