I've been interviewing COOs and Heads of Digital across the GCC consumer brand mid-market for the Plaith Operational Benchmark, and the same diagnosis keeps surfacing.
The AI work happens. Usually fast. Usually well.
An agency in DIFC runs a six-week sprint and ships a Ramadan campaign generation workflow in Arabic and English across Snapchat, TikTok, and Instagram. A freelancer builds an n8n flow that handles the Saudi regulatory approval cycle for new SKU imagery. An in-house marketing operations team prototypes a multi-brand asset localisation pipeline that handles MSA, Gulf dialects, and English voice-over alternates.
Six months later, none of it is running anymore.
Three places, three failure modes
The agency moved to the next client and the playbook went with them. The freelancer left and the prompts they wrote sit in a Slack archive nobody opens. The in-house team got redirected to whatever this quarter's CEO priority is, and the prototype is bookmarked but not maintained.
The institutional knowledge that should have compounded inside the business has dispersed across LinkedIn profiles, vendor handover docs, and laptops that aren't in the building anymore.
This is not a productivity problem. It is a structural one.
Most consumer brand businesses in the region have nowhere for AI work to live. There is no equivalent of the systems team or the platform layer you'd find in a tech company. The closest thing is usually a marketing operations function reporting into a CMO, and that function is too busy running campaigns to also be running the AI infrastructure that runs the campaigns.
Add to that the regional reality that 94% of UAE businesses are SMEs, and most don't have headroom to build a dedicated AI engineering team, even if they wanted to.
The question is not whether to do more AI work
The question I keep coming back to in these conversations is not should we be doing more AI work. It's where is the AI work going to live the next time we do it.
If the answer is still with whichever agency or freelancer happens to be in the building this quarter, the work won't compound. It just disperses again.
If the answer is we'll build an internal AI team, the headcount math is brutal — senior AI engineers in the GCC clear AED 50–80k a month, mostly absorbed by G42 and the sovereign-scale buyers, and a mid-market consumer brand can't compete for them.
If the answer is a dedicated piece of infrastructure that stays inside the business after the implementation partner leaves, the question changes. That's the shape the next eighteen months of operational AI deployment should look like in this market. It's not a smaller version of enterprise AI. It's a different shape entirely.
What does the answer to where will the work live look like inside a specific business? That's the conversation worth having.