Don’t just make your team aware of AI. Understand how they actually use it.

Because the cycle doesn't only ask. It opens with an AI-drafted document containing planted problems — a fabricated citation, a missing piece of evidence, a decision a person must own — and records what the employee catches unaided. The reflection is written straight afterwards, anchored to that result rather than to a remembered episode. You get what they say and what they demonstrably did.

No. There is no pass mark, no ranking and no gate, and a manager never sees an individual's result — only a team trend over time. That is deliberate: the moment there is something to win, people optimise for the score instead of answering honestly, and honest input is the whole value of the record. The employee sees their own result in full.

Legal practices and financial services have their own case banks, written as the work those people actually do: a research note carrying an authority that does not exist, a suitability rationale asserting a rule nobody has opened, a prompt about to take a client name into a public tool. Other sectors get the general bank, which is deliberately not dressed up as sector content. If a bank does not exist for your work yet, we would rather say so than hand your people a generic exercise.

No, not for several cycles. Each person rotates through the bank on their own count, so two colleagues on different cycles are not looking at the same document, and the support fades as they go: the first check lists the kinds of problem present, the next tells them only how many, and after that there are no hints at all and there may be several problems or none.

No. They are off until an administrator turns them on in your organisation settings, and the cycle runs as a reflection on its own until then. We default them off so that nobody meets the case content in a live organisation before someone on your side has read it.

No. It's a short, periodic reflection that sits on top of the training you already run. Skilly captures the applied judgement that shows the training actually changed behaviour, and turns it into evidence.

No, it runs alongside them. There's no content to migrate and no course to switch off. It adds a thin evidence loop, it doesn't replace a workflow.

Both, deliberately. Every question starts from a curated bank built on the SHARP framework, which is what keeps the standard consistent and comparable. Then, once you describe your organisation and label each team's function, questions are personalised to that context: a wealth manager's client-advice team sees a different scenario to a retail bank's credit team, for the same underlying habit, tuned to role seniority. Every personalised variant is reviewed and approved by your compliance manager before any employee sees it; until then, the curated original serves. Confidentiality guards are preserved verbatim and checked automatically.

In principle, yes: the reflect, score, evidence method isn't specific to AI. AI competency (the SHARP habits) is the framework it ships with today; areas like leadership, ethics or communication would run as additional frameworks on the same platform. Today, a further module is something we scope with you.

Every reflection is screened for live AI-governance risk: data exposure, unverified output, undisclosed use, over-reliance. A genuine risk raises an alert for the manager and compliance manager; closing it records the action taken, which joins the audit trail. A near-miss the person caught themselves is scored as awareness, not flagged, so honesty stays safe.

A per-employee evidence pack, printable after a human review: standing on each SHARP habit, the full scored reflection record with any manager overrides, and the risk log with actions taken. Plus the cohort dashboard and organisation summary for the board.

About 15 minutes per person per cycle, fully asynchronous. Nothing is taken away from existing training time. You choose the cadence when you set the organisation up — monthly for a new-joiner ramp, quarterly (the recommendation for an established team) or semi-annually. Managers aren't asked to run a programme either: each gets one digest per cycle, sent only when something actually needs their note or override.

Very little: you can start with just a browser. Optionally, IT can wire single sign-on (Microsoft or Google) and allowlist the domain: roughly an hour, one-time. No LMS integration, no data migration.

Skilly Work deliberately sits outside your stack: it's a browser link secured by your existing Microsoft or Google sign-on, so there's no system integration to security-review before a pilot. People are onboarded by CSV today, evidence packs export as PDF and CSV for whatever your compliance team uses, and reflection reminders are sent by email automatically as each cycle closes (employees who haven't logged an entry, and managers with team reflections awaiting review). Reminders in Teams or Slack, and automatic people-provisioning from your directory (which HR systems like Workday typically already feed), are next on the roadmap, prioritised with pilot partners.

No, and we'd advise against it. Start with one team and one cycle, look at the evidence, then expand where it's most useful. Small and real beats big-bang.

Now, small. Reflective evidence counts because it accrues: a record that starts the month a duty lands is worthless; one that starts now is years deep. Article 50 transparency is live from August 2026 and the high-risk oversight duty lands December 2027, so the deferral is the reason to start, not to wait. Start one team now and you meet those dates with a record, not a certificate.

Try the interactive demo today, with no sign-up. Live workspaces are onboarded one organisation at a time: talk to us and we'll set up your pilot and walk through a first cycle.

No. Reflections are self-authored: people write about how they worked, in their own words. Managers see behavioural patterns by team and role to coach around. It's formative, not disciplinary.

That's the employer's call, like annual compliance training; Skilly doesn't impose it or penalise anyone itself. Someone who doesn't take part shows as ‘no reflections logged’: a gap in your evidence that managers follow up through the usual channels. The content stays safe: owning a mistake is scored as awareness, not marked down.

Yes. The employee reflection flow runs in Arabic: the question banks (general and financial services), the writing scaffold and Skilly's formative feedback, with an English ⇄ العربية toggle when the organisation allows it. Manager dashboards and the audit-ready evidence packs stay in English, so the supervisory record reads exactly as a reviewer expects.

Yes. An admin can upload your AI policy or training material, and Skilly references it when scoring and screening, so it applies your approved tools and disclosure rules, not generic ones. It never overrides the SHARP rubric, and employees never see the document through Skilly.

No confidential information or personal data is needed. Reflections cover how someone used AI and the judgement they applied, not the client file, the case notes or the raw data itself.

Yes. The governing body is a setting, not a rebuild. Choose your supervisory context and every evidence pack adapts its framing: for the Central Bank of Ireland, packs carry EU AI Act references (Articles 4, 50 and 26); for the FCA, which has no standalone AI rulebook, the same record is framed for SM&CR accountability, the Consumer Duty and SYSC; for the Central Bank of the UAE, packs frame the record for the five principles in its February 2026 responsible-AI guidance, including effective human oversight; for the Qatar Central Bank, packs frame the record for its Artificial Intelligence Guideline (September 2024), which makes boards accountable for AI outcomes and mandates a human oversight protocol for any AI system. In every case Skilly Work evidences the people dimension of AI governance; it doesn't itself deliver regulatory compliance.

The Omnibus is now law (Regulation (EU) 2026/1744, in force 27 July 2026). It softened the AI-literacy duty (the Act now says in terms that no organisation has to guarantee any individual's literacy level) and deferred the high-risk rules to December 2027. So we don't sell an ‘Article 4 makes you’ mandate; the softening is exactly why a completion certificate proves nothing and a record of judgement is what's left. What didn't move: Article 50 transparency is live from August 2026, the human-oversight duty still arrives, and supervision looks at what you can show your people did. Skilly Work is that record, and for registered high-risk systems, the evidence pack adds the per-person oversight-competence section Articles 26(2) and 27 will ask deployers for from December 2027.

In one specific way. DORA (Regulation (EU) 2022/2554, applying since January 2025) makes ICT security awareness and digital operational resilience training compulsory for all employees and senior management, and requires the management body to understand ICT risk. Skilly Work is not that training and not an ICT control. What it adds is the evidence a supervisor asks for next: that where those duties meet day-to-day AI use, the understanding the training assumes was assessed rather than assumed.

The literacy duty itself carries no direct fine, which is less headline risk than most vendors imply. The real exposure is the human-oversight duty for higher-risk uses (hiring, performance, credit; finable from 2027), plus unmanaged, unverified AI already in your workflow. In each case, what protects you is what you can show your people actually did.