Outcome-based workforce analytics

See the work behind the org chart.

Clarifice reconstructs the real projects and contributions inside your company from the tools your teams already use — and shows the evidence behind every conclusion. Built to support promotion, leveling, and performance decisions. Never to automate them.

Metadata only·content never stored·human-in-the-loop·every number traces to its source
raw tool exhaust  →  reconstructed projects
/ the problem

Performance is decided on memory and visibility — not contribution.

The person who quietly unblocked three teams, paid down the tech debt everyone depended on, or drove a project that never lived in a ticket — their work is the easiest to miss and the hardest to defend at calibration. Reviews lean on what a manager happened to see, who spoke up, and what shipped most recently. Clarifice reconstructs the fuller picture, with receipts, so the conversation starts from evidence instead of recollection.

/ how it works

Connect, reconstruct, decide — with the evidence attached.

01 — CONNECT

Read-only, metadata-only

Connectors to Slack, GitHub, Jira, Google Workspace, Claude Enterprise, email, and your HRIS. Content is read transiently to extract work-facts; raw message bodies are never stored.

02 — RECONSTRUCT

The projects nobody tracked

Identity resolution and discovery reassemble each initiative: who kicked it off, who shipped it, who reviewed, what it cost, whether it landed — including the organic work that never opened a ticket. Membership overlaps; it isn't a rigid partition.

03 — DECIDE

Explainable, cohort-relative

Managers get a calm, comparative view scoped to their own team. Every claim links to the raw artifacts behind it, so a person can be shown exactly why. A human always makes the call.

/ proof

Every conclusion is traceable to its sources.

No black-box scores. A project Clarifice reconstructs is a set of real artifacts you can click into — the Slack thread that started it, the PRs that built it, the doc that specced it, the metric that proved it. Fan-out means one artifact can support more than one project; nothing is forced into a tidy bucket. If the evidence isn't there, Clarifice says so rather than inventing a story.

shipped

Swarm battery-swap automation

reconstructed from 11 artifacts · 4 people · 3 channels
Evidence
doc02-16Design: swap-station carousel & state machine
slack03-05#battery-ops: connector pins chewed up after ~30 cycles
pr03-08field-ops#3004: state machine (idle/detect/eject/verify)
metric04-08Live at Hendricks: 85s avg swap, 0 failures / 140 cycles
9
source systems unified — from chat to code to calendar
0
message bodies stored at rest
100%
of derived facts cite the raw artifacts behind them
1
human in the loop — always, by design

Start from evidence, not recollection.

See Clarifice run on a sample company and walk through how a real review conversation would change.

Product

The contribution story, reconstructed and cited.

Clarifice turns the exhaust your tools already produce into an explainable view of who drove what — the projects, the follow-through, the mentorship, and the cost — each one grounded in artifacts a person can inspect.

/ what it reconstructs

Six views, one source of truth.

Projects & contribution

Initiatives reassembled from scattered signal, with weighted attribution — lead, contributor, reviewer — and the evidence for each.

Follow-through

Commitments made and kept, across tickets and work email — the quiet reliability that never shows up in a ticket count.

Mentorship & enablement

Reviews, design feedback, pairing, and in-channel help directed at more-junior teammates — "growing others," made visible.

Cost & ROI

An opportunistic ledger — AI tokens, meeting time, person-effort — so impact can be weighed against what it took to get there.

AI leverage

How a person uses AI to multiply their output, from usage metadata — never a productivity cudgel, a picture of augmentation.

Outcomes shipped

Did it land? Deploys, experiments, and adoption tie a project to a result — or honestly show that it stalled.

/ the method

Discovery, not just dashboards.

Identity resolution stitches a person's Slack handle, GitHub login, Jira account, and email into one identity — and abstains when it isn't sure, rather than guessing. Structural + topical discovery then reconstruct projects from the join keys that exist (epics, ticket references, dedicated channels) and from the organic work that has none — a quiet cross-cutting cleanup, a problem someone just solved. Communities overlap, so a single artifact can belong to more than one project.

Everything is incremental: understanding evolves as new data arrives, and is validated at every checkpoint — not assembled once at the end.

connectors

Nine sources, one contract

chatslackmessages, threads, reactions, channels
codegithubPRs, reviews, commits, deploys
planjiratickets, epics, status, due dates
docsgoogledocs, revisions, comments
calcalendarmeetings, attendance, prep time
aiclaudeusage & work-facts (metadata)
mailemailwork threads only — 1:1 excluded
orghrisdirectory, reporting lines, levels

Reconstruction you can defend line by line.

The fastest way to understand Clarifice is to watch it rebuild a project from raw signal and click into the evidence.

Integrations

Works with the stack you already run.

Every class of information has more than one supported source, so you're never locked to a single vendor — and one tool can feed several classes. Bring your own homegrown system too: a custom connector emits the same normalized data and passes the same governance gates as a first-party one.

/ 18 vendors · 11 classes

Bring the stack you already run.

Every connector listed is built & verified — turn it on with credentials.

Identity & HRIS

Org chart, the account graph that stitches every other source to a person, and admin scope.

IDSCIM 2.0 directory (Okta, Entra ID, Google, OneLogin…)

AI usage

Which work was AI-assisted, and the token/cost footprint — content read transiently, never stored.

Claude EnterpriseAIOpenAI / ChatGPT Enterprise

Team chat

Public-channel discussion, threads and reactions — collaboration and enablement signal (never DMs).

TeMicrosoft TeamsSlSlack

Code & delivery

PRs, reviews and deploys — what shipped, who reviewed, and the path to production.

GitHubGitLab

Incidents & on-call

On-call load, MTTR and change-failure — operational reliability signal.

PDPagerDuty

Project management

Tickets, epics and status — the planned-work spine projects are reconstructed around.

GitHubGitLabJiraLinearNotion

Docs & knowledge

Doc authorship, edits and comments — durable knowledge and design contribution.

GitHubGoogle WorkspaceMSMicrosoft 365 / OutlookNotion

Email

Work-scope threading and routing metadata (personal-scope bodies are never read).

GmailGoogle WorkspaceMSMicrosoft 365 / Outlook

Calendar

Meeting invites and pairing — coordination cost and who worked with whom.

Google CalendarGoogle WorkspaceMSMicrosoft 365 / Outlook

Experiments

Launched experiments and their measured impact — did the work move a metric.

ExIn-house experiment logLDLaunchDarkly

Availability

Time-off / working days — so load and consistency are fair, not penalized for PTO.

BHBambooHRHRIn-house HR feed

One tool, several classes

A vendor isn't limited to one kind of data. GitHub, GitLab, Google Workspace, Notion, Microsoft 365 / Outlook each feed multiple classes — so a single connection can light up several at once. Switch to “By provider” above to see exactly what each one covers.

Homegrown & niche sources — custom connectors

Run something we don't list yet? A custom connector pushes your normalized data (or lets us pull it), emitting the exact same shape as a first-party source and passing the same metadata-only, work-facts-only governance gates.

Deliberately excluded — governance, not a gap
  • Slack / Teams private DMs — 1:1 DMs are not a reliable work signal and carry NLRA / privacy risk.
  • Personal-scope email bodies — Metadata only; the connector never fetches personal-scope message bodies.
  • Sentiment / mood / health / competence — Out of scope by design — we reconstruct work facts, never inferred states.

Don't see your tool?

If it holds work metadata, we can almost certainly ingest it — a custom connector takes a homegrown or niche source and maps it to the same normalized shape.

Security & Trust

We built the guardrails first.

Workforce analytics earns trust or it doesn't get deployed. Clarifice's boundaries aren't settings you can quietly turn off — they're the design. Here is exactly what is in scope, what is never touched, and how it runs in your environment.

/ scope

What's in scope — and what is never touched.

✓ In scope — work signal
  • +Public & private team channels, repos, and drives — the shared surfaces where work happens
  • +Work group & distribution-list email threads
  • +Tickets, pull requests, deploys, experiments
  • +Meeting metadata — who, when, how long (not transcripts)
  • +AI usage metadata & extracted work-facts
  • +Directory & reporting lines from your HRIS
✕ Never — out of scope
  • −Message content at rest — bodies are read transiently, then dropped
  • −1:1 DMs and personal email bodies
  • −Sentiment, mood, competence, or flight-risk inference
  • −Health data as any scoring input
  • −Protected-topic & NLRA-protected channels — excluded by content, not by name
  • −Keystroke, screen, or activity monitoring
/ principles

The rules that don't have an off switch.

Metadata-only storage

Content is read transiently to extract work-facts. Raw bodies are never persisted — we store the derived fact and a reference to its source, not the source itself.

Work-facts only

What was done, whether it shipped, what was decided, who contributed. Never how someone feels, how "good" they seem, or what they talked about off-topic.

Human-in-the-loop

Clarifice produces decision support, not verdicts. No automated promotion, PIP, or exit. A person always decides, with the evidence in front of them.

Provenance, always

Every derived fact cites the raw artifacts behind it — zero fabricated. If the evidence is thin, the view says so instead of guessing.

Absence is context, never a penalty

Time off, leave, and quiet periods are never scored against anyone. People are only ever compared within their own cohort.

Least-visibility access, fully audited

Managers see only their own subtree. Every view of another person's data is logged. Sensitive and legally-protected channels are excluded by a content classifier before anything is stored.

/ deployment

Runs where your data-residency rules require.

Managed

We host it

Fastest to stand up. Per-tenant isolation; metadata only; every access audited. Right for teams comfortable with a SOC-2-style managed service.

Collector

Raw stays in your cloud

Connectors and extraction run in your environment; only the derived model leaves. Raw signal never transits our infrastructure.

BYOC

Entirely your environment

The whole pipeline runs in your cloud, with your own model keys and retention policy. We provide the software and control plane only — for the most regulated buyers.

Bring your security team to the first call.

We'd rather answer the hard questions early. Ask us anything about storage, scope, retention, and the exclusion model.

Pricing

Priced per person, not per data point.

Bounded by headcount — the thing you actually plan around — not by how much your teams communicate. Figures below are illustrative for this preview.

Team
$8 / person / mo

For a single org up to ~150 people getting started.

  • Core project reconstruction & contribution views
  • Slack, GitHub, Jira, Google connectors
  • Manager dashboards with full provenance
  • Cohort-relative comparison & audit log
Business
$14 / person / mo

For scaling orgs that need every source and SSO.

  • Everything in Team, plus
  • All nine connectors incl. Claude, email, HRIS
  • SSO / SAML, role-based access, audit export
  • Follow-through, mentorship, cost & AI-leverage views
  • Leveling framework & calibration support
Enterprise
Let's talk

For regulated buyers and strict residency requirements.

  • Collector or BYOC deployment
  • Data residency & custom retention windows
  • Bring-your-own model keys
  • Security review, DPA, and SLA
  • Dedicated support & onboarding

Prices shown are illustrative placeholders for this preview and are not an offer. Final pricing depends on deployment model and connector scope.

Company

Evaluation should be evidence, not vibes.

The best contributors are often the least visible. Careers turn on calibration conversations that lean on whoever was loudest and whatever shipped most recently. We think that's both unfair and fixable — if, and only if, it's done with hard guardrails.

/ our stance

Why we lead with the constraints.

Plenty of tools in this space start from "capture everything" and bolt on privacy later. We did it the other way around: the boundaries — metadata-only, work-facts-only, human-in-the-loop, provenance on every claim, protected-activity exclusion — came first, and the product is built inside them. A system that helps decide people's careers has to be legible to the people it describes. If we can't show someone exactly why, we don't show it at all.

Support, never automate

The hardest calls about people should be made by people. Our job is to bring the evidence, cited and in context.

Legible by default

Anything Clarifice asserts, a person can trace back to its source and challenge. No black boxes in career decisions.

See it on a sample company.

The clearest pitch is a walkthrough. We'll show reconstruction, provenance, and the guardrails in fifteen minutes.

Request a demo

Fifteen minutes, on a sample company.

No connectors required to see it. We'll walk through reconstruction, provenance, and the trust model on synthetic data — then talk about your environment.

Preview form — submissions aren't sent anywhere. This is a design prototype.

Thanks — you're on the list.

In the real product we'd email you to book a time. For now, this confirms the flow works.

/ what happens next
  • 01We book fifteen minutes and run Clarifice live on a synthetic company — no access to your data.
  • 02You see a project reconstructed from raw signal, and click into the evidence behind it.
  • 03We walk your security team through storage, scope, and the exclusion model.
  • 04If it's a fit, we scope connectors and the deployment model that meets your residency rules.