Technology needs knowledge and practice
New software has been purchased, accounts are set up and the technology works. That completes the introduction — on paper, at least.
For the people who will use it every day, the real transition is only just beginning. Where should documents be stored? Which template is authoritative? How can changes be traced? And who helps when a task differs from the introductory example?
If these questions remain unanswered, different ways of working should come as no surprise.
For me, staff training belongs in the introduction of new IT just as much as installation and technical support. A tool only delivers its benefits when employees can use it confidently and appropriately.
Good training can make work easier and reduce avoidable mistakes. Simply attending a course is no guarantee of either.
What the evidence says about training
The potential impact of workplace training has been studied scientifically.
A meta-analysis published in April 2003 by Winfred Arthur and colleagues found positive average effects on learning, work behaviour and measured outcomes. Training methods and the skills taught were among the factors associated with effectiveness. This older research offers context, not a guarantee for a particular course today. [1]
A more specific example is a study published online in the Journal of Political Economy on 25 July 2023. Researchers estimated a 13.5 per cent productivity gain from workplace soft-skills training among Indian garment workers. Gains were greater in joint operations, consistent with improved teamwork. [2]
That figure cannot simply be applied to a German office, workshop or IT department. Nor does it establish an equivalent reduction in mistakes.
For me, these findings show that training can deliver measurable business benefits. Whether those benefits materialise in an individual company must be assessed against its actual tasks.
Efficiency begins with everyday tasks
Not every course needs to tackle a major technical topic. I would start where people regularly have to search, ask questions or redo work.
For example, two employees edit different copies of the same quotation. Someone must then establish which changes are valid and which version should go to the customer. Faster typing cannot solve that alone. A shared working method and the knowledge to implement it with the available tools are needed.
These tasks could provide a starting point for a company’s training plan:
Files and documents: use the agreed storage locations, recognise current versions and set access permissions correctly.
Spreadsheets and reporting: check data, understand formulas and assess whether results are plausible.
Order and customer management: record cases fully, assign responsibilities and document progress clearly.
Collaboration: hand over tasks unambiguously and record changes so colleagues can follow them.
IT security: verify unusual requests independently and use the agreed reporting route.
This is a possible selection, not a universal curriculum.
What matters to me is that a specific task is better understood afterwards. “We introduced the new software” describes an event. “Employees can independently create, check and hand over an order” describes a capability.
Knowledge alone does not prevent mistakes
Before treating a mistake as a training issue, establish its cause.
Was a step misunderstood? Is the instruction contradictory? Is a permission missing? Or does an unsuitable program setting force people to enter the same information manually several times?
These distinctions matter. Training cannot replace missing access. It does not automatically resolve unclear responsibilities either.
I therefore think it is wrong to respond to every correction by saying, “Staff need to pay more attention.”
Repeated mistakes call for a review of the workflow itself. A clearer template, automated input validation or a better handover may be needed. Only once the correct technical and organisational approach is clear can it be taught meaningfully.
Training and good work design belong together. People should not have to compensate indefinitely for weaknesses in their tools.
Watching is not the same as working independently
A demonstration may provide an initial overview. To build confidence with an application, I would follow it with practical tasks.
Participants should process a case themselves, check the result and explain why they chose particular steps. Uncertainty should be allowed to emerge.
A January 2008 meta-analysis by Nina Keith and Michael Frese examined error management training: active exploration and learning from mistakes during practice. It suggests advantages over error-avoidant training, particularly when applying learning to unfamiliar tasks. [3]
My practical conclusion is that a suitable training environment should cover more than the flawless standard workflow.
What happens after an incorrect entry? How is an accidental change identified? When may someone correct it independently, and when must the responsible person be involved?
These exercises belong in a prepared environment with appropriate sample data. Real customer documents and production systems are not unrestricted playgrounds.
IT security also depends on the response after a mistake
Security training should provide more than a list of prohibited actions.
Employees need to know, for example, how to check an unusual payment request, who can be reached about a suspicious message and what to do after an accidental approval.
The NCSC’s phishing guidance recommends verifying important requests through a second communication channel. Defences should not depend entirely on employees recognising every fraudulent message; technical controls and suitable processes remain necessary. [4]
I particularly value its advice against a blame-oriented culture: people should feel able to report suspicious activity even after clicking. [4]
My measure would therefore be more than who answered the course questions correctly.
I want to know whether someone can find the agreed contact and ask in time when a situation is unclear. Openly acknowledging uncertainty is better than silently trying to hide or resolve a possible problem alone.
Not everyone needs the same training
Someone in accounts handles different tasks from a field worker or technical administrator. Prior knowledge and familiarity with the programs also vary.
I would therefore not group people solely by department, age or years of service. Their actual tasks and existing capabilities should guide the plan.
NIST’s Building a Cybersecurity and Privacy Learning Program, published in September 2024, describes adaptable, audience-specific learning and ongoing review. Primarily intended for US federal organisations, its planning principles are useful here as guidance, not binding requirements for German businesses. [5]
A small company could begin modestly: shared fundamentals followed by targeted exercises for particular roles.
Experienced colleagues should be able to contribute their knowledge. At the same time, everyone must be able to ask a basic question without having to justify it.
What a useful training programme could look like
I would start with a clearly defined problem instead of immediately assembling a comprehensive annual programme.
Define the task and the goal first
“Work better with computers” is too vague a goal.
A more specific goal is that an employee can record a customer enquiry completely, attach the necessary documents and hand it over so the next person can continue without further explanation.
Before training, establish where this is not yet happening reliably.
Explain the process, then let people try it
The introduction should explain the purpose of the workflow and its important steps. Participants then work through a suitable practice case.
For a limited topic, a short session followed by application can be a useful start. Complex or safety-critical work needs more time. A particular number of minutes is not a measure of quality.
Leave behind a useful reference
After training, nobody should have to remember every detail by heart.
A short, current guide containing the agreed process, common questions and a contact person would often be more useful to me than a lengthy presentation. It needs to be stored where people can find it quickly during work.
Check later what has actually carried over
After a suitable period of application, I would revisit the same task. What now works independently? Where do questions remain? Which training assumption did not fit everyday work?
These answers identify the next improvement. Another exercise may be needed, or the workflow itself may need changing.
New AI tools need clear rules too
For AI use, I would not limit training to writing clever prompts.
My plan would also cover which information may be entered, which application is approved for the purpose, which results need professional checking and who decides whether a text, calculation or suggestion can be reused.
A useful exercise might involve comparing a draft with supplied original documents. It would only be complete once discrepancies had been identified, uncertainties recorded and necessary corrections made.
The learning goal is then a checked work product, not the fastest output.
I would also specify the tasks for which AI must not be used in the business. A technical possibility is not the same as organisational approval.
Time for training must be planned
Anyone scheduling training should create the conditions that make participation possible.
I would not send employees to a course while also expecting them to answer calls and messages continuously and complete the same workload as on an ordinary day.
Time to apply the learning belongs in the plan too. When a new method is introduced, room is needed to clarify questions and resolve uncertainty.
Managers share that responsibility. They should support and use the agreed processes themselves. A jointly trained workflow means little to me if it is abandoned at the first urgent request.
Training should not automatically turn every improvement into a higher performance target either. Time gained can also support more careful advice, better preparation and less rework.
Measure success through outcomes
The number of courses attended says little about whether work has improved.
NIST calls for evaluating impact and improving learning programmes over time. Evidence that an event took place is therefore not enough. [5]
For a specific workflow, I would examine how often rework is needed, whether handovers are complete and how long processing takes, including corrections.
Speed must not be the only consideration. A faster report is no improvement if more errors subsequently need correcting.
Before-and-after comparisons also require care: if software, responsibilities or workloads changed at the same time, an improvement cannot straightforwardly be attributed to training alone.
A calculation to put this into perspective
Suppose eight employees each save an average of six minutes per working day through a working method they genuinely understand. Assuming 220 working days, that amounts to 176 hours per year.
This is an illustrative model, not a measured saving or a promise. Training, preparation and support also take time. Nor do theoretical minutes saved automatically become recoverable staffing costs.
The calculation simply illustrates why small, repeated tasks can be worth examining closely.
My conclusion: Training belongs in the investment
When a company invests in new technology, it should also enable people to work with it.
For me, an introduction is convincing when employees can carry out their essential tasks, check results and find the right route when problems arise.
That does not require the longest possible list of certificates. It requires relevant content, practical application, accessible support and a willingness to improve the company’s own processes.
Well-trained employees do not need to memorise every exceptional case. They should recognise when a task differs from the usual workflow and know how to proceed responsibly.
Anyone expecting better results must provide more than tools. They must make knowledge, practice and time available too.
Scope and research
Research date: 9 October 2026. Findings apply to the conditions studied and cannot be transferred unchanged to every business. The examples do not promise particular savings or error rates. The illustrative calculation is 8 × 6 × 220 ÷ 60 = 176 hours, without deducting time spent on training and support.
Sources and further reading
Sources checked on 9 October 2026. Access dates are not publication dates. Research claims were checked against original abstracts and publication details. The Leuphana University research portal additionally confirmed the Keith/Frese paper because the direct PubMed request returned no article text. The full JPE study is not freely available; the findings reported here appear in its public abstract. NIST and NCSC guidance was checked directly.
[1] Arthur, Bennett, Edens and Bell: Effectiveness of training in organizations (April 2003)
Read source ↗
[2] Adhvaryu, Kala and Nyshadham: Returns to On-the-Job Soft Skills Training (online 25 July 2023)
Read source ↗
[3] Keith and Frese: Effectiveness of error management training (January 2008)
Read source ↗
[4] NCSC: Phishing attacks: defending your organisation
Read source ↗
[5] NIST SP 800-50 Revision 1: Building a Cybersecurity and Privacy Learning Program (12 September 2024)
Read source ↗