AI strategies and projects: The hope, the fear and everything in between

When it comes to AI strategies and projects we've seen the hope, the fear and the reality all within a few days.

Google Cloud Next 2025 provided an interesting backdrop to the state of agentic AI, enterprise use cases and how AI is altering corporate strategies. Here's a quick tour.

The hope

Let's start with hope shall we? Google Cloud outlined how it is working with Sphere Entertainment, Magnopus and Discover Warner Bros. to bring “The Wizard of Oz” to Sphere. The mission: Revamp the 1939 classic with AI technology that was essentially being developed on the fly. The companies developed a "super resolution" tool to turn celluloid frames from 1939 into high-definition imagery using Google Cloud models, AI outpainting and models to expand the scope of screens.

Now this project could have gone terribly wrong. If the AI enhanced any of the characters so they weren't realistic or looked like animation or CGI, the audience would revolt. I've seen enough of the Sphere version of Oz to know that you won't, but you'll have to wait until August to find out.

Nevertheless, what stuck out most about the Oz project was the passion about the creative side of the equation. Google CEO Sundar Pichai said AI is about augmenting creators and giving them new tools. "Filmmaking has been and will always be a profoundly human invention," said Pichai, who said the goal was to do things that have never been done while maintaining the original spirit of the film.

Google Cloud CEO Thomas Kurian said with a project like Oz at Sphere needs to apply AI with intent and responsibility. Models had to be fine-tuned on the source material--notebooks on cameras, film and still shots--to be historically accurate.

• Google Cloud CTO Grannis on the confluence of scale, multimodal AI, agents
• 9 Google Cloud customers on AI implementations, best practices
• Google Cloud CEO Kurian on agentic AI, DeepSeek, solving big problems

The artistic and technical collaboration behind bringing Oz to Sphere revolves around co-creation. "The artist collaborated with the technologist and then the AI spoke back to us," said Sphere CEO Jim Dolan, who noted that he felt like a partner instead of a customer of a vendor.

Now you could argue that bringing Oz to Sphere was a unique project that is hard to replicate from a collaboration perspective. The project had passion and was almost a cause. But it's hard to miss the ideal mix of ingredients where AI amplifies human creativity, and collaboration and creativity flourish.

The fear (at least for humans)

Shopify CEO Tobias Lütke made some news with a staff memo that boiled down to the following:

• AI usage is baseline expectation. Shopify employees all need to be using AI naturally and frequently as part of the day-to-day workflow.
• AI isn't an option. It has to be learned and performance reviews for employees will include AI skills and usage.
• The human hiring freeze. Teams need to demonstrate that needs can't be fulfilled by AI before requesting a hire.

Overall, Lütke said Shopify is all-in on AI and it's a cultural change. Lütke didn't say anything that other CEOs haven't said before. He was just a bit more direct about preferring AI as a force multiplier over hiring. This take shouldn't be a shocker given the steady stream of technology company layoffs in the last two years.

Lütke caught flack on social media for his memo, which was leaked so much he just posted it.

Here's why Shopify's AI approach may turn out to be a disaster. AI projects are about change management as much as technology and fear doesn't necessarily drive change over the long run.

The risk of Lütke's strategy is fairly obvious:

• Shopify's tone with AI is urgent, but some employees will be overwhelmed and left out. Change fatigue could be real.
• It's unclear from Lütke's memo how much training and support for this AI knowhow will come from Shopify.
• Cultural resistance and lack of buy-in. Lütke's requirement that teams will have to justify headcount requests by proving AI can't do the job saps morale. It takes little to make the hop that AI-first means downsizing later.
• AI usage as a performance review requirement is heavy handed and not all roles are going to be 100% AI driven.
• Lütke's memo is likely to be a poster for how AI is going to take jobs. My hunch is that Shopify employees are going to feel like the person who had to train her replacement when a downsizing was on deck.

The in-between works in progress

I spent a good chunk of my Google Cloud Next time checking out customers adopting AI agents and generative AI. These vignettes weren’t just about productivity and included a lot of transformation.

Some high level takeaways include:

• A general feeling of optimism about AI agents.
• Time to value matters.
• Google Cloud has its ground game down with its go-to-market strategy and has organized around industries.
• The use cases for AI are almost infinite. When a vendor can put more than 600 use cases in a blog, you know you have some mojo.

We're in the early innings of what's possible with AI. "Agents are in early stages right now and starting to be proven out," said Google Cloud CTO Will Grannis. "AI agents need to scale out with authentication, stability, reliability."

Here's a look at just some of the customer stories from Google Cloud Next.

• Lloyds Banking Group bets on Google Cloud for AI-driven transformation
• Healthcare leaders eye agentic AI as next frontier for clinicians, patients
• Google Cloud, UWM partner as mortgage battle revolves around automation, data, AI
• Lowe's eyes AI agents as home improvement companion for customers

The overall vibe is that enterprises are looking to scale AI to transform their businesses.

Citigroup CTO David Griffiths said the banking giant is leveraging Google Cloud as part of its multi-year transformation. "Anywhere we work with digital transformation, AI can help. We are embracing AI as a universal enabler," said Griffiths.

Griffiths outlined the strategy:

• "We've been guided by a couple of simple principles, with taking a very deliberate approach. We build a simple, scalable, secure, multi-model platform that has centralized controls and observability, so we can keep everyone safe, and we can learn and observe across the breadth of all of our AI interactions."
• "We think about the impact of AI in two dimensions: General, horizontal, assistive AI tools that have very wide applicability. These may only give you 1% to 3% of productivity back, but you scale that across the company, this really adds up. And you have to complement that with deeper AI verticals, specialized capabilities for the specialists within your workforce."
• "A scale footprint allows us to maximize the impact as this technology advances. Google is at the frontier of AI development, and we want to have a mini lag between AI innovation and AI impact."

Griffiths added that Citigroup had about 1,000 use cases in 2024 at various stages. Those use cases were horizontal and could benefit the entire organization. In 2025, Citigroup is focused on scale and depth and "industrializing our AI verticals" for customer servicing, fraud detection, finance and sales and marketing.

Citigroup’s journey

• Here’s what technology buyers say about AI, technology, transformation
• Here's why generative AI disillusionment is brewing
• Digital transformation should pay for itself with real business value

During Kurian's keynote, Verizon was cited as a Google Cloud Customer Engagement Suite customer that has advanced over the past year. Verizon was an early customer of the Customer Engagement Suite and ramped usage over the last year.

Verizon is using a personal research assistant powered by Google Cloud to give its 28,000 care reps personalized information about a customer's needs.

The wireless giant is seeing better service, lower wait times and improving customer satisfaction scores.

Compared to Citigroup and Verizon, Pearson isn't as far along with its AI plans. Pearson CTO Dave Treat said the education and learning company is transforming to "use AI to help educators and students transform learning across all stages of life."

The company is betting on agentic AI.

"We've realized it's time for us to think outside the book, and we're really just at the beginning now of creating super effective, personalized learning experiences using agents," said Treat. "We're envisioning a team of agents working together on behalf of educators and students using natural language interfaces integrating all of the tools and resources that they need guided and shaped by our learning science and trusted content."

Treat noted that agentic AI will change its engineering and software development lifecycle. Pearson is using specialized agents for code, documentations, and testing scripts.

According to Treat, Agentspace will be the control plane for multiple agents including ones from Salesforce and ServiceNow. Treat's take highlights how hyperscalers may be best suited to orchestrate AI agents across systems. "Just like humans, there's going to be the right agent for the right job," said Treat.