How to Decide When Someone Is Ready to Work Alone

Why this matters

The solo-release decision is the highest-stakes call a small shop makes with the least structure behind it. Release too early and the cost lands on a customer: a miss, a callback, a damaged system, occasionally an injury, and a tech whose confidence takes a hit that takes months to rebuild. Release too late and you burn a senior tech's hours babysitting work an apprentice could have carried, while the apprentice concludes you do not trust them and starts taking calls from the shop across town. Most shops make this call on a feeling in the gut of one person on one Tuesday. It deserves a procedure, and the procedure is not long.

Step 1: Do not confuse readiness with qualification

Where the work requires a license, certification, permit, or a regulated qualification - gas appliance work, live electrical work, confined-space entry, work at height under a fall-protection program, refrigerant handling, anything requiring a permit in your jurisdiction - readiness is irrelevant until the credential exists. No amount of demonstrated competence substitutes for the card, and no supervisor's confidence protects the shop, the customer, or the tech if something goes wrong. Check the credential first, and check what your jurisdiction requires in terms of supervision ratios for the credential the person actually holds, because those ratios often govern how many apprentices one qualified person may supervise and where.

Once the credential question is settled, everything below is about the readiness judgment the credential does not answer.

Step 2: Understand that "alone" is four different releases

Shops treat solo release as one switch. It is at least four, and they should be granted separately:

An apprentice can be perfectly ready for the first and nowhere near the fourth. Granting them together is how a shop ends up honoring a scope commitment a fourteen-month tech made in a driveway. Name which release you are granting, out loud, and write it down.

Step 3: Release by job type, not by person

Readiness is a property of a person-and-job-type pair. "Is Marcus ready to work alone" is an unanswerable question. "Is Marcus ready to run a residential seasonal maintenance visit alone, with no gas appliance work in scope" is answerable in an afternoon.

Build the list of job types you actually dispatch - usually eight to fifteen for a small shop - and treat each as its own release. This also solves the political problem, because a partial release stops feeling like a verdict on the person and starts feeling like a map with more of it filled in every quarter.

Step 4: Write the sign-off criteria before you need them

Criteria written while looking at a specific person are written to justify a decision you already made. Write them once, per job type, in advance, and keep them short. Four to seven lines is enough:

• The specific tasks that must be demonstrated unaided.
• The readings or checks that must be taken and what they are compared against.
• The stop conditions: what makes this job stop and become a call to the shop.
• The documentation that must come back complete.
• The minimum number of supervised performances before a test is even offered.

That last line prevents the most common distortion, which is a busy week producing an early sign-off because dispatch needed a body.

Step 5: Run the three-part readiness test

All three, in this order. Failing any one is a stop.

Part one: the safety veto. Ask them to walk the job's hazards and the isolation or verification steps unaided. This part is not scored, it is pass or fail, and it is a veto. A tech who cannot state how they make the work safe does not get released regardless of how strong the other two parts are, and re-testing happens after remediation, not on the spot.

Part two: the cold demonstration. They perform the job unaided, at least 30 days after the last time they were taught it, with the assessor present but silent. The 30-day gap matters: performance immediately after instruction measures short-term recall, and short-term recall fades exactly when the supervision goes away. Silence matters too. An assessor who hums, shifts, or reaches for a tool is coaching, and you learn nothing.

Part three: the judgment probe. Five short "what would you do if" scenarios, drawn from the real edge cases of that job type. The customer is not home. The equipment is not the type on the order. You find a second problem that is outside scope. A reading comes back impossible. You are running an hour behind on a day with three more stops.

Part three is the one that predicts solo performance best, and the one shops skip. Procedure gets you through the job that goes as expected. Judgment gets you through the one that does not, which is most of them.

Step 6: Set the escalation rule before the first solo, in one sentence

Before the first solo, agree the exact sentence: "If you hit anything on the stop list, or anything you have not seen before, you stop and call me, and calling me is never the wrong answer."

Then honor it. The first time a tech calls and gets a sigh, you have bought yourself a tech who guesses instead. Track early solo calls as a positive metric, not a negative one. A new solo tech who calls three times in their first month is behaving correctly; one who calls zero times is either extraordinary or hiding, and it is worth knowing which.

Step 7: Use a graduated release, then look at the first fifteen

Do not go from full supervision to full autonomy in one step. A workable ladder:

1. Solos one through five: one scheduled mid-visit check-in call, plus a same-day review of the job notes.
2. Solos six through fifteen: no call, same-day note review only.
3. After fifteen: normal review cadence, release recorded on the skills matrix with the date and the assessor's name.

Then look at the fifteen as a set, and sort every miss by cause before you judge. That sorting step is the whole discipline: a job that failed because the part was not on the truck is a stock problem wearing a readiness costume, and counting it against the tech teaches you the wrong thing about them and leaves the actual problem unfixed.

Worked example: one apprentice, one job type

An apprentice at 14 months. Job type: residential seasonal maintenance, no gas appliance work in scope, no permit required.

Prerequisites. No credential gate applies to this scope, confirmed against the state's requirements for the work in question. The job type's sign-off card lists 11 required tasks. The skills matrix shows 9 at level three (done alone and verified) and 2 at level two (done with supervision). The two at level two are pulled forward and worked in the next two weeks until both hit level three.

Supervised volume. 22 visits of this type observed, 14 performed with a senior tech present. That clears the card's minimum of 10 performed.

The three-part test. Safety veto: pass, unaided, no prompting. Cold demonstration at 34 days since the last instruction on the type: pass. Judgment probe: four of five scenarios answered well. The fifth revealed that the apprentice would have kept working past a condition that is on the stop list, reasoning that the customer had waited all week for the appointment.

The call: stop. That fifth answer is not a small gap. Working past a stop condition to protect a schedule is precisely the failure that produces the incidents that end careers, and it does not get coached away in a conversation. The remediation is specific: two weeks, the stop list drilled as a named list rather than a general principle, plus the lead tech telling them about a time they stopped a job and what happened afterward. Retest the probe with five different scenarios. That retest passes.

The graduated release. First 15 solos run on the ladder above. Result: 12 of 15 complete on the first visit, against a shop average of 88 percent for that job type, so on raw numbers the apprentice looks 8 points light. Sorting the three misses: two were parts not on the truck, one was a genuine diagnosis miss that required a return. Adjusted for the stock issue, 14 of 15, or about 93 percent. Full release granted for that job type, dated and signed on the matrix, with a follow-up talk on the one real miss.

What the release was worth in hours. The 14 supervised performances consumed roughly 1.2 hours of senior tech time each, about 17 hours invested. The job type takes about 1.5 hours and the shop dispatches roughly 6 a week; moving those to the apprentice returns about 9 hours a week of senior capacity, or roughly 432 hours across 48 weeks. The 17 hours of investment is repaid in under two weeks of dispatch. That ratio is the reason to be systematic about this rather than cautious about it: the cost of the process is small and the cost of an extra six months of unnecessary supervision is enormous.

What flips the decision

• Commercial or occupied-facility work. Higher consequence per mistake, more parties on site, tighter access rules. Push the recommendation and commitment releases later even when the on-site release is clearly earned.
• A job type where the failure is delayed. Where a mistake shows up weeks later rather than during the visit, supervised volume has to be higher, because the feedback loop that normally corrects a new tech is missing.
• Working after hours or in a customer's occupied home alone. This is as much a conduct and judgment release as a technical one, and it should be granted on its own evidence.
• A very small shop where the senior tech is also the owner. The temptation to release early is at its strongest here and so is the cost of a miss, since there is no bench. Do not compensate with delay; compensate with a shorter, harder ladder and a real escalation rule.
• Rehiring an experienced tech. Experience elsewhere is evidence, not sign-off. Run the same three-part test on a compressed timeline, because their unknowns are your local practices, your stop list, and your customer expectations, not the trade.

What doing this wrong looks like

Release by tenure. "He's been here a year, he's fine." Tenure correlates with competence loosely and with job-type competence hardly at all. The tech who has spent a year on the same two job types is a year in and still untested on the rest.

Release by shortage. Two techs quit and an apprentice is released the following Monday on three job types at once. This is the single most common cause of a bad release, and it is worth naming to yourself in the moment, because staffing pressure feels exactly like a readiness judgment while it is happening.

The unlogged release. Nobody wrote down what was granted, so six weeks later the apprentice is dispatched to something adjacent that nobody signed off on, and everyone assumes someone else approved it. Write the release down, by job type, with a date.

Full autonomy on day one of solo work. No check-in, no note review, no escalation sentence. The first hard call gets made alone by the least experienced person, at the worst moment, with the customer watching.

References

• U.S. Department of Labor, registered apprenticeship competency-based completion standards
• OSHA 29 CFR 1910.147(c)(7)(ii), demonstrated proficiency as the standard for training verification
• State licensing board requirements for apprentice supervision ratios and scope of work, which vary by state and by trade
• See related: Apprentice to Journeyman: What Changes, The Ride-Along Sequence That Actually Builds Competence, Build a Skills Matrix: Who Can Do What, Measuring Whether Your Training Is Working