Working at Al Jazeera changes how you think about technology. When the platform you're responsible for serves 100 million people in some of the world's most politically sensitive contexts, the question isn't just "does this work?" — it's "what happens when this goes wrong, and to whom?"
That question became sharper when we started integrating AI capabilities into our digital infrastructure. This post is about what I learned — not the technology decisions, but the human ones.
The Speed Asymmetry Problem
AI tools move fast. Editorial standards move deliberately. This asymmetry is the central tension in any newsroom AI initiative, and most organisations underestimate it.
When I was leading the strategic technology partnership for AI capabilities at Al Jazeera, one of the first things I had to do was build a shared vocabulary between engineering, editorial, and legal. Engineers talk about accuracy rates. Editors talk about trust. Legal talks about liability. These aren't the same things, and treating them as if they are is how you end up with a tool that technically works but that nobody uses — or worse, one that gets used in ways that create real harm.
Where AI Creates Genuine Leverage in Media
There are places where AI is genuinely transformative in a newsroom context:
- Content discovery and search. Finding relevant archival footage and articles across massive asset libraries is tedious work that AI handles well.
- Translation and localisation at scale. For an organisation publishing in Arabic and English with a global audience, this has real operational value.
- Metadata tagging and content classification. Unglamorous but high-value — good metadata is what makes search and discovery work.
- Workflow automation in production pipelines. Moving assets, triggering processes, formatting content for different distribution channels.
Where Human Judgment Is Non-Negotiable
And there are places where I'm sceptical that AI should be anywhere near the decision:
Editorial judgement on what to cover and how. AI can tell you what's trending. It cannot tell you what matters. That distinction is the entire job of a journalist.
Verification of sensitive claims. Pattern-matching on text is not the same as understanding context, motive, and consequence. In breaking news situations, the confidence that AI expresses is not calibrated to the situations where it matters most.
"The creative human always finds what pattern-based systems miss. This is not a temporary limitation. It's structural."
What Good AI Governance Actually Looks Like
The organisations getting this right are the ones that started by asking "what are we not willing to automate?" before asking "what can we automate?" That sequencing matters. It puts humans in charge of drawing the line, rather than asking them to stop a runaway process after the fact.
In practice, this means: clear ownership of AI-assisted outputs, audit trails, and escalation paths when the system produces something unexpected. It means editorial leadership being in the room when tooling decisions are made, not just consulted after the fact.
If this resonated, I'm easy to reach. Or connect on LinkedIn.
Did this resonate?
Share this article