Key takeaways
- Publish the portal when the window opens and unpublish it when it closes. An unpublished portal can't receive applications.
- Eligibility rules are hard rules; anything an application doesn't state becomes a question, not a rejection.
- Mentors and industry reviewers join as guest experts with expiring access in the first release.
- Every decision, override and rubric version is kept, so selection can be explained after the cohort starts.
Planned feature: not in Delsivo yet, and details may change before it ships.
What makes cohort selection different from fund screening
A fund screens a steady stream. A programme gets its applications in a burst, with a deadline, and has a few weeks to turn them into a cohort. Eligibility matters more: incorporation, stage, region and full-time founders are often fixed rules from funders or partners. The review team is bigger and changes between cohorts, and the programme often has to explain its choices to rejected applicants, sponsors and its own board.
That changes what the tool must do. Consistency across reviewers and a clear record of why each team was picked matter as much as speed.
Set up accelerator application review before the window opens
- 1Write eligibility as hard rules, what you look for as weighted preferences, and the questions to ask when an application is silent (Pilot).
- 2Brand the portal with your logo, colours and copy on a subdomain, and put the closing date on the landing page in your own words (Pilot).
- 3Build the application form. The pilot form supports text, number, select and file fields and required questions; conditional questions come in the first release.
- 4Create private referral links for partners, alumni and investors so each application keeps its source (Pilot).
- 5Give your selection team roles: an admin for settings, analysts for the queue, a partner who approves decisions and messages (Pilot).
- 6Publish the portal on the day the window opens.
During the window
Each application is read and screened as it arrives, so the queue is useful from the first week rather than after the deadline. Teams that apply twice, or a founder who submits for two companies, are suggested as possible duplicates for an analyst to confirm. Applicants can invite co-founders to the application and save their progress.
Unknown values build up a list of questions per applicant. You can send them as a short form while the window is still open, so applicants have time to answer before review starts.
From longlist to selection
| Stage | What happens | Release |
|---|---|---|
| Longlist | Applications that meet eligibility, ordered by fit, with mismatches reviewed by a person | Pilot |
| Team review | Analysts work the queue and record reasons for each advance or pass | Pilot |
| Outside reviewers | Mentors see only the applications shared with them, until access expires | First release |
| Interview | The same live AI interview, or its text version, for every shortlisted team | First release |
| Selection | Your programme team decides, and each decision keeps its reason | Pilot |
| Selection packet | A frozen packet with independent reviews for a selection committee | Later |
Limits for programme teams
- Applications need a PDF deck. PowerPoint support follows once extraction quality is measured.
- English first at launch.
- How guest reviewers record their views is being shaped with design partners; there is no built-in judges' scoring sheet today.
- Scout links with per-scout attribution are planned for later. Referral links work in the pilot.
- Messages to applicants are approved by a person, including passes, so plan time for that after selection.