AI and jobs: should your SMB rethink junior salaries?

Two pieces of research, published five days apart, sketch a genuine paradox for any business owner watching how AI reshapes their team. On August 17, 2026, Lloyds Bank published its monthly Business Barometer, based on a survey of 1,200 companies polled in July from a panel of 20,000 UK firms: 54% say AI has created new jobs within their organization. Five days earlier, on August 12, 2026, Germany’s Ifo Institute published the opposite-sounding finding, drawn from its June 2026 survey of over 3,000 German companies already using AI: 50.8% expect to pay new graduates and staff with under five years’ experience less than they otherwise would. Both studies are real, both are recent, and neither is wrong — together they say something more useful than either headline alone.

What each study actually found

• →Lloyds Bank Business Barometer (UK, published August 17, 2026): 54% of surveyed businesses say AI has created new jobs; the survey does not measure whether other roles were cut in parallel, so the net employment effect stays an open question.
• →Same barometer: 58% of businesses plan to increase AI-related training investment over the coming year — 43% are introducing AI-skills training, 32% are expanding existing programs, 24% now weigh AI skills in hiring decisions, and 21% are creating new AI-specific roles.
• →Ifo Institute (Germany, published August 12, 2026): 50.8% of AI-using companies expect lower starting salaries for graduates and staff with under five years’ experience; around 40% expect cuts even for staff with five-plus years’ experience. Service-sector firms report the sharpest gap — 53.3% for juniors versus 44.2% for the more experienced.
• →OECD Employment Outlook 2026 (published July 7, 2026): AI is not, at this stage, causing a widespread decline in overall employment across member countries — but it is reshaping the skills employers look for, and slowing hiring specifically at the junior end in AI-exposed sectors.

Why the two findings don’t actually contradict each other

AI can create jobs and value in aggregate while the newest, least-experienced employees see none of the upside — that is precisely what both studies, read together, describe. A 2023 NBER study by Erik Brynjolfsson, Danielle Li and Lindsey Raymond, tracking 5,179 customer support agents through the staggered rollout of a generative AI assistant, found the tool lifted average productivity by 14% — but the gain was 34% for novice and lower-skilled agents, against almost nothing for the most experienced ones. The AI effectively distributed the working habits of the best agents to the newest ones, moving them down the experience curve faster. That is the opposite conclusion from “junior work is now worth less”: the data says a junior equipped with the right AI tool closes the gap with a senior far quicker than before. Treating a junior’s AI-assisted output as a reason to underpay them, rather than as a reason to accelerate their progression, gets the economics of these studies backwards.

What it means for your SMB

French SMBs make most of their hiring and training-budget decisions for the year around the September rentrée — which makes this exactly the season these findings matter. If you deploy an AI agent on an entry-level function, the design choice you make determines which study describes your business: an agent handed to a junior as a coach that shows them how a senior would resolve a case is investment; an agent that simply replaces the junior’s judgment is substitution. The difference shows up directly in how fast that junior becomes valuable — and in what you can defensibly pay them next year. This is the exact logic behind an AI support desk built to be supervised by a junior agent rather than to bypass them: it puts senior-level responses within reach of someone six months into the job, instead of two years.

Four concrete moves before the rentrée

• →Check, inside your own team, whether AI has actually created new tasks or roles this year, or simply moved existing time around — go by your own numbers, not by general sentiment.
• →Set aside a real training budget for your juniors’ AI skills rather than leaving everyone to figure it out informally — 43% of Lloyds-surveyed businesses are formalizing this now, 32% are expanding what they already had.
• →Do not index a junior’s pay to “AI now does part of the job” — measure instead whether AI is getting them to a confirmed level faster, and set compensation on that trajectory.
• →If you deploy an AI agent on an entry-level function — support, email triage, sales qualification — design it as a coaching tool your junior supervises and learns from, not a pure substitute; the two choices produce very different outcomes a year from now.

The businesses that get this right this fall are not the ones deploying AI fastest, but the ones deciding early whether it is there to grow their juniors or to justify paying them less — the research says only one of those choices actually pays off.