Hiring now that AI does a lot of the entry-level work

Keep hiring juniors, and change what the first year is for. The entry-level tasks were never the point of the entry-level job – they were how people learned the business. Replace the tasks with supervised judgement, earlier.

By Matthew Steele4 min read28 October 2026

A senior and a junior colleague side by side at one screen in a quiet office, the junior pointing at something and the senior considering it
Illustrative scene, not a customer's premises.

What entry-level work was for

The junior who spent their first year drafting proposals, researching companies and keying data into the CRM wasn't valuable because of the drafts, the research or the data. They were valuable because by the end of the year they knew how the business sold, what a good customer looked like and what the senior people cared about – and they'd earned the right to be trusted with a decision. The tasks were how they learned. The tools have taken the tasks and left the need to learn exactly where it was.

What the first year becomes

The old first year and the new oneTwo timelines across twelve months. The old first year: doing the tasks for nine months, then first small decisions. The new first year: using the tools and checking their output against the standard for three months, running the sample checks and sitting with a senior on judgement calls for three, then owning one small agent and making the calls the tools cannot. It reaches judgement a year earlier.The old first yearDoing the tasks: drafts, research, data entry. Learning by repetition.First small decisionsThe new first yearUsing the tools, checking the outputSample checks, shadowing a seniorOwning one small agent. Making the calls the tools cannot.Month 0Month 3Month 6Month 9Month 12Checking turns out to be a better teacher than doing: fifty researched records a week against the profile teacheswhat a good customer looks like in a month. The ratio of juniors to seniors may change. The pipeline of future seniors must not.
The old first year was doing. The new one is checking, then deciding under supervision, then owning – and it gets to judgement about a year earlier.
The old first year The new first year
Months 1 to 3 Doing the tasks: drafts, research, data entry. Learning by repetition. Using the tools to do the tasks, and checking every output against the written standard. Learning the standard directly.
Months 4 to 6 More of the same, faster. Running the sample checks on a live agent for its owner. Sitting with a senior on the judgement calls the tools flag.
Months 7 to 12 First small decisions, closely watched. Owning one small agent, with the owner's card. Making the calls the tools can't, with a senior one step away.
End of year Knows the business. Trusted with a decision. Knows the business and the tools. Trusted with a decision and an agent.

Checking turns out to be a better teacher than doing. A junior who reads 50 researched company records against the profile every week learns what a good customer looks like in a month. One who researched them by hand took a year, because most of that year went on the searching.

What to hire for

Curiosity, because the tools reward people who ask them good questions. Clear writing, because a brief to a tool is writing, and so is explaining to a customer what the tool found. A willingness to check and to say "this is wrong", which we think is the single most valuable habit in a business that runs on these tools. And the things that were always the point: being pleasant to deal with, and wanting to understand the customer's business.

What not to hire for: speed at the tasks the tools do. It doesn't tell you anything any more.

The arithmetic

A team that needed three juniors to keep two seniors supplied might now need one or two. That's real, and pretending otherwise helps nobody. But the ratio is the only thing that should change. A business that stops hiring juniors because the tools do the junior work has decided, without noticing, to have no mid-level people in 5 years and no seniors in 10. Nobody's pricing that risk, and we'd say it's the most expensive line in the whole subject.

What to tell candidates

The truth: the tools do the repetitive parts, the job is to check, learn and decide earlier than it used to be, and they'll own something real within a year. Good candidates find that more attractive than a year of data entry. The ones who wanted the data entry weren't the ones you were looking for.