InRoots — An AI productivity app, not a monitoring tool

People

Who's on now18 / 25

Live · Friday 11:20

Lina K.signed in

Omar S.on-call

Philip R.late 09:42

Sara M.signed in

Rabie A.pto

Ahmed T.signed off

Presence & devicesdesktop · mobile

Status from the chat client, no agent installed

Lina K.online

Omar S.busy

Philip R.away

Sara M.online

Rabie A.offline

Ahmed T.away

Top performersthis month

Admin-only; one list, built from the same score everyone can see

1

Sara M.86%

2

Omar S.81%

3

Lina K.72%

4

Philip R.69%

5

Ahmed T.66%

Recognition only; no pay decision is attached. Under 25% → coaching first, not penalties.

Retention riskadmins only · advisory

Early, quiet, and always with a reason

Ahmed T.watch

Score down 3 weeks · 2 short days · no kudos 30d → suggested 1:1 this week

Philip R.steady

Lina K.rising

Never shown to peers. Advisory only; a suggested 1:1, never an action.

PTO & leaveauto-approved 3 of 4

Requested in chat, rules decide

Rabie A. · todayapproved

Sara M. · Mon–Tueapproved

Philip R. · sick, Wedto lead

Balances, coverage and blackout dates checked before anyone is asked.

Work

Efficiency▲ +11%

8 weeks

Whole team72%

W1W4W8

Transparencysame score, every seat

Own view = lead's view = team's view, week by week

Lina K.▲ 61 → 72

W1W8

Weights 40·25·15·10·10, every change audited. Anyone can type /pulse me.

Productivity▲ 14 tasks this week

Tasks done per person, this week

Sara M.

9

Philip R.

7

Lina K.

6

Omar S.

4

Ahmed T.

3

Tasks done▲ 9 → 14

Per week, whole team

W1W8

Counted from standups and cross-checked with Jira, GitHub and GitLab activity.

Hours workedthis week

Per person · short days flagged · PTO balance

HoursShortPTO left

Lina K.38h 20m—12 d

Omar S.44h 05m—6 d

Philip R.31h 10m2 days9 d

Sara M.39h 40m—15 d

Ahmed T.29h 50m3 days4 d

Goalson track · above · below

Against personal and team objectives, this month

Sara M. · redesign onboardingabove

Omar S. · cut deploy time 30%on track

Lina K. · 12 tickets / monthon track

Philip R. · 8 demos bookedbelow

Ahmed T. · a11y auditbelow

Below expectation triggers a coaching nudge to the person first, then the lead.

My goals & KPILina K.

Personal view · recent activity

12 tickets / mo

9/12

Review 4 PRs / wk

4/4

KPI this month

72%

11:02closed #1031 · +1 Progress

10:30kudos from Omar · +1 Responsibility

09:05signed in · office

Work locationconsent-gated

Hours-weighted, this week

Office 54%Home 38%On the go 8%

IP classified on-host against a local database you refresh weekly. Nothing is sent out. Employees opt in and can see their own record.

Operations

Blockers & escalation2 open

2-hour SLA, the chain walks itself

Invoice export fails on large accounts26h

Lina → Omar (routed) → lead if silent 2h

Staging cert renewal1h

Ahmed → nudged at 2h → chat, WhatsApp, email until owned

Support tickets & SLA7 open · 1 escalated

Severity low, medium or high; unacknowledged high severity pages the on-call within minutes

#1042 Login loop on iOShigh · escalated

#1039 Invoice PDF blankmedium · 1h open

#1037 Add SSO for tenant Xlow · 22h open

#1031 Export CSV encodingresolved · 3h

High severity pages the on-call; no acknowledgement in 5 minutes pages the backup, and at 15 minutes it reaches the fallback room and WhatsApp. Resolution credit goes to whoever actually fixed it.

Campaignsfeedback · ack · consent · votes

Sent in chat, tracked to the last person

Policy ack · v4

24/25

IP consent

22/25

Offsite date vote

16/25

Q3 feedback

10/25

Non-responders get one reminder per day until it lands, in their own channel.

Audit log & task statuslive

Every decision, every change, as it happens

11:20:14 pto.auto_approve sara.m 2d · coverage ok

11:21:21 task.status #1031 → done · lina.k

11:22:28 blocker.route invoice-export → omar.s

11:23:35 consent.sign ip-policy v4 · ahmed.t

11:24:42 ticket.escalate #1042 high · on-call paged

11:25:49 oncall.handoff omar.s → rabie.a

Intelligence

Adaptive learningtrends · flags

Baselines per person, role profiles, targets that move with the team

Trend · Lina▲ sign-in earlier 3 wks

Trend · team▲ blockers cleared 2× faster

Flag · Philippresence check missed 3×

Flag · Ahmedsign-in via unknown VPN

Drift monitormodel re-checked weekly

Flags go to the person first with the evidence. Nothing is silent.

Neural enginetrained natively on your data

Decisions and auto-approvals, with confidence and a reason

PTO auto-approved · Sara0.97 · coverage ok

Blocker routed · Omar0.91 · owns module

Sick request · Philip0.62 · to lead

Anomaly · Ahmed VPN0.88 · flagged

Runs on your model, in your perimeter. Below the confidence bar it asks a human.

Bots shaped by your instructionspresets + your own

Personas shaped by the instruction text you write; each answers in its function's language

TETechnicalDEDevOpsSESecurityHRHRFIFinancePRProjectSASalesONOnboarding+Your own

Behaviour and automations are settings, with manager and admin overrides. No web browsing or retrieval; the reply comes from the provider you configured.

Integrations

Chat & calendarone brain, every channel

Post where you already talk

K9 Chatnative · E2EE

Slackbridged

Microsoft Teamsbridged

WhatsAppon-call reach

Google Calendarvia n8n

Microsoft 365via n8n

Work & cloudverified against real work

Standups cross-checked against Jira, GitHub and GitLab activity; your cloud managed from chat

Jiraactivity

GitLabactivity

GitHubactivity

Bitbucketsoon

VPSienative

Ops, observability & automationevents out, actions in

Every decision is an event you can ship to your stack, and any webhook can trigger InRoots

n8nnative

ZapierAPI · webhooks

Splunkevents out

Grafanaevents out

PagerDutyevents out

Datadogevents out

SendGridemail

Any SMTPemail

CustomAPI · webhooks

Bring your own model16 providers · air-gapped with Ollama, voice off

Point InRoots at any of these, or at a local model. Standup text never has to touch a public API.

Anthropic

OpenAI

Gemini

Mistral

Ollamalocal

Groq

DeepSeek

Perplexity

Cohere

OpenRouter

Together

Cerebras

Hugging Face

Hyperbolic

Cloudflare AI

n8nyour pipeline

How it works

Four steps, one minute a day

1

Post or speak

"Signed in. Shipped the export fix. Blocked on invoice PDFs." One line in K9 Chat, Slack or Teams, or say it to the assistant. Typos are forgiven.

2

AI reads and verifies

Intent, hours, tasks, blockers and location are parsed, then cross-checked against Jira, GitHub and GitLab activity. Presence is confirmed with a light ping, never a silent screenshot.

3

Goals and coaching

Personal objectives get follow-ups. Stuck tasks get a nudge. Improvement suggestions cite the person's own work. Managers get a weekly digest instead of a daily hover.

4

One fair score, seen by all

Five weighted pillars, per-role profiles, a line-by-line breakdown behind every number. The employee sees it first. "/pulse me" in chat returns it any time.

Sovereignty

Growth you own. Data you keep.

Bring your own model

16 providers including Anthropic, OpenAI, Gemini, Mistral and Groq, or point it at a local Ollama and go fully air-gapped with spoken replies off. Standup text never has to touch a public API.

On-host geo and IP intelligence

A local IP database — keyless db-ip Lite or MaxMind, refreshed weekly by a script you run — classifies office, home and mobile on-host. Nothing is looked up externally. Location is consent-gated and visible to the employee.

Signed consent and immutable audit

Policies are versioned and e-signed with a tamper-evident hash. Every decision, weight change and auto-approval lands in an audit log you can export.

Self-hosted or hosted

One container, SQLite or Postgres, deploys anywhere. Or let us host it. Same code, same features, your choice of where the data sleeps.

No telemetry

The product does not phone home. No usage beacons, no crash reporters calling out, no hidden analytics. Zero bytes leave unless you configure a destination.

Minimal token footprint

Daily call budgets, memoisation and deterministic scoring keep inference cost near zero even on paid providers. The score is math; the AI only explains it.