AI consulting and AI products, built in Kuwait.
I've built and run digital businesses in Kuwait for fifteen years. Now I build AI products, agents and automation for companies here — starting from the workflow that actually costs you money.
Kuwait is not a generic AI market
The businesses that ask me about AI in Kuwait tend to have three things in common. They run on a mix of systems bought at different times and never fully connected. A meaningful share of their customer communication happens on WhatsApp rather than email. And their content, contracts and customer records are bilingual, with the Arabic side usually the messier one.
That combination decides what actually works. A polished English-only chatbot will fail on the first Kuwaiti dialect message. An automation built on a clean CRM assumption will fail because the real record lives in someone's inbox. The interesting work is rarely the model — it's getting to the point where a model has something reliable to act on.
It also means the honest first answer is sometimes no. A company with three hundred support conversations a month does not need an agent; it needs two saved replies and a form. I would rather tell you that in week one than bill you for six months of building it.
AI product development
Building the product itself: customer-facing apps, internal tools, and AI-powered features inside systems you already run. In practice that spans a booking or quoting flow that drafts its own responses, an internal tool that replaces four spreadsheets and a WhatsApp group, or a product feature that classifies and routes incoming work.
The technical decisions that matter early are boring ones. Where does the data live, who is allowed to see what, what happens when the model is wrong, and can a non-engineer correct it without a deploy. Getting those right is what separates a feature that survives a year from one that gets switched off in month two.
AI agents
An agent is worth building when it can complete meaningful work, not when it can hold a conversation. The ones that hold up in Kuwaiti businesses:
Support agents that answer from your own documentation in Arabic and English, and hand off cleanly when they are unsure.
Research agents that gather and summarise sources on a recurring question — competitors, tenders, market movements — on a schedule.
Content agents that draft in a house voice and hand a human the last 20%, rather than publishing unattended.
Operations agents that watch a queue, apply rules, and escalate the exceptions a person actually needs to see.
Internal knowledge agents that answer staff questions from policies, contracts and past projects, with citations back to the source document.
Lead qualification agents that read an inbound message, enrich it, score it and route it before anyone opens it.
The test I apply before building any of these: if it gets the answer wrong at 2am, what breaks, and who finds out? If nobody can answer that, the agent isn't ready to be autonomous yet.
AI automation
Removing repetitive work from the middle of a process, usually across systems that were never designed to talk to each other.
CRM workflows — enrichment, deduplication, follow-up drafting and stage hygiene that nobody has time to maintain manually.
WhatsApp workflows — routing, triage, structured capture and handover, which in this market often matters more than email automation.
Content operations — briefing, drafting, bilingual variants, and the approval trail around them.
Customer support — classification, first-draft replies, and surfacing the three tickets that need a human today.
Reporting — pulling from the systems the numbers actually live in, instead of a monthly copy-paste ritual.
eCommerce — product data, descriptions in both languages, returns triage and post-purchase messaging.
Internal processes — approvals, onboarding, document handling and the small administrative work that quietly consumes a team.
Arabic AI, done properly
Most bilingual products in this region are English products with an Arabic layer applied at the end. You can see it immediately: mirrored layouts that break, numerals and dates in the wrong system, a font chosen for Latin text carrying Arabic badly, and search that fails the moment someone types without diacritics.
For AI products the stakes are higher, because the model inherits every one of those decisions. Prompting in translated English produces stilted Arabic. Dialect input — which is how people actually write in Kuwait — degrades models tuned on Modern Standard Arabic. Retrieval over a bilingual document set returns nothing useful unless the indexing accounts for both.
I design the Arabic experience as the primary one where the audience is Arabic-speaking, and treat English as the parallel build. It takes longer at the start and saves a rebuild later.
AI experience in Kuwait
I have been CTO of Al-Qabas, one of Kuwait's established newspapers, since 2016 — leading digital transformation across the organisation, overseeing technology infrastructure and platform development, and working on how content is delivered to and engaged with by readers. That is a decade of running a real platform in this market, through the shift from print-era publishing to a digital product business.
It is also where I learned the thing that most informs the AI work: at a publisher, the systems that decide what a reader sees are judged every single day by whether readers come back. There is no room for a feature that demos well. Either engagement moves or it doesn't.
Alongside that I am co-founder and CTO of Wavai, a Kuwait-based web and eCommerce practice; CEO of Menasa since 2020; and I hold operating seats at Bleep and SHASHA. The through-line is fifteen years of shipping into this market rather than advising about it.
Who is actually doing the work
Selected rather than exhaustive — the roles and evidence relevant to an AI engagement here.
Faisal Karkoh — AI Product Builder, Kuwait
Based — Kuwait City. Fifteen years building and operating digital businesses across Kuwait and the UAE.
Al-Qabas — Chief Technology Officer since 2016. Quoted in the partner announcement for the AI recommendation and search platform — the first of its kind for a media outlet in the region.
Wavai — Co-Founder & CTO since 2010. The Kuwait web and eCommerce practice behind much of the delivery work.
Menasa · Bleep · SHASHA — Chief Executive, Chief Operating and Chief Technology Officer respectively. Operator, not adviser.
Shipped AI — AhlanAI, an Arabic-first directory of 500+ AI tools. AlooChat, a conversational automation platform. The Al-Qabas archive and recommendation work.
Published — Nine templates accepted into Notion's own gallery, and 55 articles on AI tooling, agents and implementation.
Languages — English and Arabic, with Arabic built in from the first version rather than translated afterwards.
Every claim here is stated on this site with a source behind it — see /about and /work.
What an AI engagement actually looks like
Most of what gets called AI consulting stops at a recommendation. These end with something running.
A conversation about the business, not the technology
What is slow, what is manual, what is being lost. AI is not the starting point — the cost of the current process is. Some of these conversations end with me saying the problem is not an AI problem.
One process, traced end to end
Who touches it, where it waits, how often it runs, and what it costs as it stands. This is where a vague brief becomes two or three specific things worth building.
A written recommendation with the trade-offs in it
What to build, what to leave, what it will cost to run, and where it will break. If the honest answer is that this should wait, that is what it says.
The smallest useful version, built
Not a pilot that lives in a slide deck. Something that runs against real data, that the team can use, and that can be measured against the process it replaces.
Run alongside, then handed over
It runs next to the existing process until it is trusted, then the team owns it — with documentation, and someone named as responsible for it.
Selected Kuwait-related work
Digital transformation of an established Kuwaiti newspaper — infrastructure, platform development and the content delivery experience.
Kuwait-based web and eCommerce practice building Shopify stores, WordPress sites and Webflow experiences for GCC brands.
Arabic-first AI tools directory — 500+ tools across 20+ categories, built with the Arabic experience as the primary one.
Alsadu
A digital platform for Sadu weaving and Kuwaiti craft heritage, currently in active development.
An AI platform for automating conversations and customer support — the agent and automation work this page describes, built as a product.
Where this tends to land
Sectors where I've either operated directly or delivered work in Kuwait and the wider GCC:
Media and publishing — content operations, personalisation, archive search.
eCommerce and retail — product data, bilingual catalogues, support and post-purchase.
SaaS and digital products — AI features inside an existing product.
Startups — MVPs where the point is to learn quickly whether the idea holds.
Professional services — proposal drafting, document handling, internal knowledge.
Operations-heavy businesses — approvals, scheduling, reporting and internal tooling.
Customer support teams — triage, drafting and escalation across Arabic and English.
Common questions
What does an AI consultant in Kuwait actually do?
In practice, two things: work out which parts of a business genuinely benefit from AI, and then build those parts. The first half is mostly subtraction — most of what a brief asks for is not worth automating yet. The second half is why I do not stop at advice: a recommendation nobody implements has cost you money and changed nothing.
Do you advise, or do you build?
Both, and the building is the point. Fifteen years of running businesses here means the advice comes from having operated the things I am recommending changes to — but the engagement ends with something shipped, not a document.
How do you handle Arabic and English together?
Bilingual from the start rather than translated afterwards. For AI products that matters more than it does for a website: an assistant that handles Gulf Arabic dialect, mixed Arabic-English input and right-to-left interfaces has to be built and evaluated that way from the first version.
What types of AI products can you build?
Customer-facing apps with an AI layer, internal tools and knowledge assistants, AI features added to a product you already run, agents that own a defined workflow, and automation across the systems you use. If it needs a data model, a UI and a deployment, it's in scope.
Do you work with businesses in Kuwait?
Yes — Kuwait is home base and where most of my operating history is. I've been CTO of Al-Qabas here since 2016 and co-founded Wavai, a Kuwait-based web and eCommerce practice.
Can you build Arabic AI products?
Yes, and Arabic-first rather than Arabic-translated. That covers RTL layout, Arabic typography, dialect handling, bilingual retrieval and search, and prompting written in Arabic rather than translated from English.
Do you build AI agents?
Yes, where the workflow justifies one. An agent is worth building when it can complete real work with defined permissions, tool access and a human approval path. If the same result comes from a form and two saved replies, I'll say so.
Can you integrate AI with existing systems?
That's usually the majority of the work — CRM, WhatsApp, eCommerce platforms, internal databases, document stores and reporting. Integration is planned up front rather than discovered halfway through, because that's where timelines slip.
How do engagements usually start?
A thirty-minute call to find the workflow with the highest potential impact, then a short scoping pass. If there's a build worth doing, the first milestone is something running that a real user can judge — not a document.
Related
AI hub — Everything in one place
AI Product Development — How a build actually runs
AI Agents — Production-readiness checklist
AI Automation — Workflows worth removing
Arabic-first AI Products — The differentiator, in detail
About Faisal — Background and operating history
Selected work — The projects these claims point back to
Web and digital products — When the build is a platform rather than a model
AI Consulting — The advisory side on its own
Insights
Insights
Plan Big Refactors Safely with Claude Code — How to use Claude Code plan mode and the `opusplan` stack to run large, cross-cutting repository changes through a spec-first, branch-safe workflow tied to PRs and CI.
Cursor + Codex as an editor–agent workflow — How to pair Cursor as your AI IDE with Codex as a scripted agent for safe, scoped repo-wide changes—with real costs and a concrete end-to-end pattern.
Cursor indexing that works on large codebases — A concrete indexing and context blueprint for making Cursor reliable on large monorepos: what to index, what to ignore, and how to repair stale context.
Letting Codex Into Prod Without Losing Sleep — How to let Codex read and open PRs against your production repo without ever letting it deploy, run live migrations, or see real secrets.
Safe AI Coding Agent PR Workflow for Production — A production-safe, tool-neutral pull request workflow for AI coding agents: isolate branches, tighten CI, keep merges human, and track SLOs and costs.
Design a Safe Codex Workflow on GitHub — A concrete, production-safe way to let Codex work on branches and PRs in GitHub Actions without ever letting an agent merge into main or bypass human review.
Want to automate part of your business?
Want to automate part of your business?
Let's find the workflow with the highest potential impact first, then decide whether it deserves a build.
Browse the site
Home
about
story
work
expertise
ai
ai ai product development
ai ai agents
ai ai automation
ai ai consulting
ai arabic ai products
ai kuwait
toolkit web
toolkit claude
toolkit lovable
toolkit notion
toolkit webflow
toolkit shopify
toolkit wordpress
toolkit ai solutions
services
services business strategy
services growth planning
tools
blog
listening
books
stack
contact
quote
privacy
terms