A team accountability system for agencies. One company, one instance.

Project tools store work. They don’t make anyone do it.

An editor goes quiet, a deliverable moves a day, then three, and you hear about it from the client. Every person sits on one pyramid, coloured by what they actually delivered, computed from real tasks and real due dates with nothing to fill in. When someone needs chasing, the message is written for you from that task's own record, it arrives as a notification on their phone, and the chase is logged.

In production since 10 July 2026, running a Nairobi marketing agency every working day. 12 client accounts, 25 content calendars, 1,415 tasks moved across in one pass. Your own server, your own database.

The Kazi team pyramid: an org tree of avatar cards labelled on track, watch, behind and no data, with overdue counts
Demo workspace, fictional data

The problem

The tool is not the problem. You are the follow-up system.

The board is full of cards and the client got nothing this week.

An editor goes quiet on a Tuesday. You hear about it on Thursday, from the client.

Two people carry the delivery. You know which two. They are also the two who leave.

Every approval routes through you, so your calendar is the bottleneck and your memory is the system.

What changes

The week you have now. The week you get.

Nothing below needs a new habit from your team. They work the tasks they already work, and the record writes itself off that.

Now

  • You are the follow-up system. Nothing moves unless you ask twice.
  • You find out work slipped when the client complains.
  • You feel who your good people are. You cannot prove it.
  • Your manager cannot chase what he cannot see, so you hold the whole picture.
  • Clients ask for updates you have to go and find.
  • You already pay for a project tool and the week still looks like this.

Once it is running

  • Chasing costs one click. The message is written from that task’s own state, how many days it has slipped and what it is sitting in, and it arrives as a notification on their phone.
  • The day work slips, the card changes colour and carries the overdue count. Nobody had to tell you.
  • Completion, on-time rate and a five-week trend sit under every name, computed from the work.
  • Managers you nominate open the same screen you do, including one shared list of what is on fire and what is blocked across the team.
  • Clients watch their own calendar in their own portal and stop asking.
  • Every chase is on the record: who asked, how, and when. The Monday conversation runs on the record instead of your memory.

Storing the work was never the hard part.

The pyramid

Your whole agency on one screen, coloured by delivery.

Who reports to who, drawn as a tree, with your editors, account leads and designers on it. Each card takes its colour from the tasks they already work in. Click anyone and the evidence is underneath: their open work, what is overdue, and by how long.

The rule, in full.

  • On track. Active work, nothing overdue.
  • Watch. A task or two slipping, worth a look.
  • Behind. Half or more of their open work is overdue, or more late than on-time deliveries in the last 14 days.
  • No data. No dated work in the window, so no claim is made.

No forms. No timesheets. Nobody self-reports. The colour comes from real tasks, real due dates and real completions, so a tidy status update moves nothing. Every card states its reason in plain English.

The manager dashboard team tab: the pyramid above per-person performance cards with delegation level badges and a top performer panel
Demo workspace, fictional data

Management, computed

Managing people becomes reading a screen, not remembering a week.

Performance metrics

Completion, on-time rate and a five-week trend per person, computed from the work itself every time you look. Nobody fills anything in, so a tidy status update moves nothing.

Role Anchors

One short document per person that defines the job: key result areas, a benchmark, a goal. The definition of the role sits next to the evidence of the role.

The delegation ladder

Four levels, printed on every profile. L4, wait to be told. L3, recommend then act. L2, act then report. L1, act independently. Everyone can see the level they are on and the one above it.

The chase, logged

Click the task, and the check-in is drafted from that task's own state: what it is, how many days it has slipped, what it is waiting in. Send it as a task comment or a DM and it pushes to their phone. Each one is recorded, so the card reads “Nudged (2)” with who asked, how, and when. Nothing sends itself, and a person reads the draft first.

A team member's personal scoreboard: score gauge, delegation badge, due tasks and five-week statistics, with the same pyramid the manager sees
Demo workspace, fictional data Everyone sees their own numbers, the same ones their manager sees.

The whole workspace

Nobody types a status. The system reads the work.

  • Content calendars with list, board, calendar and table views
  • Custom fields, checklists, attachments and file approvals
  • Chat, channels and DMs with images and formatting
  • Client portals: clients see their calendar, nothing else
  • Docs and SOPs, searchable, with AI answers citing sources
  • Real push notifications on phones and laptops
  • Recurring tasks, automations and goal tracking
  • Installs as an app on iPhone, Android, Mac and Windows
  • Calendar feeds, a migration importer and an API for AI tools
A content board with pipeline columns from Open through Published
Demo workspace, fictional data
The same workspace on a phone: bottom tabs and the personal dashboard
Demo workspace, fictional data

Switching

Nothing gets retyped.

A full workspace moved across in one pass: 1,415 tasks, 1,494 comments, 6 spaces, 90 lists, 97 custom fields, and the docs. Re-runs gap-fill and never overwrite: on a live instance, 30 values the team had typed by hand came through a full re-import untouched, 30 before and 30 after. Whatever you run today, the tasks, the comments, the fields and the docs come over, and you check them with me before anyone switches.

Your instance, your data

This is not SaaS. That is the point.

One agency per installation. Your client list, your rates and your team's delivery record never share a database, a server or a login screen with anyone else's.

  • Hosted for you, or on your own machine on the Scale plan
  • Flat pricing per instance. Adding a person never raises the bill
  • API keys stay on the server, never in a browser
  • AI switches on with a key you control, and the workspace runs at full strength without one

When the money asks

Investors read your accounts. Nobody reads how you run.

Funds, accelerators and boards arrive at the same question in the end: is this a company, or one person holding it together? Working in one system answers that as a by-product of the work, not as a scramble the week before a meeting.

Reviewing the operational causes behind Kenyan companies that had raised real money and still failed, PwC put it plainly: investors lean on financial, legal and tax due diligence while overlooking the operational mechanics of the businesses they fund, which lets weakness persist until it surfaces as a liquidity crisis. That operational layer is the one this software is built to record.

A spend trail

Unbudgeted spend goes up as a request with an amount and a reason. Who approved, who objected and when is all on the record. How a decision got made stops being a search through old chat messages.

Minutes that exist

Record the meeting, get minutes and action points back, turn each one into a task with a named owner and a date. The gap between what a board decided and what anyone actually did closes on one screen.

A chart with evidence under it

Who reports to whom, and each person's delivery record underneath the name. Key person risk stops being a worry somebody raises and becomes something you can point at.

History nobody typed in

Completion and on-time rate per person, computed from the work itself. Numbers that were never assembled for an audience are the ones worth putting in front of one.

  • Read only access for a board member or an investor, scoped to exactly the lists you choose and nothing else
  • Every approval, comment and status change carries a name and a timestamp, kept as long as the instance lives
  • The monthly update stops being a two day job, because the numbers are already there
  • Export the lot whenever you like: it is your Postgres database, your dumps, your volume

When the money asks, the answer is already on the record.

Proof

Built by an agency, on itself, in production since July.

  • 10 July 2026 Went live as the workspace of Nios Media, a Nairobi marketing agency, replacing the tool it ran on before.
  • 26 July 2026 Public on its own domain, over a tunnel with no open ports on the network.
  • 31 July 2026 A full re-import ran against the live instance: 759 tasks refreshed across 22 lists, and the 30 values the team had typed by hand came through untouched.
  • 25 August 2026 A second company went live on its own instance, its own database and its own domain.
  • 1 September 2026 A client walkthrough produced fifteen requests. All fifteen were closed in the release round that followed.
  • Today 12 client accounts, 25 content calendars, a team in it every working day.
  • Engineering 34 release rounds shipped, each one gated on automated checks. 37 test suites live in the repository and run against a running stack before anything reaches a client.
  • Security An internal audit across authorisation, injection and infrastructure, patched and re-verified. API keys stay server-side, containers run as a non-root user.
  • Backups A compressed dump every night, 14 kept on a rolling window, and a restore that has been run for real rather than written down.

Every number on this page is open on the call.

Pricing

Every price, printed.

A one-time setup project, then a flat monthly. No per-seat charges, no surprise tiers on a sales call.

Shilling pricing applies in Kenya and East Africa. Dollar pricing applies elsewhere.

Core

Teams up to 15

KES 150,000USD 1,900setup, one time

KES 30,000USD 200per month

  • Your own domain, database and nightly backups
  • The full workspace and the team pyramid
  • Mobile apps and push notifications
  • 1 client portal
  • Data migration available, KES 60,000
  • AI ready, bring your own key
  • Email support, two business days

Recommended

Standard

Teams up to 40

KES 280,000USD 3,500setup, one time

KES 55,000USD 350per month

  • Everything in Core
  • Data migration included, from your current tool
  • AI features on, key included
  • Unlimited client portals
  • Support next business day
  • Quarterly review call

Scale

Teams of 40 plus

from KES 500,000from USD 6,000setup, one time

KES 95,000USD 700per month

  • Everything in Standard
  • Host it yourself if you prefer
  • Source licence included
  • Same-day support, named contact
  • Monthly review call
  • Custom automations scoped in

Add-ons, printed too: extra training session KES 25,000. Custom automation or integration from KES 80,000. Pay 10 months up front and get 12. All prices exclude 16% VAT.

Price it against the hire, not against software.

The other way to fix this is a person. An operations coordinator whose whole job is asking where things are, chasing the ones that slipped, and writing up who did what. That is a salary every month, statutory contributions, leave, a laptop, a recruitment round, an induction, and a notice period at the end of which the whole picture walks out with them. Standard is KES 55,000 a month, flat, printed above, for a team of up to 40. It starts the day it is installed, it holds the record for every person at once, and hiring your next editor never changes the invoice.

What setup includes

The fee is a project, not a licence key.

  1. Discovery call. Your structure, your pipeline, your clients.
  2. Instance provisioned. Your domain, HTTPS, backups wired.
  3. Workspace built. A space and a calendar per client, your own status pipeline, your portals ready to hand out.
  4. People in. Team invited, roles set, the pyramid arranged to match how your agency actually reports.
  5. Data moved. Imported from your current tool, verified together.
  6. Training. A live session plus the written team manual.
  7. First 14 days. Hands-on support while it beds in.

The questions that decide it

Asked and answered.

Will my team actually use it?

There is nothing for them to learn. The accountability layer asks your team for exactly one thing: reply when they get a message. No timesheets, no status forms, no weekly self-assessment, nothing extra typed in, because the numbers compute from tasks, due dates and completions that already exist. Everyone gets a home screen with their own numbers on it, the same numbers their manager sees, on the same day. Setup includes a live training session and a written team manual.

Is this spying on my staff?

No. It reads what is already true: a task moved, a due date passed, a piece shipped. There are no screenshots, no keystroke logging, no location tracking, no idle timers and no timesheets. Nobody fills in a form about themselves. Everyone sees their own numbers on their own home screen, the same numbers their manager sees. The work it does is stopping your reliable people from carrying the quiet ones invisibly, which is how you lose the reliable ones. Nothing on the system messages your team on its own, either. Every check-in is drafted, read by a person, and sent by a person, under their name.

What happens if you disappear?

The stack is yours structurally, not contractually. It runs on your server, you hold the environment file, the volumes and the nightly database dumps, and restoring from them is a documented procedure that has been run for real. The agreement puts it past argument: if the monthly service stops for 90 days, the source code is released to you under a perpetual internal use licence.

Where does my data live?

In your own Postgres database, in your own Docker volume, on the machine you choose: Nairobi, Europe, your own office. Nightly compressed dumps, 14 kept on a rolling window. No other company is in your database, so there is nothing to share and nothing to leak sideways. AI calls leave your server when you switch them on, and the key stays server-side, never in a browser.

Do we pay per seat?

No. Flat per instance within your tier's team size. Hiring your fifteenth or fortieth person changes nothing on the invoice, and every price is printed on this page.

We already run everything on a project tool.

Then you already know the shape of the problem. A board stores the work and shows you what exists. It has no opinion about who is delivering, it never tells you that a designer has gone quiet, and it never chases anybody, so the chasing is still you, twice a week, from memory. That missing layer is what this is: the colour computed from real delivery, the per-person record nobody typed, one shared list of what is on fire, and a chase that takes one click and stays on the record.

Moving off what we run today sounds painful.

So did the agency this was built for, and it moved in one pass: 1,415 tasks, 1,494 comments, 97 custom fields and the docs. Re-runs gap-fill and never overwrite: on a live instance, 30 values the team had typed by hand came through a full re-import untouched, 30 before and 30 after. Migration is included on Standard, and it is verified with you before anyone switches over.

Will our clients see our internal mess?

No. A client portal is scoped to the lists you put in it, so your client sees their own calendar and their own comments, and no spaces, no other clients, nothing internal. Their view also translates your internal pipeline into plain language, so a status named UPDATE REQUIRED reads as In review on their screen. They watch their own work land, which is most of why they stop emailing you for updates.

It is too expensive.

Count the hours you spent last week asking people where things were, and the meetings that existed to produce a status. Multiply by what an hour of your time is worth. You are already paying for this, in the wrong currency, and the invoice is a week of your attention rather than a line in the accounts. The prices are printed above, flat per instance, and hiring a person never raises them.

Could we not build this ourselves?

You could. The software is the small part. What takes the time is deciding what counts as behind, and making it hold on real data without insulting anyone: half or more of a person’s open work overdue, or more late than on-time deliveries in the last fourteen days, and grey when there is nothing to judge, so no claim is made about someone who has had a quiet fortnight. Every release round is gated on automated checks, with 37 test suites in the repository that run against a running stack before anything reaches a client. What you are buying is the rules and the record, not a login.

Does it work on phones?

It installs as a real app on iPhone and Android, with genuine push notifications, a five-tab mobile layout and a calendar built for small screens. Laptops install the same way from Chrome or Edge.

Is our data used to train AI?

No. AI features switch on per instance, calls go out when someone uses them, and nothing in your workspace trains a model. Leave the key blank and the whole workspace runs without it.

See it live

Twenty minutes. The real system, not slides.

Pick a slot and the live pyramid, a client calendar, the portal your client would log into and the nudge log all open on screen. Bring the person in your team you would most like a straight answer about.

The booking calendar loads here.

Open the calendar

Not ready to talk? Get the pricing sheet and the security notes by email.

Book a demo call →