If you are a business leader in 2026, AI is impossible to ignore and increasingly difficult to navigate. Your LinkedIn feed is full of success stories. Every conference promises transformation. Consultants arrive with solutions to problems you did not know you had. And somewhere in the middle of all that noise, you are trying to work out what is actually worth your attention and what is just expensive theatre.
I have been working in data and AI for more than fifteen years, and on AI specifically for close to three years. I have built business intelligence platforms for an airline, led CDP deployments across ten markets of a luxury automotive brand, ran segmentation models on over 20 million customers for a QSR client, and spent the last few years building AI applications from scratch. I have seen more technology waves than I can count, and I have learned, sometimes the hard way, which of those create real commercial value and which just create impressive slide decks.
What I am sharing on this blog is not academic. It is not a fellowship transcript or a listicle of trending tools. It is over fifteen years of practitioner experience applied to the question that actually matters: where does AI create genuine business value, and how do you get there?
The marketing oversells AI, and the sceptics dismiss it too easily. The real value sits in between, and reaching it takes knowing how to navigate the noise.
From BI Developer To AI Builder
I started my career writing code. Then, like many people who moved into leadership, I gradually stepped away from the technical work and into strategy, delivery, and team management. By the time I was running the Business Intelligence Competency Centre at a reputed Airline in Dubai, a fifteen-person team that delivered a programme generating over ten million dollars in cost savings, I was deep in the business of translating data into decisions, not building the pipelines myself.
Fast forward to 2024, and AI changed that equation. Not because leadership roles became less important, but because the tools became accessible enough that building again was possible and necessary. If you are going to advise clients and organisations on AI adoption, you need to have built things. You need to have hit the walls, made the tradeoffs, and understood why some elegant-looking architectures fall apart in production.
About eighteen months ago, I found myself doing something I had not done in over a decade: writing code. The difference this time is that the building blocks include large language models, APIs, intelligent agents, and retrieval systems. The engineering instincts from my early career turned out to be surprisingly transferable; it is the same discipline of breaking a problem into its components, understanding the data flow, and designing for edge cases. It just looks very different now.
The fellowship I did in 2025 gave structure to what was becoming a fast-moving and sometimes overwhelming space. Every week brought new tools, new frameworks, new announcements. The programme did not teach me to use AI; I was already building with it. What it did was give me a map: foundational concepts, architectural patterns, and crucially, the ability to evaluate what is genuinely useful versus what is noise. That distinction is worth more than any specific tool.
The moment that genuinely stopped me was seeing how easily an AI system could be connected to the tools a business already runs: calendars, CRMs, email, databases, not by rebuilding those systems, but simply by telling the AI they exist. What I had assumed would require months of custom integration work turned out to be a configuration exercise. That single realisation changed how I think about the cost and timeline of AI adoption for any organisation that has been hesitant to start.
Three AI Systems I Have Built And What They Taught Me
Theory is necessary. Working solutions are better. Here are three AI systems I have built that are solving real problems. Each one taught me something that changed how I think about AI implementation.
1. Avenote AI-Powered Content Studio |
|
|---|---|
| What it does | Avenote transcribes voice recordings, text, and documents, then transforms them into multiple formats: summaries, translations, articles, mind maps, blog posts, social media content, and more. It handles the full content repurposing workflow that most creators and businesses do manually, inconsistently, and at significant time cost. |
| The problem | Content creators and marketing teams waste hours repurposing a single piece of content across platforms. A recording that should become a blog post, a LinkedIn article, a newsletter, and social snippets ends up staying as a recording because the manual work is simply too time-consuming. |
| Key lesson | AI solutions succeed or fail based on workflow design, not technical sophistication. |
The use cases that have emerged gladdened me. A product manager told me she uses it to take the chaos of raw ideas and research and turn it into presentation-ready material; the Mind Map feature in particular has become a core part of her workflow for breaking down complex problems before she brings them to her team.
A senior voice engineer pointed to something different: the ability to capture content from any webpage instantly via the bookmarklet, then combine that with audio transcription to turn conversations (interviews, briefings, discussions) into structured, actionable summaries. Two very different roles, two very different workflows, both built around the same underlying capability: turning unstructured input into something you can immediately use.
You can access the app on https://avenote.co.
2. AI Solution Advisor Matching Businesses to the Right AI |
|
|---|---|
| What it does | Instead of businesses wading through hundreds of AI tools with no guidance, the Solution Advisor asks structured questions about their specific needs, then matches them with relevant tools from a knowledge base of over 200 AI solutions. It generates detailed implementation proposals and provides an interactive Q&A for follow-up questions. |
| The problem | Decision paralysis. Small and medium businesses know they should be using AI but have no idea where to start. They do not need another ranking of the top fifty tools. They need specific, contextual recommendations for their situation, their budget, and their team's capability. |
| Key lesson | Generic AI advice is worthless. Context-specific guidance is the entire value proposition. |
I haven’t publicly launched the AI Solution Advisor, since it uses my personal API tokens but can be demoed separately. If you would like to run an assessment for your business, please reach out to me and I will run an assessment for free. It takes about three minutes and produces a recommendation specific to your situation.
3. Customer Support Multi-Agent RAG-Powered Response System |
|
|---|---|
| What it does | This system automatically responds to customer queries from email and Google Maps reviews by drawing from your business's knowledge base: product documentation, FAQs, past interactions, and policies. The multi-agent architecture routes queries to the right specialist agent and escalates to a human when necessary. |
| The problem | Small businesses cannot afford round-the-clock support teams, but customers expect fast, accurate responses. Manual handling of repetitive questions drains time and creates inconsistency. Every slow or unhelpful response is a trust cost. |
| Key lesson | The quality of your knowledge base determines the quality of your AI responses. Before you build the agent, build the knowledge base. |
What Makes 2026 Different From Previous AI Waves
We have been here before. Watson was going to transform healthcare. Chatbots were going to replace call centres in 2016. Robotic process automation was going to eliminate back-office work by 2019. Each wave brought genuine capability, legitimate hype, and eventually a more measured reality. What makes this moment different?
2011 IBM Watson |
2016 Chatbots |
2019 RPA |
2026 This Wave |
|||
|---|---|---|---|---|---|---|
✗ Overblown Healthcare AI will transform diagnosis |
→ | ✗ Overblown Bots will replace call centres completely |
→ | ✗ Overblown Automation will eliminate all back-office work |
→ | ✓ Different Accessible, agentic, and data-powered |
◉ Capable & accessible models Frontier AI is now accessible via APIs at pennies per interaction. No data science team required. |
◈ Mature tooling ecosystem No-code platforms, orchestration frameworks, and MCP mean a weekend prototype is now realistic. |
|
|---|---|---|
◆ 'Should we?' not 'Can we?' The question has matured from capability to governance, ROI, and strategic fit. That is the real shift. |
◇ Data as the differentiator Everyone accesses the same models. The winner is whoever has richer, cleaner, better-governed data. |
|
1. The Models Are Genuinely Capable, And Now Within Reach
Today's frontier language models can understand context, reason across long documents, write code, analyse data, and produce work-quality outputs across a remarkable range of tasks. More importantly, they are accessible through APIs that cost pennies per interaction. You do not need a data science team or a million dollar infrastructure investment to deploy AI in a business context. The barrier has dropped from enterprise-only to any-organisation.
2. The Tooling Ecosystem Has Matured
Three years ago, building AI meant writing custom neural networks. Today there are no-code platforms that let non-developers build AI workflows, standardised orchestration frameworks that handle the infrastructure, and protocols like MCP (Model Context Protocol) that let AI systems communicate with your existing business tools. What previously required a team of engineers can now be prototyped in a weekend.
3. The Conversation Has Shifted From 'Can We?' To 'Should We?'
The most important signal that this wave is different is the maturity of the questions being asked. Businesses are no longer asking whether AI can perform a task; they are asking where it adds the most value, what the governance implications are, and how to measure ROI. That kind of question only gets asked once a technology has stopped being an experiment and started becoming infrastructure.
4. Data Strategy Has Become The Differentiator
Here is the one that most AI-focused commentators understate: the businesses that will extract the most from AI are not those with access to the best models since everyone has access to the same models. The differentiator is data. Rich, clean, well-governed first-party data. Organisations that have invested in their data foundations: their CDPs, their measurement frameworks, their governance, are positioned to build AI capabilities that competitors without that foundation simply cannot replicate. This is why data is the key component in any AI discussion.
A Deloitte survey published in early 2026 captures this shift precisely for the GCC region. In 2024, 52 percent of organisations surveyed had not adopted generative AI at all. By 2025, that figure had dropped to 29 percent in a single year. More telling than the adoption rate is what organisations are doing with it: early use cases were dominated by basic tasks like email drafting. By 2025, research and analysis had become the primary use case at 41 percent, with accuracy and quality improvement at 38 percent. Businesses have stopped experimenting with AI for its own sake and started using it to do specific work.
GCC Non-Adoption Rate 52% → 29% 2024 2025 Non-adoption fell 23 points in a single year |
How Use Cases Have Shifted THEN Email drafting & basic tasks NOW Research & analysis 41% NOW Quality & accuracy improvement 38% Deloitte Middle East Survey, Feb 2026 |
|---|
Source: Deloitte Middle East, GenAI Adoption in the GCC, February 2026
https://deloitte.com/middle-east/en/services/tax/research/genai-adoption-in-the-gcc-2026.html
None of this makes adoption automatic. The tools are more capable and more accessible than they have ever been, but capability was never the real constraint. Judgment is, knowing which problem is worth solving, which shortcut will cost you later, and which pitch is dressed-up nothing. That is the lens I bring to everything I write here, built from shipping these systems rather than watching from the sidelines.
If you are trying to work out where AI actually fits in your business, get in touch. I would rather talk through your specific situation than hand you another generic framework.