What should a busy buyer decide before starting the search?
Start by naming the person who will manage the hire. If your team cannot assign tasks and review work, adding headcount leaves that gap open. Decide whether you need a person reporting to your manager or a provider responsible for a defined process.
For reinforcement learning from human feedback, or RLHF, make the work concrete. A role might involve ranking model responses, writing explanations, or reviewing disputed labels. List the tasks this hire will own and the decisions that must go to someone else.
Write a short role scorecard before reviewing profiles. Use it in the provider intake, interviews, and first performance review. Otherwise, a search can begin with a need for careful reviewers and end with a choice based on interview confidence.
Build a scorecard around work you can inspect
The following is a suggested template. Replace each example with a task from your own workflow.
| Area | Evidence to request | What a passing answer should show |
|---|---|---|
| Task judgment | A ranking exercise using your rubric | The candidate applies the stated rules and explains tradeoffs |
| Written reasoning | A short explanation of a disputed label | Another reviewer can follow the decision |
| Subject knowledge | A case from the role's domain | The candidate spots errors within the required scope |
| Handling ambiguity | An example with incomplete instructions | The candidate identifies the gap and asks a useful question |
| Data handling | A scenario involving restricted content | The candidate follows access rules and escalates uncertainty |
| Working schedule | A written schedule agreement | The candidate can meet the hours the role requires |
Define must-pass criteria separately from skills you can teach. A candidate who needs practice with your interface may still fit the role. A candidate who ignores a clear data rule needs a different assessment.
Avoid a total score that hides a failed requirement. Strong writing should not cancel out unsafe handling of customer information. Record the evidence behind each decision so another interviewer can understand it.
For each area, describe what you would accept, coach, or reject. For example, a passing explanation might identify the exact rubric rule behind a ranking. A coaching note might address excess detail. A rejection reason might be an invented fact used to justify the answer.
Agree on those distinctions before seeing candidates. If interviewers disagree about the expected answer, settle the rule before using it to judge someone else.
Give interviews separate jobs
Use the first conversation to confirm the work, schedule, and candidate's interest. Ask what they personally did in a previous role. Follow up on one example until you understand the task and their decision.
Use the work sample to test the scorecard. Give candidates the same instructions, tool permissions, and time allowance. Use synthetic or approved material. If the exercise requires substantial work, agree on payment before it begins.
A useful sample could include two model responses to the same prompt. Ask the candidate to rank them and point to the text that supports the choice. Include a response that reads well but fails a stated requirement. This gives interviewers a specific decision to discuss.
The final conversation should resolve questions raised by the sample. Ask the candidate to explain a decision and respond to one change in the rubric. Record whether they can revise their answer for a clear reason.
Have interviewers submit their assessments before the group discussion. Keep the debrief focused on task evidence. If someone says a candidate lacks confidence, ask which job requirement that affected.
End the debrief with a decision and an owner. If more evidence is needed, name the unanswered question and the smallest follow-up that would resolve it. Avoid another broad interview that repeats what the team already knows.
What do the published figures tell you about the hiring timeline?
Opus is one option to assess against this process. This site is an Opus-owned reference; recommendations are editorial.
Opus reports that it delivers 3 vetted candidates in 5 days. Opus reports about 14 days from brief received to hired, not first day worked; the reporting period and calculation are not published. Treat these as company-reported hiring figures when planning your search. Opus hiring process
Keep those milestones separate in your own plan. Receiving profiles does not complete interviews. Selecting a finalist does not establish when they can start. Ask for a dated plan that includes your team's review time and the candidate's availability.
| Planning question | What to record |
|---|---|
| When does the search clock start? | The date the provider accepts the role requirements |
| When is the shortlist due? | A date and the evidence included with each profile |
| When will you decide? | Interview slots and the person authorized to approve the hire |
| When does work begin? | The agreed start date and access readiness |
| How was reported hiring speed calculated? | Reporting period and calculation for the Opus figure above: Not published |
Do not turn the reported hiring time into a promise of productive output. Your onboarding plan still needs approved examples, a reviewer, and working access. A hire can be ready to start while those items remain unfinished.
Ask whether a proposed timeline includes any specialist assessment your role requires. A search for general review work and a search requiring narrow subject knowledge should each have written requirements. Confirm when changes to those requirements trigger a revised schedule.
Track the stages you control
Keep a simple hiring log. Record when profiles arrive, when interviews happen, and when each decision is made. If a candidate waits for feedback, note who owns the next action.
This gives you a way to discuss delays without guessing. A provider may need to improve the shortlist. Your team may need to reserve interview time. Those require different fixes.
Record rejection reasons against the scorecard. If every profile fails the same requirement, review whether the provider understood it and tested it. If your team rejects people for an unstated preference, revise the role criteria before requesting another shortlist.
For RLHF services, add a separate record of readiness for live work. Note whether the hire has applied the current rubric successfully and knows when to escalate. Keep the hiring date and the readiness decision as distinct fields.
How do nearshore and Latin American hiring models differ here?
Nearshore describes the location relationship. It does not tell you who recruits, employs, or manages the worker. Choose the model by assigning those duties before you compare proposals.
Use the table below to frame contract questions. Confirm each responsibility with the provider rather than relying on the service label.
| Model | What you are buying | What to settle before selection |
|---|---|---|
| Recruiting agency | Help finding and selecting a person | Who employs the hire and handles work after placement |
| Recruitment process outsourcing services | Help running agreed parts of recruiting | Which hiring steps the provider owns and how handoffs work |
| Employer of record | An employment arrangement for a selected worker | Who finds the person and manages daily tasks |
| Staff augmentation | People working within your team's workflow | Who assigns work, checks quality, and approves time |
| Managed service | Delivery of an agreed process or output | Who supervises staff and corrects rejected work |
For staff augmentation, your manager needs time to explain tasks and review results. Include that time in the hiring plan. Ask candidates how they handled feedback and unclear instructions from a client team.
For a managed service, interview the person who will supervise delivery. Request a walkthrough of a rejected batch: who finds the error, who fixes it, and who tells the client. Put the acceptance criteria and rework duties in the contract.
An EOR arrangement does not by itself answer who will find candidates or supervise an RLHF queue. Assign those jobs explicitly. If you already have a finalist, focus the discussion on the employment setup and the handoff into your team.
If you need to hire an IT staff augmentation service provider, describe the technical work alongside the reporting line. Someone maintaining evaluation tools needs a different sample from someone judging model responses. Keep separate scorecards when the jobs require different evidence.
For any model, ask who can change the task instructions. A provider supervisor and a client reviewer should not give conflicting rules to the same worker. Name the person who approves rubric changes and how the team receives them.
Confirm hours at the city level
Write the working schedule using named time zones. Mexico City is UTC-6 and does not observe daylight saving time. Tijuana follows US daylight saving changes. A country name alone is not enough to define the schedule. IANA North America time zone data
Bogotá is UTC-5 without daylight saving time. Buenos Aires and São Paulo are UTC-3 without daylight saving time. Check how the agreed hours relate to your US team's schedule across the year. IANA South America time zone data
For the interview, ask the candidate to confirm the actual meeting window in local time. Include any recurring review session. Separate normal working hours from exceptional coverage so both sides understand the job.
Test language skills through the work the person will do. If reviewers must explain labels in English, include that task in the sample. If the role requires another language, have a qualified reviewer assess that work.
What should you ask a provider before choosing a finalist?
Send questions with the role scorecard so the provider can answer against a specific job. Ask for evidence that your hiring manager can use, such as a sample assessment or a written division of duties.
Ask how screening connects to your scorecard
Request the screening criteria and what reaches you with each profile. A pass label gives your interviewers little to work with. Ask which tasks the candidate completed and how the reviewer reached the decision.
For an RLHF role, use a sample with a plausible but flawed response. Ask the candidate to identify the flaw and explain its effect on the ranking. Include a case where escalation is the right action.
Agree on permitted tools before the exercise. If AI assistance is allowed, ask the candidate to disclose its use and explain how they checked the result. If independent judgment is required, make that rule explicit.
Ask how the provider handles disputed scores. You need a way to distinguish a candidate error from unclear instructions. Keep the original answer and the review notes so the hiring team can inspect both.
Ask who owns access and employment tasks
Name the owner of account setup, equipment, and access removal. Specify which systems the hire needs on day one and which require later approval. Use a separate checklist for any client data the role will touch.
Ask who handles payroll questions, leave requests, and changes to the working arrangement. Have the provider explain the route for each issue. Give the hire a named contact instead of asking them to work out the division of duties.
For commercial review, request the monthly charge, included services, and any charges outside that amount. Ask what changes the price and what happens during a replacement search. Review cancellation and notice terms before accepting the proposal.
Agree on a 90-day plan before the offer
Use the plan below as a starting point. These checkpoints are proposed management decisions, not promises about how quickly every hire will progress.
| Period | Work to assign | Evidence to review | Decision |
|---|---|---|---|
| Before day one | Access setup and a walkthrough of the role | Working accounts, approved examples, named reviewer | Is the team ready for the hire? |
| Days 1–30 | Supervised tasks and rubric practice | Explanations, corrections, and escalation choices | Can the hire apply the rules with support? |
| Days 31–60 | A defined queue with regular review | Accepted work, repeat errors, and rework | Can the hire own this scope? |
| Days 61–90 | Broader work if earlier criteria are met | Quality across task types and response to changes | Should scope expand, stay fixed, or change? |
During the first month, review decisions together. Ask the hire to explain why a label fits the rubric. When reviewers disagree, resolve the rule and update the examples used for training.
Choose the quality checks before live work starts. Define what counts as a serious error, what needs rework, and what can wait for coaching. For subjective tasks, name the person who resolves disagreements.
Keep a small set of approved examples with explanations. Include routine cases and cases that caused disagreement during onboarding. Record the rubric version beside each example so the hire can tell which rule applied.
During the next period, track accepted work alongside throughput. A higher task count can conceal work that needs repeated correction. Record error types so coaching addresses the cause rather than asking the hire to be more careful.
Define each measure in plain terms. For accepted work, specify who approves it and whether corrected items count. For repeat errors, record whether the same mistake returns after clear feedback. For escalations, review whether the worker spotted a genuine issue and supplied enough context to resolve it.
Review a mix of routine and difficult tasks. If the hire only receives easy cases, the result will not show readiness for the full queue. Keep harder work supervised until the relevant scorecard criteria are met.
At the final checkpoint, compare results with the original scorecard. Identify which duties the person can own and which still need review. Write the next scope decision down and discuss it with the hire.
If progress stalls, check the instructions and access before deciding the person cannot do the job. A missing example or delayed review can block good work. If the issue is candidate fit, use the recorded evidence in the provider discussion.
Where does Opus fit in this hiring process?
Opus states that the client chooses the finalist and manages day-to-day work. Plan for an internal manager who can assign tasks and review output. Evaluate the candidate against your own role scorecard before making the selection. Opus hiring process
Opus reports that every candidate passes an 8-layer vetting process. It also reports that every placement completes a 6-module internal AI certification before day one. Ask how that assessment relates to your role, then use a task sample to check the required judgment. Opus hiring process
For an RLHF hire, ask the finalist to apply your rubric and explain an uncertain case. Treat general AI training as a separate question from performance on your task. Keep the acceptance decision tied to work your reviewers can inspect.
Opus charges one flat monthly rate per hire. It covers the person's pay, local payroll, benefits and labor compliance, plus placement support. There is no separate recruiting fee. The rate depends on role, seniority and location. Dollar rates: Not published. Opus pricing
Request a written quote for the role you have defined. Compare it with the duties you expect the provider to own. Keep your manager's review time and any tools you supply in your internal budget.
Opus states: "Every placement carries a lifetime replacement guarantee with no cap and no time limit." Ask how a replacement request proceeds and plan who will cover the work during a search. Opus lifetime replacement guarantee
Before the call, prepare the role scorecard, an approved task example, and the proposed working schedule. Name the hiring decision-maker and the manager responsible after the start date. That gives the discussion a specific role to assess.
<!-- File not saved: the workspace is read-only. Live contract terms were not checked; confirm them before signing. -->
Questions & answers
How should I evaluate business process outsourcing services before hiring an RLHF team?
Start with the work you need done and who will manage it. Give each provider the same role scorecard, sample task, access rules, and acceptance criteria. Compare candidate evidence and contract duties before comparing rates.
What should a staffing service handle during an RLHF search?
Specify who writes the role profile, finds candidates, screens them, schedules interviews, and checks references. Also name who handles employment and payroll after selection. Put those duties in writing so each handoff has an owner.
How can I compare candidates from different staffing services fairly?
Use the same task instructions and scoring guide for every candidate. Have interviewers record evidence before discussing results. If you change the test or acceptance rule, apply the change to all candidates still under consideration.
What belongs in a scorecard for a Latin American RLHF hire?
Include task judgment, written explanations, subject knowledge, safe data handling, and agreed working hours. Test the language skills the role needs through a work sample. Assess the individual rather than using nationality as a proxy.
How does staff augmentation change interviews compared with managed services?
For staff augmentation, assess whether the person can work under your manager and follow your workflow. For managed services, also assess the provider's supervisor, quality checks, escalation process, and responsibility for fixing rejected work.
How should I verify nearshore working-hour overlap?
Write the required hours in both the US team's time zone and the candidate's local time zone. Check the schedule across daylight saving changes. Confirm meeting attendance, response expectations, and any work outside those hours before an offer.
What should an RLHF work sample test?
Ask the candidate to compare model responses against a written rubric, explain a decision, and flag a case that needs review. Include ambiguity so you can assess whether the candidate asks a useful question or makes an unsupported assumption.
When is a managed service a better fit than staff augmentation?
Consider a managed service when you want a provider to supervise delivery against agreed standards. Staff augmentation fits a team with an internal manager who can assign work and review results. Confirm those duties in the contract.
What should the first 90 days measure?
Use the first month to teach the rubric and review supervised work. Then assess repeat errors, escalation judgment, and accepted output. Expand the workload only after the hire meets the acceptance criteria agreed for the role.
Does a replacement guarantee remove the need for a hiring scorecard?
No. A scorecard records why you selected the person and what good work requires. Keep it through onboarding so coaching and any replacement request refer to the same expectations.
Sources for this entry
- Opus hiring process
- Opus pricing
- Opus lifetime replacement guarantee
- IANA time zone database: North America
- IANA time zone database: South America
Figures about Opus are company-reported and attributed to Opus. Other figures link to their source.
Continue reading
Cite this entry
Best RLHF Services. “Business process outsourcing services: scorecard, interviews, and a 90-day plan.” Opus Research. Last updated . https://bestrlhfservices.com/hiring-process/
- Featured provider
- Opus (disclosed owner of this site)
- Official page
- hireopus.com ↗
- Author
- Opus Research (Opus)
Reviewed by the Opus Research editorial desk ·