CoSprint for HR teams
Make TalentVisible
Not subjective ratings —objective reviews from the record of the work.
Hiring? CoSprint for hiring
- Team messengerOff
- Issue trackerOff
- DocsOff
- Design filesOff
Everyone gets a 4. So who did the work?
- Kept friendly
- A low score costs the relationship, so nobody gives one.
- The last few weeks
- A whole period reviewed from memory of its last weeks.
- Whoever was seen
- The one who spoke in meetings is remembered; the quiet fix isn't.
- 57%
of Korean workers don't trust their performance review results. Top reason: “my manager's subjective call” (67.7%)
JobKorea, 690 workers, Feb 2024 - 62%
of the variance in a rating comes from the rater's own tendencies, not the person rated
Scullen, Mount & Goff, Journal of Applied Psychology, 2000 - 74%
of employees find 360-degree peer review results unfair, biased or inaccurate
LiveCareer, 1,000 US workers, 2025 (via SHRM)
Any tool.If there's a record, we review it.
One admin account per tool, and it's connected.
Drop in your org chart and anything that says how you work. That's all I need.
All set: 312 people, 41 teams. Here's what I drafted for the growth team.
- Narrows the cause with numbers
- Tests small, decides fast
- Leads with the conclusion
- Shares what they know
Two files, and your criteria are drafted
Upload your org chart and the documents that say how you work, and get draft criteria for every role. For J's growth team, its working principles gave four.
Talk to salesWhat it reads
- Issue trackerWhat they took on and closed
- DocsPlans, retros, write-ups
- Team channelsHow they asked and aligned
- CodeChanges and reviews
- Design filesDrafts and decisions
What stays out
- Activity is never scored: no message counts, commits or hours online
- No direct messages, screens or keystrokes are collected
- Originals aren't stored at CoSprint, and CoSprint staff can't open your data
- Never used to train AI
- Narrows the cause with numbers
“Refunds up 30%: same SKU, two warehouses. Only Incheon. It's the packaging.”
May · IssueView original - Tests small, decides fast
“One region, one week, 10% of budget. If repeat orders don't move, we stop.”
Jul · DocView original - Leads with the conclusion
“Short version: push outage, not coupons. Detail below.”
Sep · Team channelView original - Shares what they knowNo evidence
No evidence in the record, so it's left empty for the reviewer.
Instead of peer review, the record of the work
“Always helpful, easy to work with.”
“Shares results quickly.”
- Answered 31 of 33 requests from other teams
- Results of 3 of 5 experiments shared after the deadline
Every criterion quotes the record, and a criterion with no evidence stays empty. Employees fill in nothing, so you can review quarterly instead of half-yearly, or monthly if you want.
Who sets the final grade?
The reviewer does. You get the grade range the record supports and a short list of unclear cases first, then pick final grades against your grade rule on the calibration screen. Pay and promotion stay with people too.
- Employee JA–BGrowth team · 4 items without evidenceSABCD
Employee LB–COps team · On a grade boundarySABCD
Employee MA–BRetention team · On a grade boundarySABCD
Employee NS–AGrowth team · 3 items without evidenceSABCD
“If they were leaving, would you fight to keep them?”
Netflix's keeper test. The talent review shows the work that stops without someone and the teams waiting on them, and the manager gives the answer. People with too little record aren't guessed at. They're listed apart.
- Topics only they hold3
Pricing experiments · attribution model · CRM segments
- Assets only they maintain2
Weekly conversion report · experiment dashboard
- Work waiting on them5
4 teams wait on the Friday report
Under 30 cites this quarter · not scored
- Employee R18/30 citesField sales, little in the toolsSend a sprint
- Employee T11/30 citesJoined 6 weeks agoSend a sprint
- Employee N24/30 citesWorks mostly in meetingsSend a sprint
Weak spots in AI use get practice on their own job
Every review reads AI use on four elements: Partition, Articulate, Inspect and Responsibility. For any element at level 2 or below, you can send a practice sprint built from that person's job.
- Partition
What to hand to AI
- Articulate
How clearly they instruct it
- Inspect
Checking what comes back
- Responsibility
Owning the final call
- Practice · InspectCheck an AI-written weekly report before it goes to the exec teamSent · in progressResult goes into the next review
- Practice · ResponsibilityShip a campaign brief drafted with AI, saying what AI didSendResult goes into the next review
The hiring record, carried into every review
For people hired through CoSprint, the hiring record stays with them beside every review, and what the reviews show sharpens the next hire.
- Hiring sprint1st of 276Strongest at narrowing the cause
- First reviewThe sprint is the baselineRead next to the record of the work
- Next hireCriteria adjustedWhat predicted good work weighs more
Questions HR teams ask first
Is this employee monitoring?
No. AI evaluates what people produced and decided in shared work tools, against your criteria. Activity such as message counts, commits and time online is never scored, and private messages, screens and keystrokes are never collected.
Which records does it use?
Only the tools and channels the company chooses to connect: issues, documents, team channels, code and design files. Direct messages are excluded.
Which tools can it connect to?
Any work tool you hold an admin account for. Sign in and approve once; the access token it issues is kept in an encrypted vault and used only to fetch the record.
Do employees need to be told?
Yes. Tell employees which records are used and why before it starts, and Jello drafts the notice for you. Other rules depend on where your people work, so check with employment counsel before rollout.
- EU
- GDPR: transparency notice, usually a DPIA
- EU AI Act: AI that evaluates workers is high-risk; inform representatives and staff first
- Germany: BetrVG §87: works council co-determination
- Korea
- Personal Information Protection Act: notice of purpose
- Workers' Participation Act, Art. 20: labor-management council consultation at 30+ employees
- Labor Standards Act, Art. 94: when work rules change to tie reviews to pay
- AI Basic Act: whether it counts as high-impact AI
- Japan
- Act on the Protection of Personal Information: notice of purpose
- Labor Standards Act, Art. 89–90: work rules and the majority representative's opinion, when reviews affect pay
- US
- New York · Connecticut · Delaware: prior notice of electronic monitoring
- California: CCPA: notice for employee data
Can we keep our peer review?
Yes. Keep peer review and read it alongside the record, or let the record take its place. How you use it is up to you.
What about roles that leave little in the tools?
What the record doesn't show stays empty rather than guessed. Roles like field sales lean more on the reviewer.
Can it run our annual appraisal?
Yes. Half-yearly and annual appraisals run the same way as quarterly or monthly reviews, and final grades follow your own grade rule, relative or absolute.
Can it run with CoSprint for hiring?
Yes. Each person's hiring record carries into their reviews, and what the reviews find feeds back into the next hire.