What is actually going on under the AI hood?
24 Nov 2025
What is actually going on under the AI hood?
Some of my fav slides from Benedict Evans latest report:
1. Where has all the AI investment gone?
It’s eye watering the amount of investment that has gone into AI and the infrastructure/ecosystem that enables it. But what have we got to show for it? More models, that are slightly better and more niche experiments, everyone hoping they create ‘the one’ that dominates the market. On the ground companies are finding it tough to show cold hard results for the time and money they're putting into initiatives, will 2026 start showing ROI?
2. Who should lead the AI charge?
Is it the CEO/CTO/CMO, which one? Is it all of them? Is it a Deloitte or Publicis you need to come in and help? Is it an indie specialist? Reality is, it’s so new, it’s a mix of leadership roles and specialists. And what is the most helpful thing right now is to start with the problem in mind, with the customer at the heart of it. So if you want to improve customer satisfaction using GenAI, does it help or detract from that?
3. It’s time to think about brand alongside just models
Models are the product at the moment and ChatGPT, with their ‘first out the blocks' advantage’, is still on the whole leading share. As we saw with their recent campaign, how do you start building a brand world around a hyper growth industry and fast moving models? No one yet has felt like they’ve distinctly started to carve this out. We’re in the ‘all grey desktop’ era…
4. 800m WAU but only 5% paying for it (apparently)
We will likely see an explosion of different ways to try and capture value from customers from ad led models to subscription or even potentially countries subsidising it to increase illiteracy of different AI tools. Who will be able to create a sustainable model that consistently adds value people/companies want to spend on
5. It’s still embedding into the day-to-day
The whole “everyone is using it” is true to a level but maybe not to the size and scale we all think. People might be using it monthly and weekly but not as much daily. The embedding into everything we do will take time, upskilling and a clear point of value that a lot of use cases are currently missing, which goes back to the point 1 around the return seen on investments.
Image credit: Benedict Evans, full report here.
Image credit: Stephen Shore, Uncommon Places.