Caliber scores candidates on four levels of AI leverage: L1 Manual, L2 Assisted, L3 Augmented, L4 Architect. Levels are a maturity stage, not a percentile. Different roles need different levels. Pick the one your role needs.
A year ago, claiming you use AI was a hiring differentiator. Today every candidate says it. The question stopped being whether someone uses AI and started being how. Caliber scores candidates on a four-level rubric of how much leverage they actually get from AI on real work. Here is the rubric.
L1: Manual
Does the job by hand. AI is absent, or used only cosmetically (a sentence here, a phrasing fix there). The work gets done, but at the original cost. Sometimes legitimate: high-trust documents, regulated environments, or the candidate's first month before they have tooling set up. Sometimes a tell that the candidate has not been pushed.
L2: Assisted
Uses AI as a faster search box. Single prompts, single outputs, no reuse. The pattern is "paste this in, copy that out." Faster than L1, but no compounding. The work product is brittle: ask the candidate to do it again with new inputs and they start over. Verification is rare. Most candidates in the AI conversation today are L2 and don't know it.
L3: Augmented
Built reusable prompts and light systems. Reads as the operator who has been doing this for a while and gradually evolved their workflow. The prompt itself is an artifact. Verification is built in: spot checks on numbers, flags on anomalies, humans on the edges. Same task next week is faster, not just as fast.
L4: Architect
Redesigned the workflow around AI. The candidate connected tools, built a skill or agent, and now spends their time on judgment, not typing. The recurring pieces of the job run themselves. The candidate can explain the tradeoffs: why this part is automated, why this part is human, what fails and how they catch it. They are not faster at the old job. They have built a different job.
Why levels, not scores
The levels are not a percentile. Two L3 sales operators in the same week look almost identical. Two L4s diverge based on what they chose to architect away. Levels capture a maturity stage. A score in the middle of an arbitrary scale tells you nothing actionable.
The hiring use
Different roles need different levels right now. A senior individual contributor at a fifty-person company hiring into a high-leverage motion (sales, ops, marketing, recruiting) is best at L3 or better. A first-year analyst at the same company might come in at L1 and grow. The question is what level your role needs, not what is impressive in the abstract.
Where Caliber fits
We built Caliber to score this on a real scenario, not a quiz. Pick a role, read the scenario, watch four candidates do the same job at L1 through L4, and see what good looks like at the level you are hiring for. The interactive sample on the home page runs through Sales, Operations, and Marketing scenarios. The first three pilot orgs per quarter run as a manual service at no charge while we onboard the self-serve product.