Most conversations about AI in business start with a forecast. This one starts with results that organisations have already published, and it is honest about what each one proves.
We collected evidence across twelve industries for our own industry pages. Where a figure comes from a controlled study we say so. Where it is a company's own claim, a vendor case study or a projection, we say that too. The full list, with links, sits on each industry page.
The controlled studies
A handful of rigorous studies set the baseline for what AI does to everyday knowledge work.
- Customer support. A study of 5,179 support agents found AI assistance raised issues resolved per hour by 14% on average and by 34% for less experienced agents (NBER, 2023).
- Professional writing. In an experiment with 453 professionals, AI cut the time taken on writing tasks by 40% and raised quality by 18% (MIT News on Noy and Zhang, Science, 2023).
- Consulting work. In a Harvard Business School and BCG study of 758 consultants, those using AI finished tasks more than 25% faster with more than 40% higher quality, but only on tasks within AI's capability. On a task outside it, they did worse (Harvard Digital Data Design Institute).
- Software. Developers with an AI assistant completed a set programming task 55.8% faster (Peng et al., 2023).
The pattern matters more than any single number. Gains are real and large when AI is applied to the right tasks, and they can turn negative when it is applied to the wrong ones. Knowing which tasks are which is the entire game.
Results close to home
Some of the strongest published examples come from the UAE.
- Utilities. DEWA says AI cut the time to refund a security deposit from four days to eight minutes (ZAWYA), and that its business requirement documents now take one day instead of a week (Microsoft).
- Energy. ADNOC reported USD 500m of value from more than 30 AI tools in 2023 (World Oil).
- Retail. Majid Al Futtaim cut customer feedback analysis from seven days to three or four minutes, saving about USD 1m a year (Microsoft).
- Insurance. Tokio Marine Insurance UAE reports claims processing 90% faster, with triple the daily volume (Gulf News).
- Real estate. The Dubai Land Department's AI platform monitored 279,000 property adverts in six months and corrected 29% of them automatically (Gulf News).
- Hospitality. Hilton cut breakfast food waste by 62% across 13 UAE hotels using AI waste tracking (Hospitality Net).
- Healthcare. M42's screening centre in Abu Dhabi runs 2,000 AI-assisted chest X-rays a day, with radiologist workload down 37% at the testing stage (M42).
What the results have in common
Read side by side, the examples share three features.
They target a specific, high-volume process. Refunds, claims, adverts, feedback, X-rays. None of them is "AI for the whole company". Each picked one workflow where the same work repeats thousands of times.
They keep a person in the decision. Claims are approved by handlers, scans are confirmed by radiologists, adverts are governed by the regulator. AI does the reading and preparing. People decide.
They were measured against a baseline. "Four days to eight minutes" only exists because someone knew it took four days. Organisations that cannot say how long a process takes today cannot prove what AI changed.
How to read vendor numbers
Much of the evidence in the market comes from vendors, and some of it is excellent. Still, three questions are worth asking of any figure.
- Was it measured or projected? Several large numbers in energy and logistics are expectations, not results.
- Who reported it? A company's own figure is useful. An independent study is stronger.
- What exactly was measured? Klarna's AI assistant resolved customer errands in under two minutes against eleven before (Klarna). That is a speed result. Later reporting on Klarna's costs is a different story, which is why we never use it as a cost headline.
Where to start
The organisations above did not start with a platform. They started by knowing exactly where their time went. That is why every Thirty Days engagement begins with AI interviews of every employee: so the first solutions target the processes that cost the most, and the results can be measured against a real baseline.
If you want to see what the evidence looks like for your sector, start with the industry pages or talk to us.
