A sales associate asks an AI product adviser which television fits a customer’s room. A service manager reviews an AI-written reply about a warranty. Upstairs, an executive decides whether an automated workflow can contact thousands of customers without a human reading every message.
All three employees are “using AI.” That’s about where the similarity ends.
Put them through the same annual awareness course and they’ll share some vocabulary. They still won’t know where their own authority stops, which mistakes matter most in their job, or when a doubtful answer needs another pair of eyes. For consumer-electronics businesses adding AI to sales, support and operations, that gap is becoming hard to ignore.
The job changes what AI literacy means
New EU rules make the question more urgent. The EU AI Act’s AI-literacy requirement has been in effect since February 2, 2025. As of August 2, 2026, national authorities are responsible for enforcing it. That doesn’t mean every TWICE reader falls under the same legal requirements. It does give companies a useful prompt: what should a person know before an AI system becomes part of the job?
The European Commission’s current AI literacy guidance doesn’t prescribe one fixed level for every employee. It points to technical knowledge, experience, education, training, the setting in which a system is used and the people affected by it.
That matches what the CE channel is already seeing. ProSource has discussed role-specific training for technicians, project managers and sales leaders. Different work calls for different preparation even before AI enters the picture.
Map the role before assigning the course
Start with the work, not the course catalog. Which AI system does the employee use? What decision does it support? What can the employee approve, and what has to be escalated? Who could be affected by a bad output?
The sales associate may need to know when a recommendation lacks enough information about the customer’s room, budget, existing equipment or the reason they’re hesitating in the first place. A support manager needs to spot invented warranty terms and decide when a reply requires legal or technical review. An executive needs visibility into who owns the workflow, how failures are reported and which decisions should never be handed over completely.
A practical role-specific competency management approach can document what each job requires, what evidence shows readiness and when the requirement should be reviewed again. Salespeople don’t need to become AI experts. They need to know what they can decide for themselves and when to call in someone else.
Teach the decision, then test it
Attendance is easy to count. Applied judgment takes more effort.
Give employees situations they might actually face. Ask the associate to handle an AI recommendation that conflicts with what the customer has said. Give the service manager a polished reply containing one false policy detail. Ask the executive to identify the owner and stop condition for an automated campaign.
Assessment should reflect the role. A manager might observe a task, review a scenario exercise or sign off on supervised work. The most important part is the connection between the assessment and the decision the person is allowed to make when a customer, employee or business process could be affected. A generic quiz about AI definitions won’t show whether somebody can recognize a risky output under pressure.
NIST’s AI Risk Management Framework Core supports this kind of division. It calls for clear roles and lines of communication, training that lets people perform their assigned duties, defined human oversight, and periodic review.
The human side matters here. TWICE has already examined how sales-floor AI can undermine an associate when it knows the catalog but misses the customer. Training shouldn’t teach the employee to recite the machine’s answer more confidently. It should help them recognize what the machine cannot see.
Refresh the map when the system changes
AI literacy expires faster than the certificate suggests. A tool that once drafted internal notes may later produce customer-facing recommendations. A vendor may change the model, connect a new data source, or add an automated action. Any of those changes can alter the knowledge and supervision a role needs.
Set review triggers around the work: a new system, a new use, a new audience, or a change in decision authority. Then check whether the employee can still use the tool within its limits and knows where to take an uncertain result. Reassessment doesn’t have to mean another hour-long course. Sometimes a short scenario and a manager conversation will tell you more.
A company-wide course can cover the basics. From there, training should reflect what people actually do. Employees should leave knowing what the tool can do, where it tends to fail, and when a decision needs a human handoff. A completion certificate can’t show that. Their judgment can.
See also: The Floor Staff Didn’t Get Replaced By AI, They Got Undermined By It
