TL;DR
- The September registration and module totals describe an early stage of RD Inteligente. They do not measure the programme’s eventual success.
- The public course outline includes checking answers, privacy and fraud awareness. Its participation certificate has stated requirements.
- Dominican employers can build on that learning with a low-stakes exercise using fictional records and a written answer key. Keep customer files out of it.
What happened: an early national training snapshot
The Dominican Republic launched RD Inteligente on 10 September 2026, with ITLA responsible for the free programme and a goal of reaching one million people. For a hotel supervisor arranging staff development or a small-business owner deciding where to start, that creates a practical question: what learning does the certificate record?
The presidency’s 28 September update reported about 54,000 registrations and 4,968 people who had completed at least one module. It also recorded 16,757 new registrations in the preceding week. Those are historical figures, not live counts for October.
The programme website, checked on 13 October, describes six common units taking ten hours. Alongside AI basics and prompting, it lists critical checking, privacy, fraud awareness and a first practical application. The published curriculum gives employers a starting point for a useful discussion about their own work.
The numbers need a time window
Dividing the reported module total by registrations gives approximately 9.2%. That snapshot mixes people who joined at different times. It cannot establish a dropout rate, a final completion rate or why someone had not yet finished. The release also says “16 days” after launch, although its publication date falls 18 calendar days after the launch announcement.
A useful follow-up report would group learners by the week they joined, allow each group the same study period and show how many started, completed a unit and finished the common course. It would distinguish participation from assessed performance. A forecast based on one week of sign-ups cannot tell an employer when a million people will finish.
The wider Caribbean context explains the interest in basic training. ECLAC’s October 2025 account of the Latin American AI Index estimated that Latin America and the Caribbean generated 14% of worldwide visits to AI tools while representing 11% of internet users. It placed the Dominican Republic in the adopter group. These regional usage figures describe activity, not proficiency, and predate RD Inteligente.
There is a reason to involve office and administrative teams. The ILO’s May 2025 occupational-exposure study found clerical work most exposed to generative AI. Globally, one in four workers were in occupations with some exposure, with 3.3% of employment in the highest category. The authors expected job transformation to be more common than replacement. Those estimates concern tasks and potential exposure; they do not count Dominican job losses or evaluate this course.
Read the certificate’s actual requirements
The programme’s terms say its level-one certificate is issued after learners work through all six units and answer the checks. It is a participation certificate, with no equivalence to an academic degree or a manufacturer’s professional certification. Describing it as attendance alone would leave out those checks.
We reviewed the public outline and terms, not a learner account or a completed assessment. We have not evaluated the questions, scoring or what graduates retain. An employer should ask which skills were practised and what evidence the certificate supplies before using it to assign responsibility.
Someone can complete an introductory course and still need supervision when a task involves unfamiliar language, an exception to a policy or a financial consequence. A practical exercise can expose where that support is needed. It should also let staff explain why an answer is uncertain and when they would ask a colleague for help.
Adrian Dunkley’s published FAQ uses the term Shadow Literacy for the judgement needed to recognise plausible errors and know when to verify or withhold delegation. Applied here, it suggests asking staff to show their checking process. The exercise below is our editorial suggestion; it is not an ITLA assessment or a validated test of that concept.
What this means in the Dominican Republic
Consider a fictional Punta Cana guesthouse. A receptionist asks an assistant to draft a bilingual reply about airport transfers. The supplied booking sheet says the transfer is available on request, but the draft promises it is included in the room price. A fluent translation could make the mistake harder to notice. The reviewer needs to find the promise and check it against the booking terms. Both language versions need the same correction before the reply goes out.
A fictional wholesaler in Santiago could try the same exercise with a delivery note. The source lists twelve boxes, but the draft invoice summary says twelve individual items. A reviewer needs to check the unit as well as the number. Both examples use invented records. Colleagues have a precise error to discuss without seeing any guest’s personal details.
A bank or insurer faces more consequential mistakes. An introductory certificate should not, on its own, authorise someone to upload customer records or rely on an assistant for a lending or claims decision. Start with approved training material and wholly fictional cases. Any move into live work needs a separate review of data handling, permissions, system reliability and the person responsible for decisions.
Other Caribbean employers can use the same approach with their own language and procedures. A good result on one set of hotel replies says little about insurance wording, and a correct English answer does not establish accuracy in Spanish. Choose examples from the actual role and have someone who knows that work check the answer key first.
What to do this week: a safe checking exercise
- Pick one low-risk task with your training lead, such as explaining a fictional guesthouse booking. Write a short source sheet and five sample answers. Make two contain specific, realistic errors. Label the whole packet as training material. A subject expert should verify every answer against the source before the session.
- Ask a small volunteer group, perhaps 10 colleagues, to review the packet before their next study session. Allow the same time and source access for everyone. Have each person mark what they would use, correct or refer for help, and explain the evidence behind the choice. Tell them the purpose is learning rather than appraisal.
- Measure accepted errors and rejected correct answers separately. If all 10 people review the two faulty answers, there are 20 chances to accept a known error. Record the count and denominator. Also record whether people found the supporting source and how long checking took; a single accuracy score misses those habits.
- Review the mistakes together after relevant learning. Ask what made each answer convincing and which source resolved it. Within two weeks, repeat with a fresh packet of comparable difficulty, the same instructions and no hints from the first answer key. Note who participated both times before comparing the results.
- Write a short training note within a week of the second session. Identify recurring errors, useful checks and any device, language or time barriers. Choose one change, such as a checklist for price and currency, then test that change separately. Do not treat a small improvement as proof of course effectiveness or permission for unsupervised live work.
You do not need to send the packet to an AI service. A trainer can prepare and review the sample answers offline. If an approved assistant helps draft them, use invented information throughout and check the draft yourself. Removing names from genuine records may still leave details that identify a person.
The exercise needs no universal pass mark. The cost of a missed error depends on the task, and ten volunteers cannot establish a safe error rate for a whole business. Treat an uncertain answer as a chance to practise escalation, rather than rewarding confident guessing. Keep individual results private and report team-level lessons where possible.
Risks to learners and customers
The main risk is overconfidence: a manager sees a certificate and removes checks from work that affects a customer. Keep existing approvals in place while people learn. Give staff time to study and permission to pause when an answer cannot be verified; otherwise the certificate can become another target to meet under pressure.
The exercise itself can mislead. Staff may remember the planted errors, a second packet may be easier, or a rushed reviewer may miss an answer they would normally catch. Discuss the circumstances before drawing conclusions. Do not use these informal results to rank employees, decide promotions or select people for redundancy.
Privacy is another practical limit. A well-intended training session can expose a customer if someone pastes a real case into an unapproved service. Keep this exercise fictional, avoid collecting extra identity information and agree how long to retain the training notes. These precautions apply to any Caribbean workplace using the examples.
What this analysis cannot establish
We have not tested the enrolled learner experience or independently measured outcomes. The September release provides an early snapshot; it cannot tell us how those learners progressed afterwards. The public curriculum shows what the programme intends to cover, while the ILO and ECLAC sources provide broader context. None demonstrates that RD Inteligente improves workplace accuracy. A fair evaluation would need comparable groups, suitable assessments and enough time for learning to show up in practice.
For an employer, the immediate decision is narrower: make study time available, understand the certificate and practise the checks needed for the job. See our related note on the Caribbean AI skills gap.