Cyprus businesses should organise their data, set clear rules for AI agents and train staff before adding more AI tools, says essere.ai chief executive Kyriaki Parmakki. Her advice is practical: start with one repeatable task, keep a person responsible for consequential decisions and check whether the change delivers a measurable benefit.
Watch Desk analysis
What happened
In an interview with the Cyprus Mail, Parmakki described a gap between people’s use of AI and company adoption. She cited figures presented at an event in Cyprus saying that, in 2025, 44 per cent of people in the country used generative AI, compared with 33 per cent across the EU. The corresponding figures for companies using AI were 9 per cent in Cyprus and 20 per cent across the EU.
Her suggested starting point is operational groundwork, not another strategy document. Businesses should bring client, task and document information into a reliable system they control, write down what an AI agent may read or do, and specify when a person must approve an action. Parmakki also recommends moving staff from personal AI accounts to company accounts and testing one recurring process for four weeks before expanding.
Why it matters
The advice addresses a familiar gap: an employee may feel more productive with AI without the company seeing a measurable improvement. Parmakki cited research involving 1,700 executives, saying 80 per cent reported personal productivity gains while 37 per cent of companies saw benefits reflected in results. Those are figures she cited, not a guarantee that every organisation will see the same pattern.
The proposed safeguards are concrete. Parmakki says companies should log agent actions and review those records weekly, and keep people in charge of decisions involving payments or people. That gives businesses a way to test a tool without treating “the agent did it” as an acceptable audit trail.
Our read
The useful message is not “buy AI” or “wait for the perfect AI strategy”. It is to fix the plumbing, choose a task that happens regularly, and measure what changes. A four-week trial with a clear pass-or-fail test is a sensible antidote to both endless pilots and grand claims made after one impressive demo.
What to watch
- Whether companies replace personal AI accounts with tools and access rules they manage.
- Whether small, repeatable trials produce measurable business results.
- How organisations log agent activity and handle approvals for sensitive actions.
Discussion spark: Should businesses be expected to establish written rules and human approval before letting AI agents act, even if that slows adoption?
Sources and evidence
- Source update (29 October 2026, 06:02 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.