Sam Altman’s recent warning to OpenAI staff that the coming months would be "difficult" as Google rolls out Gemini 3 caused a stir in the tech world. For many, the idea that Google has taken a decisive lead seems to be rather a shock. For us it is simply the validation of a strategy we committed to months ago.
We have been arguing for a while that despite the early consumer dominance of ChatGPT, the real AI race would be won on infrastructure, integration, and data. Our preference for the Google ecosystem has never been about ‘fanboyism’, it was always based on capability and a calculated decision based on where we saw the enterprise market going.
At our "Chat Revolution" event back in the summer we outlined why the old ways of customer experience using broad segmentation to provide a few content variants is long overdue for replacement. Without question the future is individually personalised communication agents picking up on customer data points and generating genuinely engaging, individually personalised dialogue.
To deliver this at scale, you need more than a clever model; you need an orchestration engine. This is why we chose Google’s Vertex AI. It allows us to move beyond simple chatbots to complex "Controller Agent" architectures where a master agent intelligently routes user intent to specialized sub-agents, e.g. for specific product queries, support tasks or to bring multiple request strands together into a coherent whole.
We found that building these robust, next-generation customer services was faster, cheaper and more scalable within the Vertex ecosystem than trying to cobble together similar capabilities elsewhere.
We have also been arguing that the biggest hurdle for enterprise AI is trust. A model that is prone to hallucinations becomes a massive problem when it is directly supplying information to customers.
This is where Google’s level of mastery with the world of information on the internet becomes a huge asset.
In our recent study on LLM visibility we've observed Gemini's ability to intelligently ground its responses using Google search results and this offers a level of reliability that competitors struggle to match.
We aren't just looking for creativity; we are looking for accuracy. Google bridges the gap between traditional search SEO and generative answers in a way that is uniquely valuable for brands.
One of the primary reasons we foresaw Google’s resurgence is the sheer depth of their integrated data. It isn't just about the model; it’s about what the model can access.
Google’s dominance in search and their years of building the Knowledge Graph provided a structured data foundation that allows Gemini to conduct deep, accurate research instantly. We are seeing this power unleashed now with Google’s new agentic shopping developments.
By combining Gemini with the Google Shopping Graph which understands billions of products and their attributes the AI isn't just guessing, it is acting with the power of the world's organized web data behind it.
This allows for a level of "agentic" performance (by which we mean actually getting things done) that standalone models struggle to replicate.
We were early adopters of AgentSpace, now fully realized as Gemini Enterprise. Whilst we get as excited as the next person by the amazing things you can do with Veo 3.1 and we have no end of fun with Nano Banana, we are also motivated by the rather more mundane realities of business such as enterprise data security and workflow integration.
The ability to have an AI that lives securely within our existing workspace understanding our documents, spreadsheets and internal data within our own secure enterprise data environment is the difference between a toy and a tool.
The "game over" comments from some industry insiders regarding Gemini 3’s performance are telling, but we have felt the race was something of a formality since we realised we needed an ecosystem and not just a chatbot.
It is also clear that Google is going all-in on AI. CEO Sundar Pichai recently stated that when it comes to AI infrastructure, "the risk of under-investing is dramatically greater than the risk of over-investing". Nor is this just talk with recent commitments like the $40 billion investment in AI infrastructure in Texas Google are deploying resources on a scale that makes long-term dominance almost inevitable.
The coming months will be tough for some but ultimately fierce competition benefits us all. Whether it’s OpenAI fighting back with renewed innovation or newcomers like DeepSeek proving that elite performance is achievable on smaller budgets, it is this competition which drives the rapid progress that we should all welcome.
But for now, we are working hard on using Google’s best-in-class capabilities to deliver genuinely revolutionary use cases and the coming months and years look incredibly exciting.
Malcolm Clifford - Data Science Strategy Director
Joe Crawforth - Head of Research and Development
References:
1 https://www.eweek.com/news/sam-altman-economic-headwinds-google/
3 https://www.jaywing.com/views/news/what-google-will-reveal-at-our-manchester-marketing-event/
4 https://blog.google/products/shopping/agentic-checkout-holiday-ai-shopping/
6 https://blog.google/inside-google/company-announcements/google-american-innovation-texas/