How Will AI Literacy Help Us Get Value From AI?

Most financial services firms have already bought the tools.

Trading desks have AI-assisted analytics, operations teams have automated reconciliation aids, and generative AI assistants are available to staff across the business. Yet many of these initiatives stall well short of the value that was promised in the business case.

The reason is rarely the technology itself.

It is whether the people using it understand what it does, where it can be trusted, and when it needs to be questioned. That understanding is AI literacy, and it is the factor that determines whether AI investment turns into AI value or simply adds cost and risk.

This article sets out why literacy is a precondition for value, not a supporting activity to be addressed later.

Why AI investment doesn't automatically produce value

Deploying an AI tool is a straightforward commercial transaction. Realising value from it is not.

In practice, firms that roll out AI tools without preparing their workforce tend to see a familiar pattern. Some staff adopt the tool enthusiastically but apply it uncritically, trusting outputs they haven't been trained to evaluate. Others avoid it altogether, reverting to manual processes because they don't understand how it works or don't trust its conclusions. Both responses undermine the business case.

The result is inconsistent adoption, unrealised productivity gains and, in some cases, new operational or compliance exposure that wasn't present before the tool arrived. None of this shows up in a procurement evaluation. It only becomes visible once the tool is live and being used, or not used, in practice.

The missing ingredient in these cases is rarely better technology. It is a workforce equipped to use the technology appropriately.

How firms have traditionally approached AI rollout

Financial services firms have long experience rolling out new technology, and much of that experience shapes how AI is introduced today.

The common pattern is deploy first, train later, if at all. A tool is procured, a short launch briefing is delivered, and staff are expected to build proficiency through use. For conventional software, this often works well enough. The system behaves predictably, and errors are usually visible and correctable.

AI tools behave differently. Their outputs can be plausible and confidently presented even when wrong. Their reasoning is not always transparent. Their performance can vary by context in ways that are not obvious from a short demonstration. A one-off training session, or none at all, leaves staff unable to judge when the tool's output deserves confidence and when it doesn't.

This is not a criticism of past practice. It reflects the fact that AI tools place a different, ongoing demand on judgement than most technology firms have previously deployed. Traditional rollout approaches were not designed with that demand in mind.

Where literacy changes the equation

A literacy-led approach starts from a different premise: that understanding must be built deliberately, across roles, before or alongside deployment, rather than assumed to follow automatically from access.

This means staff understand, in terms relevant to their own role, what the AI tool is doing, what its known limitations and failure modes are, and when human review is required rather than optional. It means risk and compliance teams understand enough about how outputs are generated to oversee their use credibly. It means leaders understand enough to ask the right questions when reviewing performance.

Critically, this does not mean removing human judgement from the process. It means positioning judgement where it adds the most value, on exceptions, ambiguous cases and oversight, rather than on interpreting basic tool mechanics for the first time under operational pressure.

When literacy is built this way, adoption patterns change. Staff use tools consistently rather than sporadically. Outputs are applied appropriately rather than blindly trusted or reflexively ignored. Risk exposure becomes visible and manageable rather than hidden inside individual working habits.

What this means operationally for leaders

For leaders sponsoring AI initiatives, this has several practical implications.

First, literacy needs to be sequenced alongside deployment, not treated as a follow-up activity once a tool is already in production. Waiting until after go-live to address understanding means value is lost during the period when habits and trust in the tool are being formed.

Second, literacy requirements differ significantly by role. A trader's need to understand an AI-assisted pricing tool differs from an operations analyst's need to understand a reconciliation aid, which differs again from a compliance officer's need to oversee both. A single generic briefing rarely serves all of these needs well. This role-specific dimension is significant enough that it is addressed separately elsewhere in this knowledge base.

Third, this is an ongoing capability, not a one-off event. AI tools change, staff turn over, and regulatory expectations are evolving. The UK Government's AI Skills Compact reflects this shift, adding a reputational and regulatory dimension to what was previously treated as a purely commercial choice. Firms are increasingly expected to demonstrate that their workforce is equipped to use AI responsibly, not simply that they have deployed it.

Treating literacy as foundational, rather than incidental, is what allows AI investment to convert into sustained operational value.

Example

A London-based fixed income trading desk rolls out an AI-assisted trade reconciliation tool to reduce manual matching effort. Six months in, usage is patchy: some operations staff trust the tool's flagged exceptions without review, while others avoid it entirely, still reconciling manually because they don't understand how it reaches its conclusions. Neither behaviour realises the intended value.

After introducing a structured literacy programme covering how the tool works, its known failure modes, and when human review is mandatory, staff begin using it consistently and appropriately. Exception handling improves, reconciliation time falls, and the operations manager can demonstrate to risk committees that AI use is both effective and appropriately governed.

FAQs

• Is AI literacy the same as AI training?

  Not quite. AI literacy is the outcome: staff understanding what a given AI tool can and cannot do, where its risks lie, and how to apply judgement to its outputs. Training is one mechanism used to build that outcome, alongside others such as hands-on practice, coaching and embedding literacy into day-to-day processes. An organisation can deliver training without achieving genuine literacy if the training doesn't translate into applied understanding.

• How do we know if our organisation has an AI literacy gap?

  Several practical signs tend to indicate a gap. Usage of AI tools is inconsistent across otherwise similar teams. Some staff appear to trust outputs without questioning them, while others avoid the tool altogether. Risk or compliance teams raise concerns about how AI-generated outputs are being used. None of these require a formal assessment to notice, though a structured review can help quantify the gap once it's suspected.

• Does building AI literacy slow down AI adoption?

  It can appear to in the short term, since it takes time to build understanding rather than simply granting access to a tool. In practice, skipping this step tends to be a false economy. Firms that deploy without literacy often see slower, less consistent value realisation overall, along with elevated operational and compliance risk, once a tool is in live use.