Jump to section
I have hired people who were impressive across a boardroom table and then could not close a file. I have also passed on quiet candidates who, given a real task, produced better work than the confident ones. After enough of both mistakes, I stopped trusting the interview as my main signal. I started paying people to do a slice of the actual job. It changed everything.
This is an argument, so let me commit to it. For most legal roles, from a law clerk to an associate to a bookkeeper, a short paid working interview predicts on-the-job performance far better than any conversation you can have across a desk. The catch is that you owe the candidate real guardrails, and firms that skip those guardrails are not testing skills. They are taking free work.
Why the Confident Interviewee Keeps Fooling Us
The traditional interview measures one thing well: how good someone is at interviews. That is a real skill, and it is almost never the skill you are hiring for. The candidate who makes steady eye contact, tells a tidy story about a hard client, and asks two thoughtful questions at the end has shown you they can perform in a fifty minute conversation. They have told you nothing about whether they can draft a clean affidavit under a limitation period crunch.
We know this and we ignore it, because the interview flatters the interviewer. We feel like sharp judges of character. In reality we are pattern matching against people who look and sound like us, which is how firms end up hiring the same profile over and over and calling it culture fit. A good resume screen filters out noise, and a well built job description attracts the right pool, but neither one tells you what happens when the person actually sits down and works.
The interview measures how good someone is at interviews. That is almost never the job you are hiring for. Sam Whitecloud
What a Real Task Reveals That Questions Cannot
Hand a candidate a redacted, fictional matter and ask them to draft a demand letter, or review a lease and flag three risks, or reconcile a small ledger. Watch what happens. The confident interviewee sometimes freezes. The quiet one sometimes produces the cleanest work you have seen all month. Either way, you are finally looking at the thing you are paying for.
A task surfaces the qualities questions cannot reach:
- Judgment. Do they spot the issue that was not in the instructions? Do they ask a clarifying question before charging ahead, or after they have made a mess?
- Care. Is the work proofread? Are the client names consistent? Small carelessness on a test is loud carelessness on a real file.
- How they handle ambiguity. Real matters are underspecified. Watching someone navigate a gap in the instructions tells you more than any behavioural question about a time they overcame a challenge.
- Speed that is honest. Not raw speed, but whether they can produce competent work in a realistic window instead of an idealized one.
Tip. Score the task against a rubric you write before anyone submits, not against your gut afterward. Decide in advance what a strong answer looks like, and grade every candidate against the same sheet. It keeps the test fair and keeps you honest.
Pay Them, Scope It, Never Ship Their Free Work
Here is the line that separates a legitimate assessment from exploitation, and I want to be blunt about it. If a candidate's output could plausibly be used on a live client matter, you have crossed it. The test must be built on fictional or thoroughly anonymized facts, and the deliverable must never touch a real file. The moment their work has commercial value to you, you are no longer assessing. You are staffing your practice for free.
So pay them. A flat, fair rate for the time the task should take, sent whether or not you extend an offer. This is not a nicety. It is the thing that makes the whole exercise defensible, and it does three good things at once:
- It respects the candidate's time and their professional standing.
- It filters the pool honestly, because people commit differently to paid work than to a homework assignment.
- It signals the kind of firm you are before the person has even started, which matters for the culture you are trying to build.
Watch out. A multi day "trial project" that happens to advance a real client file is not a working interview. It is unpaid or underpaid labour dressed up as assessment, and in some cases it raises employment standards questions. Keep the task short, fictional, and paid.
Designing a Test That Mirrors the Actual Job
A good task is small, self contained, and representative. If the role spends its days drafting correspondence and managing deadlines, do not test them on appellate research they will never touch. Mirror the real work.
| Role | A representative task | Rough time |
|---|---|---|
| Law clerk | Draft a fictional demand letter from a short fact pattern | 60 to 90 minutes |
| Associate | Review a sample agreement and memo the top three risks | 90 minutes |
| Legal assistant | Build a filing checklist and calendar the deadlines for a mock matter | 45 minutes |
| Bookkeeper | Reconcile a small fictional trust ledger and flag any issue | 60 minutes |
Keep the total commitment under two hours. Give clear instructions and a realistic window, and tell candidates what tools they may use. If your firm runs everything through one system, say so, and consider giving a short sandbox in the platform your team actually uses so nobody is guessing. Many practices now run their intake, drafting, and deadlines inside a single hub like A1 CMS, and it is fair to let a candidate see the environment they would work in. What you are looking for is not perfection. It is competent, careful, honest work under conditions that resemble a Tuesday afternoon.
The Line Between a Skills Test and Unpaid Labour
Ask yourself three questions before you send any task, and if you cannot answer all three cleanly, redesign it. Could this output be used, in any form, on a real client file? Am I paying a fair rate for the estimated time? Is the scope short enough that no reasonable person would call it a job? A clean yes to fair pay and no to the first question is the difference between a firm that assesses well and one that will earn a bad reputation among the exact people it wants to hire.
The guardrails are not just ethics. They are self interest. Candidates talk. The internationally trained lawyer you underpaid for a "trial" will tell the next three. Treat the working interview as the start of a relationship, and pair it with the rest of a decent process, honest reference checks and a real onboarding plan for whoever you choose, and you build a name as a place worth working.
So yes, put a paid task in your process. Stop using the confident interview as your main deciding signal. Watch people do a slice of the job, pay them properly for it, keep it fictional, and keep it short. You will hire better, you will hire fairer, and the people you turn down will still think well of you. That last part is worth more than any single hire. If you want more on the people side of running a firm, the whole People and Hiring library is built around exactly this kind of decision.
