Sakhile Bhayi
Sakhile Bhayi, signature logo

Approach

How I think about business information, how I approach a change, and the four kinds of work I offer.

What I believe about business information

Most growing businesses already have plenty of software. Their information sits in pieces across it, and people join the pieces by hand. The first fix is to connect what is already there.

  • Take facts from the source.

    A number should come from where the work was first recorded: the till, the job card, the scanner, the machine. Every retyping is a chance to get it wrong.

  • Show how old every number is.

    If you can't tell whether a figure is from this morning or last Tuesday, you can't act on it. My operational systems show how fresh each reading is and label anything simulated.

  • Never fill a gap with a guess.

    Missing data is shown as missing. In Dot.Central, a machine is never marked as stopped just because its data is missing.

  • Keep what people add apart from what systems recorded.

    A person can explain a figure. They should not be able to quietly change it.

  • Turn recurring errors into rules.

    If the same mistake turns up every week, the system should stop it where the data is captured.

  • Govern AI before you make it clever.

    AI can suggest. A person approves, and the system records what was decided. A system only gets to act without approval once its recorded results justify it.

How I approach a change

  1. Find where the information actually breaks.

    Which figures are captured more than once, which reports disagree, and which ones people have quietly stopped trusting.

  2. Agree a baseline before anything changes.

    For example, the hours it takes to produce the weekly report, or the gap between two systems' versions of the same figure.

  3. Turn the recurring errors into rules the system enforces.

  4. Build the connections.

    I use parts of the Dot.Eco-system where they fit and your existing tools where they don't, so you see one picture with the source and age of every number.

  5. Prove the new system reproduces the trusted numbers before anyone switches over.

    Then measure the change against the baseline.

When something fails, I look for the cause behind the symptom and turn it into a permanent guard.

What I offer

Four kinds of work. Each is scoped to what I can deliver and support myself.

  1. Information-flow diagnostic

    For
    Owners and operations or finance heads of growing businesses whose reports are stitched together by hand.
    You get
    A written map of where your figures are captured twice, disagree or go stale; the recurring errors, written as rules a system could enforce; a baseline we both agree; and a plan for which connections to build first.

    Start here. It sets the baseline everything after it is measured against.

  2. Connected reporting and visibility

    For
    Businesses that have had the diagnostic, or already know which figures they can't trust.
    You get
    Your existing systems and capture joined into one view that shows the source and age of every number, with recurring reports written from shared templates. Measured against the baseline.

  3. Practical AI, with a person deciding

    For
    Organisations that want AI to save real time without handing it the keys.
    You get
    One or two workflows where AI earns its place, and the controls around it: what it may recommend, who approves, what gets recorded, and what it may never do alone.

  4. Product and technology strategy for founders

    For
    Founders who know what their product should do, but not how to scope it or build it with AI and still trust the result.
    You get
    Product scoping, architecture choices, and a way of building with AI-assisted tooling under automated tests and controlled deploys. It is how I built and run the Dot.Eco-system.

Write to me about the diagnostic

What's already built

These run in the Dot.Eco-system today, and the status column says how far each has got.

WhatWhere it runsStatus
Daily production worked out from a machine's own countersDot.MinesBuilt and tested
One view of operations with the age of every readingDot.Mines, Dot.CentralDeployed
A control room that adds judgement without editing source factsDot.CentralDeployed
Recurring reports on shared templates, with an audit log that cannot be editedDot.DocLive
Signed data hand-offs between connected platformsInfoDot and the platforms it connectsDeployed
Monitoring that recommends fixes for a person to approve, with a memory of past incidentsDot.Brain, Dot.MemoryRunning
AI agents built to wait for approval and to be suspended when they driftDot.AgentsDeployed
Insights that show the source behind themDot.AnalyticsDeployed

What I don't take on

I'm a full stack developer, but I don't take general development billed by the hour. I take on work that starts from a business problem.

How I build the Dot.Eco-system

I architect the systems and build them with an AI-assisted workflow. I set the product decisions, the rules and the order in which work is merged, review what ships and run it in production. That is how one founder built, and now runs, the Dot.Eco-system.

AI-assisted work is fast, so the controls matter more. Most Dot.Eco-system platforms ship through automated tests and deploy only after they pass. Dot.Mines alone runs more than a thousand automated tests on each change. When something fails in production, the cause becomes a guard so it can't happen the same way again.

For a client, this means the person who designs the system is the person who builds and runs it. For a founder, it is a way of working I can help you set up.

Why work with me

I've worked where the numbers are first recorded: in mining operations, in site technology and mine control systems. Much of the connecting layer already exists: signed hand-offs between platforms, freshness on every operational reading, recurring reports and monitoring are built and running in my own platforms. As a mine control systems engineer and full stack developer, I've worked on both sides of the problem: where the numbers are first recorded, and the software that joins them. You deal with one person end to end: I design the system, build it under automated tests, and run it.

We'd measure the work against a baseline we agree at the start.

Write to me

Write to me

If your reports take hours to assemble and still disagree, write to me.

I'm also glad to hear from founders building a product, from organisations putting AI to work, and from investors and partners.

You can also email info@sakhilebhayi.com.

A few lines is enough: what your business does, and where the numbers don't add up.

I use what you send only to reply to you. It reaches my inbox through Web3Forms, the form service this site uses, and is never added to a mailing list.