This page explains how OSCABE LTD uses Artificial Intelligence on oscabe.com, what it does, what it does not do, and what your rights are. It supports our Privacy Policy and reflects the EU AI Act and the UK Data (Use and Access) Act 2025.
1. Where AI is used on OSCABE
1.1 AI engineering advisor chat. our AI advisor is an AI advisor that helps Clients describe an engineering brief and receive a shortlist of CE-verified Engineers. It uses a large language model (currently Anthropic Claude, served via OpenRouter) to:
- Extract structured requirements (skills, seniority, industry, timeline) from your conversation.
- Search our Engineer database for the best skill match.
- Generate a natural-language reply describing why each suggested Engineer fits.
1.2 Match-score ranking. We rank suggested Engineers by a deterministic match score based on the overlap between the extracted skills and each Engineer's verified skill set, plus rating and project-count tie-breakers. This part of the pipeline is rule-based, not AI.
1.3 Where AI is NOT used. our AI advisor does not make hiring decisions. Clients select Engineers from the shortlist. CE Verification is performed by a Chartered Engineer (a human). Timesheet approval, ratings, payouts and contract decisions are made by humans.
2. Is this a "solely automated" decision?
No. AI shortlists are advisory; a human Client always decides whom to engage, and CE Verification is human. No solely-automated decision with legal or similarly significant effects is made about any candidate.
Under UK law, the Data (Use and Access) Act 2025 (in force 5 February 2026) replaced Article 22 of the UK GDPR with new Articles 22A to 22D, which permit a wider range of solely-automated decisions provided safeguards are in place (meaningful information, the ability to make representations, human intervention, and the right to contest). Because OSCABE keeps a human in the loop, that regime does not bite on our matching. We nonetheless extend these rights to Engineers as a matter of good practice:
- Disclosure. This page describes the AI involvement, the logic and the influential factors.
- Right to human review. Any Engineer can ask for human review of any shortlist they appeared (or did not appear) on, by emailing info@oscabe.com.
- Right to express a point of view. You can submit additional context that the matcher will consider going forward.
- Right to contest. You can contest the matching outcome and we will reassess.
- Right to withdraw from AI matching. Engineers can opt out of AI-shortlisting in the Engineer Portal; you will still be matchable manually.
3. The logic, in plain language
For each Client brief our AI advisor:
- Uses a small, fast LLM to extract: required skills, seniority, headcount, duration, industry, urgency. Output is JSON.
- Queries the database for CE-verified Engineers whose declared skills overlap with those required, limited to engineers marked available or partially available.
- Computes a percentage match score per Engineer based on skill overlap.
- Sorts by match score, then by average client rating, then by completed-projects count.
- Sends the top 4 to a larger LLM with the Client conversation and asks it to write a natural-language reply explaining the fit. The LLM is instructed not to invent facts.
- Returns the reply and the structured Engineer cards (with rate breakdown) to the Client.
4. Influential factors
The factors that materially influence whether an Engineer appears in a shortlist are:
- Verified skills - the skills listed on the Engineer profile that have been CE-confirmed.
- Availability - Engineers marked "Not available" never appear in shortlists.
- CE Verification - only CE-verified Engineers appear in shortlists.
- Rating and project count - used as a tie-breaker, not the primary filter.
- Public status - Engineers must have made their profile public.
We do not use ethnicity, religion, age, gender, sexual orientation, country of origin, or any other protected characteristic as a factor in matching.
5. Sub-processors involved
our AI advisor sends Client conversation text and Engineer match metadata to OpenRouter, which routes the request to Anthropic's Claude. Both providers operate under EU SCCs Module 2 + UK Addendum and enforce a zero-retention policy on prompts; see Sub-processors.
We do not share Engineer personally identifying information (real name, contact details) with the LLM until that Engineer is CE-verified, public and has consented to AI matching at signup.
6. EU AI Act alignment
The EU AI Act (Regulation (EU) 2024/1689) treats AI used to evaluate or filter candidates as a high-risk system (Annex III). The obligations for stand-alone high-risk Annex III systems have been deferred to 2 December 2027 under the EU "Digital Omnibus" amendment. Although our AI advisor is not making solely-automated hiring decisions, we apply high-risk-grade controls now:
- A documented risk-management process.
- Data governance over training and inference data quality.
- Logging of inference for auditability.
- Human oversight by every Client decision and by Engineer review on request.
- Transparency by publishing this notice.
- Cybersecurity controls in line with our Privacy Policy.
7. Contact
Questions about our AI advisor or AI-related concerns: info@oscabe.com.