This week in Las 30,000 people gathered in Las Vegas to hear the latest and greatest from Google Cloud. What they heard was generative AI all the time. Google Cloud is primarily a cloud infrastructure and platform provider. If you didn’t know this, you might have missed it in the onslaught of AI news.
Not to downplay what Google had to offer, but much like Salesforce did at its travel roadshow in New York City last year, the company only mentioned its core business in passing – except in the context of generative AI, of course.
Google has announced a series of AI improvements designed to help customers take advantage of the Gemini Large Language Model (LLM) and improve productivity across the platform. That’s obviously a worthy goal, and during the main keynote on the first day and the developer keynote the following day, Google peppered the announcements with a healthy number of demos to demonstrate the power of these solutions.
But many seemed a bit too simple, even considering they had to be squeezed into a keynote with limited time. They relied primarily on examples within the Google ecosystem, where almost every company has much of its data in repositories outside of Google.
Some of the examples actually felt like they could have been implemented without AI. For example, during an e-commerce demo, the presenter called the salesperson to complete an online transaction. It was intended to demonstrate the communication skills of a sales bot, but in reality the buyer could have easily completed this step on the website.
That’s not to say there aren’t some powerful use cases for generative AI, be it creating code, analyzing a corpus of content and being able to query it, or being able to ask questions about the log data to understand why a Website is down. Additionally, the task- and role-based agents the company has introduced to help individual developers, creators, collaborators, and others have the potential to leverage generative AI in tangible ways.
But when it comes to building AI tools based on Google’s models, rather than using the ones that Google and other vendors develop for its customers, I feel like they face many of the obstacles that arise could, gloss over the path to a successful generative AI implementation. While they tried to make it sound easy, in reality it is a huge challenge to implement advanced technology in large organizations.
Big changes are not easy
Similar to other technology leaps over the last 15 years – be it mobile, cloud, containerization, marketing automation, whatever – it was implemented with many promises of potential gains. But these advances each bring their own level of complexity, and large companies are proceeding more cautiously than we might imagine. AI seems to be a much bigger boost than Google, or frankly any of the major players, admits.
What we’ve learned from these previous technological shifts is that they come with a lot of hype and lead to a lot of disillusionment. Even after several years, we have seen that large companies that should perhaps be taking advantage of these advanced technologies are still just experimenting with them or even abandoning them altogether years after their introduction.
There are many reasons why companies fail to reap the benefits of technological innovation, including organizational inertia; a fragile technology stack that makes it difficult to adopt newer solutions; or a group of internal company naysayers shutting down even the most well-intentioned initiatives, be they legal, HR, IT or other groups who, for a variety of reasons, including internal politics, continue to simply say no to substantive change.
Vineet Jain, CEO of Egnyte, a storage, governance and security company, sees two types of companies: those that have already made a significant move to the cloud and will have an easier time adopting generative AI; and those who move slowly and are likely to have difficulty.
He speaks to many companies that still have much of their technology in place and still have a long way to go before they start thinking about how AI can help them. “We talk to a lot of ‘late’ cloud adopters who haven’t started their digital transformation journey or are still at a very early stage,” Jain told TechCrunch.
AI could force these companies to think hard about driving digital transformation, but they may struggle to start so far back, he said. “These companies need to solve these problems first and then leverage AI once they have a mature data security and governance model,” he said.
It was always the data
With major players like Google, implementing these solutions sounds easy, but as with any sophisticated technology, looking simple on the front end doesn’t necessarily mean it’s straightforward on the back end. As I’ve heard many times this week, the “garbage in, garbage out” principle still applies to the data used to train Gemini and other large language models, and that’s even more true when it comes to generative AI.
It starts with data. If your data house is not in order, it will be very difficult to get it in shape to train the LLMs for your use case. Kashif Rahamatullah, a Deloitte director who oversees his company’s Google Cloud practice, was largely impressed by Google’s announcements this week, but still acknowledged that some companies that lack clean data are having trouble implementing them generative AI solutions will have. “These conversations can start with an AI conversation, but it quickly becomes, ‘I need to fix and clean my data, and I need to have everything in one place, or almost in one place, before I.’ Start getting the real value out of generative AI,” said Rahamatullah.
From Google’s perspective, the company has developed generative AI tools to more easily help data engineers build data pipelines to connect to data sources within and outside the Google ecosystem. “It’s designed to really accelerate data development teams by automating many of the very labor-intensive tasks involved in moving data and preparing it for these models,” says Gerrit Kazmaier, vice president and general manager of databases, data analytics and Looker at Google, said TechCrunch.
This should be helpful in connecting and cleaning data, especially in companies that are further along the digital transformation journey. But for companies like the ones Jain mentioned that haven’t taken meaningful steps toward digital transformation, this could lead to greater difficulties, even with the tools Google has developed.
That doesn’t even take into account that AI comes with its own challenges beyond just implementation, whether it’s an app based on an existing model or, especially, trying to build a custom model, says Andy Thurai, an analyst at Constellation Research . “When implementing both solutions, companies need to think about governance, liability, security, privacy, ethical and responsible use and compliance of such implementations,” Thurai said. And none of this is trivial.
Executives, IT pros, developers and others who went to GCN this week may have been looking for Google Cloud’s next steps. But if they haven’t set out to pursue AI, or simply aren’t ready as an organization, they might have come back from Sin City a little shocked by Google’s total focus on AI. It could take a long time for companies that lack digital skills to fully leverage these technologies, beyond the more comprehensive solutions offered by Google and others.