What Baseline AI Literacy Should All Staff Have?

AI tools are now embedded in everyday financial services workflows — from customer service chat assistants to trading decision support and document summarisation.

Yet many organisations still treat AI literacy as a concern only for technical or innovation teams.

With regulators increasingly expecting demonstrable AI governance, and the government's AI Compact raising expectations around workforce readiness, firms need a clear, minimum standard of AI understanding that applies to everyone — not just those who build or configure AI systems.

This article sets out what that baseline should look like, why it applies across every role, and how it differs from more specialised, role-specific AI training.

Why every role now needs some AI literacy

AI no longer sits in a single department.

A customer service representative may never touch an AI model directly, but they might field questions from customers whose loan application was assessed with AI assistance. A compliance officer may not build AI systems, but they are expected to understand how those systems influence decisions they are required to oversee. A trader may rely on AI-supported analytics without ever adjusting a single parameter.

In each case, the member of staff is affected by AI outputs and decisions, even though they are not the one operating the technology.

This is the core reason baseline AI literacy cannot be limited to technical or innovation teams. Once AI influences a workflow, everyone downstream of that workflow needs enough understanding to interpret its outputs sensibly, recognise its limitations, and know what to do when something looks wrong.

How firms have traditionally approached staff training

Most financial services firms already run structured staff training programmes, typically built around three models.

Generic compliance e-learning covers broad regulatory obligations — data protection, financial crime, conduct risk — often through annual, one-size-fits-all modules.

Role-specific technical training targets the skills a particular function needs, such as underwriting criteria, trading systems or claims handling procedures.

Ad-hoc awareness sessions are run when a new tool or process is introduced, often as a one-off briefing rather than an ongoing programme.

Each of these models has served firms well for stable, well-defined topics. But AI does not fit neatly into any of them. It changes quickly, cuts across every function, and touches both technical and non-technical roles simultaneously. A generic compliance module is too shallow. A role-specific technical course is too narrow, reaching only the teams directly using the tool. And a one-off awareness session does not keep pace with how quickly AI use cases evolve inside the business.,

This gap is why organisations increasingly need a distinct, organisation-wide AI literacy baseline, separate from — but complementary to — these existing training models.

What a genuine AI literacy baseline looks like

A defensible baseline standard should give every member of staff a working understanding of four core areas.

What AI is and is not. Staff should understand, in plain terms, that AI systems identify patterns in data and generate outputs based on probability, not certainty or human-style reasoning. They should know that AI can be wrong, confidently so, and that its outputs require appropriate scrutiny.

Where AI is used in their organisation. Staff should know which processes they interact with — directly or indirectly — involve AI, whether that is a lending decision tool, a document summarisation assistant, or a fraud detection system.

The key risks AI introduces. This includes recognising the possibility of biased outcomes, fabricated or inaccurate outputs (often called hallucinations), inappropriate handling of sensitive data, and over-reliance on AI-generated outputs without human judgement.

When and how to escalate a concern. Perhaps most importantly, staff should know exactly who to contact if an AI-influenced output seems wrong, unfair or unclear, and understand that raising such concerns is expected, not discouraged.

This baseline is intentionally non-technical. It does not require staff to understand model architecture or data science. It requires them to understand consequences, boundaries and escalation paths — the same way staff are expected to understand financial crime red flags without being trained as investigators.

Operational considerations for rolling out a baseline standard

Establishing the baseline is only the first step. A few operational realities determine whether it remains effective over time.

The baseline must be refreshed as AI use cases evolve within the firm. A standard written for last year's chatbot pilot will not adequately cover this year's AI-assisted underwriting tool.

Different roles will need this baseline plus additional role-specific training. The baseline is a floor, not a ceiling — underwriters, compliance staff and technology teams will all need further, more specialised instruction layered on top.

Leadership and governance functions should model the same baseline expectations. If senior management treat AI literacy as something only front-line staff need, the initiative loses credibility and cultural weight. A baseline standard works best when it is visibly applied at every level of the organisation, reinforcing that this is a matter of culture and mindset as much as technical knowledge.

Example

A London-based retail bank rolls out an AI-assisted lending agreement approval tool. A customer service representative, who does not use the tool directly, receives a customer query about why their loan application was flagged for additional review.

Because the representative has baseline AI literacy, they understand that an AI system contributed to the initial risk flagging, know this is not a final decision, and know exactly which internal team to escalate the query to for human review.

The customer receives an accurate, confident explanation and a clear next step, rather than confusion or a mishandled escalation, because baseline AI literacy extended beyond the technical teams who built the lending tool.

FAQs

• Does every employee really need to understand AI, even non-technical staff?

  Yes. AI increasingly influences outputs and decisions across the business, even in workflows that non-technical staff manage or support. Staff who never configure an AI tool directly may still need to explain its outputs to a customer, recognise when something looks wrong, or know when to escalate a concern to a qualified colleague.

• How is baseline AI literacy different from AI training for specific roles?

  Baseline literacy is the shared minimum every member of staff should hold, regardless of function. Role-specific training builds on top of that baseline with skills relevant to a particular job, such as prompt techniques for underwriters or formal oversight duties for compliance staff.

• Is this a regulatory requirement?

  Specific rules vary by firm and activity, but regulators and government initiatives such as the AI Compact increasingly expect firms to demonstrate workforce AI competence as part of good governance. Baseline literacy is best treated as a practical necessity rather than a purely voluntary initiative.