Fellowship selection, across rounds.
The fellowship-specific problem
Fellowships are different from one-shot scholarship awards for two reasons. The selection runs over multiple rounds, and the reviewer pool changes between rounds. The long-listers are often staff; the shortlisters are often external experts; the final panel is often the board. Three different sets of eyes reading the same applications, and historically, three different versions of the truth.
By the time the final panel meets, the staff longlister's reasoning is buried in a spreadsheet column from six weeks ago. The shortlist scores arrived as a CSV from a consortium partner without rationale. The panel ends up re-reading half the applications because nobody trusts the upstream scoring enough to skip the read.
That re-reading is where fellowships break. The panel runs out of time. The last 20 applications get a faster, looser treatment than the first 20. Whether that's biased depends on which 20 were last in the queue.
What FairLens does for multi-round cycles
One rubric, three rounds
The same criteria, weighting, and definitions apply at longlist, shortlist, and final panel. Reviewers at every stage see the same structure; what changes is what gets cut, not how it's measured.
Scores carry forward with their evidence
When an application advances from longlist to shortlist, the previous round's scoring and evidence citations come with it. The shortlister doesn't start cold; they start with the longlister's reasoning visible and contestable.
Calibration before the cycle starts
Reviewers grade a small set of synthetic applicants before scoring any real ones. The calibration agent surfaces where reviewers disagree with each other, and where they disagree with the published rubric, so those gaps get resolved before they affect real candidates.
For the selection panel vs. for the sponsor
What the panel sees
For each shortlisted applicant: the full evidence record from earlier rounds, the upstream reviewers' scores and reasoning, the panel-specific structured summary, and the AI's suggested score with citations. Panel time goes into decisions, not re-reading.
Disagreements between upstream reviewers and the AI are surfaced explicitly, so the panel can spend its discussion time on the genuinely close calls.
What the sponsor or board gets
After the cycle: a cohort report showing the longlist-to-shortlist-to-award funnel, score distribution at each stage, reviewer-disagreement map, and demographic fairness analysis. The sponsor can see not just who was selected but how the selection happened.
Two cycles in, the report includes year-over-year comparisons. By the third cycle, the data starts to inform the next call: which criteria correlate with which outcomes, which scoring patterns predicted which fellow successes.
Post-award is included
Fellows get a magic-link portal after the award. Quarterly check-ins, publications, employment milestones, and cohort-level convenings all log back to the same record the panel reviewed. Twelve and twenty-four months in, the donor report writes itself.
For fellowship operators specifically, the post-award arc matters more than for one-shot grants: the fellowship brand is built on cohort outcomes, and a structured outcome record is what lets you demonstrate that brand to the next funder.
What FairLens isn't
FairLens isn't a cohort-management system. Once fellows are selected, scheduling convenings, managing mentorship pairings, and tracking attendance live somewhere else. The post-award outcome tracking is the record-of-decision layer; the operational cadence is up to you.
It also isn't a substitute for an independent selection committee. The AI surfaces evidence and suggests scores; the committee decides. Every override is logged with a reason. If the panel wants to advance the applicant the AI ranked 47th, they can. They just have to write down why.
Run your next fellowship cohort on FairLens.