Thinking from the field.
Practical perspectives on cloud strategy, AI adoption, platform engineering, and digital transformation — written for technology leaders and their teams.
New series
565 articles across 69 series
From Conversation to Cart
Engineering agentic retail that survives production: from a shopper’s request to a valid, available, governable basket, and everything around the model that makes that possible.
15 partsThe AI Productivity Dividend
When AI reduces the effort a role needs, someone decides who receives the gain. Reduced-time roles, outcome-based job design, inclusive talent access and the ways shorter work fails.
6 partsArchitecture Must Earn Its Complexity
Start with boundaries, not microservices. A modular monolith as the default hypothesis, evidence-gated extraction, data ownership, fitness functions and the distributed-systems tax.
9 partsIndividual articles
63 articles
An AI ecosystem map can hallucinate too
Polished diagrams of a vendor's AI ecosystem circulate faster than the documentation they summarise. Some boxes are real products, some are previews, some are the mapmaker's inference and some were never there. This article sets out a method to check a map before it shapes an architecture or a proposal.
Shadow AI is a governance design problem before it becomes a disciplinary problem
Most shadow AI in consulting and transformation work is not a secret chatbot. It is a useful tool used where the approval boundary is unclear: the wrong account, an embedded feature, a synthetic example of a client problem. Two fictional cases, the eight parts of approval, a decision routine with four answers, and a path from shadow use to a governed capability.
A public-safe specification starter kit
Five plain Markdown files that carry one change from intent to evidence, ready to copy: intent, specification, acceptance criteria, a decision log and an evidence record. Plus a filter for making a specification safe to share with a vendor, a coding agent or the open-source community without leaking names, systems or costs.
Before You Build an AI Agent, Design the Handover
Successful AI agent deployment hinges on designing effective handovers between AI systems and human operators, balancing automation with human oversight.
Don't watch the AI type. Watch the structure around it.
Everyone has seen an AI write code. The hard part is the structure around it: who decides, what the agent may touch, and how you prove it built the right thing. A worked example that takes one AI search feature from customer feedback to production, and the five lessons it leaves for South African teams.
Discovery Is Not a Delay; It Is How You Protect the Date
Discovery phases in solution architecture are crucial for managing risks and ensuring project timelines are met, rather than delayed.
Simplified Technical English: an aerospace writing standard for the agent era
ASD-STE100 was written in the 1980s so that aircraft mechanics who read English as a second language would not misread a maintenance procedure. Its rules turn out to fit a very modern problem: instructions that AI agents, multilingual teams and citizens must act on correctly the first time. What the standard is, where it fits, where it does not, and how CloudNala uses it.
From Tender Feed to Pursuit Intelligence
Transforming tender processes from mere feeds to actionable pursuit intelligence can enhance decision-making and competitiveness.
Multi-Partner Delivery: Make Accountability Visible
In multi-partner delivery, clear accountability is crucial for success, requiring visible structures and processes.
The Cost of Duplicate Internal Systems Is Bigger Than Hosting
Duplicate internal systems not only increase hosting costs but also create inefficiencies and complexities that hinder organisational agility and innovation.
From idea to production: how CloudNala builds products for the African market
Building products for the African market requires a balance of innovation, local understanding, and practical execution to ensure success from idea to production.
Is your organisation ready for agentic AI? A practical readiness checklist
Agentic AI promises software that does work rather than merely answers questions, but most organisations are not yet ready to let it act on their behalf. This practical readiness check helps South African leaders find out where they really stand before they commit budget.
The Architecture Is Not the Diagram
A clean architecture diagram can win a steering committee and still leave a system nobody can run. The real architecture lives in the decisions, constraints and trade-offs behind the boxes, and that is what South African organisations should be asking their architects for.
AI Can Read the Tender; It Cannot Own the Bid
AI has made the reading, sorting and first-drafting of tender documents dramatically faster for South African suppliers. What it has not changed is who carries the commitment, the risk and the judgement behind a bid, and confusing the two is becoming an expensive mistake.
Who Supports the Solution After the Workshop?
Workshops, hackathons and innovation sprints are good at producing working prototypes and poor at deciding who will keep them alive. If nobody owns the solution on the Monday after the workshop, the organisation has bought a demo, not a capability.
Agentic AI beyond the demo: 48 concepts that make an agent production-ready
An agent that impresses in a demo is not yet a system you can trust with authority. This field guide sets out 48 concepts across seven layers, from purpose and authority through to operations, that decide whether an AI agent is ready for production.
22 production concepts for AI agents, in plain language
Most introductions to AI agents cover the model, the tools, memory and the loop. The concepts that decide whether an agent can be trusted with real authority are usually missing. Here are 22 of them, each with an everyday comparison and a worked example.
Build on the Platform You Have, but Do Not Inherit Its Assumptions
Building on a platform you already own is usually the smartest and cheapest place to start. The risk is not the platform itself, but the unexamined assumptions about users, connectivity, data and cost that come bundled with it.
AI models available in South Africa: what "available" really means
A model can be available to a South African customer without being hosted in South Africa. An endpoint can sit in Johannesburg or Cape Town while the inference runs on another continent. Verified against first-party documentation on 9 September 2026.
Practical cloud cost optimisation: 10 changes you can make this week on Azure, AWS or GCP
Most cloud cost advice asks you to redesign your architecture before you can save anything. These ten changes are different — they are configuration, not transformation, and a competent engineer can land most of them in a week.
AI governance: the framework your organisation needs before scaling AI
Establishing a robust AI governance framework is crucial for organisations before scaling AI solutions to ensure ethical, compliant, and effective deployment.
The case for multicloud: when to use Azure, AWS and GCP together — and when not to
Using Azure, AWS, and GCP together can offer flexibility but also complexity; understanding when multicloud aligns with your strategic goals is key.
Responsible AI Accuracy: Stop Promising a Single Percentage
AI accuracy metrics need context, not oversimplified promises, to guide responsible strategy and decision-making.
A Go/No-Go Gate Is a Strategy Tool, Not Sales Administration
A Go/No-Go gate is a strategic decision-making tool that aligns tender opportunities with business goals, not just a sales checkpoint.
Why digital transformation fails without an operating model change
Without adapting the operating model, digital transformation efforts risk failure as misaligned structures and processes hinder technological integration.
Replit in 2026: what it is, who it's for, and where it fits in your development workflow
Replit in 2026 offers a unique, collaborative development environment that integrates seamlessly into modern workflows, catering to diverse teams and evolving needs.
Building for Africa: why cloud architecture here needs a different approach
African cloud architecture must address unique challenges like connectivity, regulation, and diverse business needs to succeed.
How to build a data strategy when your data is a mess
Building a data strategy amidst chaos requires prioritizing business goals, establishing governance, and leveraging technology to unlock data value.
Seven solution architecture mistakes that cost enterprises millions
The most expensive architecture failures rarely show up in a diagram or a demo. They surface eighteen months later, in the run costs, the rework, and the integration nobody scoped.
Small Teams Need Better Decision Systems, Not More Tools
For small teams, improving decision systems is more critical than acquiring more tools, enabling strategic choices and efficient operations.
Public Digital Services Need a Service Model, Not Only an App
Public digital services require a comprehensive service model rather than just an app to effectively meet citizen needs and government objectives.
The B-BBEE advantage in tech procurement: what suppliers need to know
Understanding B-BBEE in tech procurement can unlock new opportunities for suppliers by aligning with South African socio-economic goals.
The Hidden Architecture of a Good Proposal
A well-crafted proposal goes beyond meeting requirements; it reflects an understanding of the client's business and positions your solution as the clear choice.
Agentic AI Needs an Operating Model, Not Just an Agent
Agentic AI requires a robust operating model to truly benefit organisations, focusing on integration, accountability, and adaptability rather than just deploying AI agents.
Zero trust security: a practical starting point for South African organisations
Implementing zero trust security in South Africa involves understanding organisational tensions, strategic planning, and practical trade-offs to protect digital assets effectively.
Containerisation for the South African developer: when Kubernetes is overkill
Containerisation offers flexibility and scalability, but overengineering with Kubernetes can be costly and complex. Assess your needs to choose the right container strategy.
Cloud migration in regulated industries: what banks and insurers need to know
Navigating cloud migration in regulated sectors like banking and insurance involves balancing innovation with compliance, risk management, and operational resilience.
The architecture decisions startups always regret
Startups often make architecture decisions that seem cost-effective initially but lead to regret as they scale. This article explores common pitfalls and offers a framework for better decision-making.
Responsible AI in South Africa: beyond the hype, into practice
Explore how organisations in South Africa can implement responsible AI practices that balance innovation with ethical considerations.
CI/CD platform comparison: GitHub Actions, GitLab CI and Azure Pipelines for South African teams
The CI/CD platform you pick shapes your delivery speed, your security posture and your monthly bill for years. Here is how GitHub Actions, GitLab CI and Azure Pipelines actually compare for teams building and deploying from South Africa.
The SMME Guide to Cloud: Enterprise-Grade Infrastructure on a Startup Budget
Cloud is often sold to SMMEs as an enterprise product with an enterprise price tag. It isn't. Here's how growing South African businesses can get genuinely enterprise-grade infrastructure without an enterprise budget or a full-time cloud team.
Cloud Cost Audits: The 10 Things We Check Before Recommending Any Migration
Cloud waste is rarely one bad server. It's weak governance, unclear ownership, poor tagging and the absence of cost accountability, compounding quietly every month. Here's the 10-point checklist we run before any migration or optimisation recommendation.
How We Built TenderCity: Lessons From Building a South African Tender Intelligence Platform
Government tenders in South Africa are public, but winning them isn't. Here's what we learned building TenderCity — and how those lessons show up in every client engagement since.
Tender Response Automation: How Suppliers Can Stop Rebuilding the Same Documents Every Week
Most suppliers rebuild the same company profile, CVs and compliance documents for every RFQ and tender. A reusable capability pack turns that repeated effort into a five-minute assembly job.
SMMEs Don't Need More Tools — They Need a Digital Operating Model
Most growing South African businesses are already digital — WhatsApp, Excel, banking apps, social media. The problem isn't a lack of tools. It's that none of them talk to each other.
The 90-Day Government MVP: How to Build a Citizen Service Without Boiling the Ocean
Large public-sector digital programmes stall. A focused 90-day MVP — scoped to one service journey — can prove the model, test adoption and create the momentum to scale without the risk of a full transformation.
Public Sector Digital Services Are Not Websites
The biggest reason public-sector digital projects fail is not budget or technology. It is the assumption that building a website is the same as building a service. It is not.
South Africa's AI Policy Reset: What Businesses Should Do Before 2027
South Africa's AI policy environment is still taking shape. Organisations should not wait for final regulation to prepare their AI governance. The risks are already real, and the window for getting ahead of them is closing.
Data Foundations Before AI: Why Your AI Strategy Will Fail Without Clean Data
AI does not fail because the model is weak. It fails because the organisation does not have usable data, clear ownership, trusted processes or business accountability. Here is what you need to fix before AI delivers any value.
Public sector digital services: why they fail and what changes it
South African government digital projects have a poor track record — not because the problems are unique, but because the approaches are predictable. Here is what actually causes these failures and what the organisations that get it right do differently.
Infrastructure as code: why you need it and how to start small
Most South African organisations manage cloud infrastructure the wrong way — through the console, through memory, through one person who knows where everything is. Infrastructure as code fixes this. Here is what it is and how to start without a full platform team.
Building an MVP that won't embarrass you at Series A
The word 'MVP' has become permission to build things badly and call it strategy. Here is what technical decisions made in the first three months actually look like eighteen months later — and which ones will define how far your startup can go.
AI use cases that actually work in South African businesses right now
There is a significant gap between AI use cases that look compelling in vendor presentations and those that deliver measurable value in the South African business context. Here are the ones that actually work — and the conditions that make them work.
DevSecOps isn't a tool — it's a culture change. Here's how to start.
Most organisations that say they are doing DevSecOps are doing DevOps with a security scanner bolted on. That is not the same thing — and the difference shows up at exactly the wrong moment. Here is what the culture change actually looks like.
GitHub Copilot, Amazon Q and Kiro: which AI coding tool is right for your team?
Three serious AI coding tools, three different bets on how software development should change. Here is a practical comparison for South African development teams deciding which to adopt — and which to leave for later.
Cloud cost governance: the framework every South African organisation needs
Most South African organisations know they are overspending on cloud. Fewer have a governance structure that actually prevents it. Here is the framework that does — practical, provider-agnostic, and built for how SA organisations actually operate.
Amazon Kiro: what AWS's AI-native IDE means for development teams in 2026
Kiro is not another Copilot plugin. It is AWS's attempt to redesign how software gets built from the ground up — spec-first, agent-driven, deeply integrated with AWS. Here is what it actually is, what it changes, and whether your team should care.
Five signs your digital transformation programme is already off track
Most digital transformation programmes do not fail spectacularly. They drift — slowly, quietly, until the budget is gone and the outcomes are not. Here are five signs to look for before it is too late to correct course.
Azure in South Africa: what's available, what's not, and how to plan around it
Azure has two regions in South Africa, but not every service runs there. Understanding the gap between what Microsoft offers globally and what is available locally is essential before you design anything.
The case for platform engineering: why your dev team is slower than it should be
If your developers are spending more time managing infrastructure, configuring pipelines, and chasing access than writing code, you have a platform problem. Here is what platform engineering actually is and why it is worth solving.
What POPIA means for your data and AI strategy
POPIA has been enforceable for four years. Most organisations have compliance documents. Fewer have genuinely thought through what it means for how they build AI systems. Here is what actually matters.
Agentic AI: what it actually is, and why it matters for your business
Everyone is talking about agentic AI. Most explanations make it sound either trivial or terrifying. Here is a plain-language breakdown of what it actually is, what it can do today, and how South African businesses should be thinking about it.
Why South African businesses are still overpaying for cloud — and what to do about it
Most South African organisations moved to cloud with good intentions. Many are now paying significantly more than they should. The reasons are consistent — and fixable.
Coming up
- Google Cloud in South Africa: what's available, what to use it for, and how it compares
- Technical Debt Is Often Decision Debt
- Lean Architecture Does Not Mean Incomplete Architecture
- Your AI Proof of Concept Is Not a Production Plan
- AWS in South Africa: the Cape Town region, what it offers, and how to plan your architecture
- When the Existing Ecosystem Arrives Late
- Data Readiness Is a Business Conversation
- From Replit to production: migrating your app to AWS, Azure or GCP