Evidence that a cohort can work within your stated AI controls.
Aiskilljourney treats enablement as a controlled programme of work. Attendance alone is insufficient. We record what people could do before instruction, what they practised under supervision, and what they can demonstrate against your approved tools and written policies at the close of the engagement.
Corporate learning for generative AI is scrutinised in a way that generic software training often is not. Boards ask whether staff understand permitted uses, whether confidential information is kept out of unapproved systems, and whether outputs are checked before they enter a customer file, a credit paper or an operational instruction. Our evidence model is built for those questions. It is designed so that a named sponsor inside the organisation can file a concise pack alongside the organisation’s own policy documents.
The method is the same across departments of fifty people and across multi-country groups of several thousand, though sampling intensity and the form of the sponsor report are scaled to cohort size. We work with licensed tools the organisation already holds, and with the governance language already in force. Where a policy is silent, we record the gap rather than invent a rule.
Four classes of evidence, kept in one sponsor file.
Every engagement produces a structured file. The file is owned by the client. Aiskilljourney retains only what the contract allows for quality review and for answering follow-up questions within the agreed period. Personal comments about named individuals are avoided unless the client’s own human-resources process requires a named record and has authorised it in writing.
The first class is administrative: enrolment, role grouping, hours attended, and completion of required modules. These figures are dull and necessary. They show that the people who were meant to be in the room were present for the work, and they allow a sponsor to reconcile invoices against delivery.
The second class is baseline and end-point assessment. Before instruction begins, participants complete a short, role-appropriate test of policy literacy and of practical judgement on realistic prompts. The same instrument, or a parallel form, is used at the close. Scores are reported at cohort and role-group level. Individual scores are released to the client only when the statement of work says so.
The third class is observed practice. Facilitators watch participants complete supervised tasks on the organisation’s approved systems: drafting within a permitted template, redacting before a prompt is submitted, citing sources, and escalating when a tool output is uncertain. Observation notes are standardised. They describe behaviours against a checklist, not impressions of personality.
The fourth class is artefact sampling. With the client’s consent, a sample of exercise outputs is retained in the sponsor file. Artefacts are stripped of customer data. They exist to show the standard of work the cohort produced under the rules in force on the day of the exercise, and to illustrate residual errors that the organisation may wish to address in line management.
Supervised practice on approved systems.
Method
Assessment is written against your tools and your policy text.
We do not import a generic vendor quiz and call it governance. Instruments are drafted after we have read the organisation’s acceptable-use rules, data-classification scheme, and any sector instructions the sponsor identifies as binding on staff. Where those documents conflict, we ask the sponsor to nominate the controlling text before assessment is locked.
Baseline
The baseline is short enough to complete in a sitting and long enough to separate guesswork from working knowledge. Items cover permitted tools, prohibited data classes, human review before external use, and the route for incidents. Practical items ask the participant to mark a prompt as acceptable or not, and to say why in a sentence.
During delivery
Facilitators log completion of exercises and note recurring errors at group level. Recurring errors are fed back the same day where the timetable allows, so the cohort can correct the pattern while the material is still in use. Logging is operational, not theatrical.
Close and follow-up
The end-point assessment repeats the literacy and judgement items. A written sponsor note lists residual risks observed in the room: for example, over-reliance on fluent but unsourced summaries, or uneven application of redaction. A light follow-up questionnaire can be issued after sixty days if the statement of work includes it. Follow-up measures self-reported use and remaining uncertainty; it is not a surveillance programme.
Confirm the rule set
Sponsors supply the policy pack and name the licensed tools in scope. Out-of-scope products are excluded from exercises and from scoring.
Lock the instruments
Assessment items are frozen before the first session. Last-minute policy changes are recorded as amendments, with a date, so scores remain interpretable.
Run observed tasks
Participants work in pairs or small groups on tasks drawn from their function: operations, finance, customer correspondence, or control functions, as agreed.
Compile the file
Administrative data, cohort scores, observation summaries and sampled artefacts are assembled into a single pack for the named sponsor.
Sponsor reporting
A file a risk committee can read without a briefing deck.
The sponsor report is written in plain sentences. It states the cohort composition, the hours delivered, the instruments used, the results at group level, and the residual issues facilitators observed. It does not offer legal opinions. It does not rank the organisation against unnamed peers. It does not convert learning data into a claim about model accuracy, vendor selection, or production-system assurance.
Cohort register
Role groups, locations, and completion status. Useful for internal audit when reconciling who was trained on a given control.
Literacy results
Baseline and close scores for policy items, reported as group distributions. Commentary explains which topics remained weak.
Practice notes
Checklist outcomes from supervised tasks: redaction, citation, escalation, and refusal to use an unapproved channel.
Residual issues
A short list of patterns still visible at close, written so line managers can decide what to reinforce in ordinary supervision.
The purpose of the pack is to show, in documents the organisation already understands, that staff were taught the rules in force and were observed applying them on the tools the organisation actually licences.
Sponsor review of the outcome file.
Marked exercise materials retained in sample form.
Quality of the record
Controls on our own work, so the file remains usable.
Evidence is only useful if it can be defended when someone later asks how a score was produced. Facilitators work from a marking guide. Dual review is applied to a sample of end-point assessments on larger cohorts. Amendments to an instrument after delivery has started are logged. If a session is interrupted or a tool outage changes the exercise, the report states the fact and the effect on what could be observed.
Data handling follows the contract. Exercise outputs that might contain client information are kept inside the client’s environment wherever that is feasible. When materials must leave that environment for marking, they are minimised and returned or deleted on the timetable in the statement of work. We do not use participant work to train public models.
Language in the report is checked against the disclaimer that course content is educational. If a sponsor asks us to certify that the organisation is “compliant”, we decline that wording and offer instead a description of what was taught, tested and seen. Certification of regulatory standing sits with the organisation and its advisers.
Sampling
On cohorts above a few dozen people, observed practice is sampled by role group rather than recorded for every seat in every hour. The report states the sampling fraction. Sponsors who require full observation should say so before the timetable is fixed, because it changes staffing.
Limits
What an enablement file cannot prove.
A well-kept learning file shows that people were instructed and assessed against a defined rule set on defined dates. It does not prove that every subsequent prompt in the organisation will be well judged. Behaviour at a desk six months later depends on supervision, system defaults, and the pressure of ordinary work. We will write that limitation into the report when a sponsor asks us to.
The file also cannot stand in for technical assurance of a model, a retrieval system, or a vendor. Those are separate disciplines. Mixing them in one document creates a false sense of coverage. Our programmes address human use, policy literacy, and the habits that keep licensed tools inside their permitted scope.
Nor does a rising cohort score mean that a particular product should be rolled out more widely. Product decisions belong to architecture, security and procurement. We will not dress learning metrics as a business case for spend. If a sponsor needs a paper on rollout options, that is a different request and should be scoped as such on the contact page.
Finally, evidence from one function should not be copied across to another without a fresh baseline. A control-function cohort and a customer-operations cohort meet different prompts and different data classes. We keep role groups separate in the numbers so that a later reader does not average unlike work.
Ask for the specimen contents list before you commission delivery.
Organisations evaluating Aiskilljourney may request a specimen contents list for the sponsor file: section headings, the form of the score tables, and a redacted illustration of an observation checklist. The specimen contains no live client data and no invented testimonials. It exists so that internal audit, learning, and risk functions can see whether the record will fit their filing standards.
When you write to us, name the approximate cohort size, the functions in scope, and whether individual-level scores are required by your human-resources process. Those three facts determine staffing, sampling, and the shape of the report. Enquiries are answered within one business day, Monday to Friday, 09:00–18:00 Singapore time.
Training and enablement services. Course content is educational and does not constitute legal or regulatory advice. Outcome files describe instruction, assessment and observed practice; they do not certify regulatory standing.
This site stores a small preference cookie if you accept the banner, so that the notice is not shown on every visit. Essential cookies needed to send the enquiry form may be set by the server. Read the Cookies page for the full list and retention periods.
