Dream big, but start with small, practical steps.

Business leaders have a compulsion to go big. It’s kind of their mandate. There’s nothing wrong with being ambitious or having lofty dreams. And the pressure is on for everyone to have an AI strategy or an AI-powered product. But that kind of "big goal, vague plan" approach associated with moonshots is exactly what dooms projects (and careers). 

Plenty of businesses involved in AI are doing moonshots at significant cost. OpenAI raised USD 6.6 billion in late 2024, having already used the USD 10 billion it raised in 2023. But I’m guessing your company doesn’t have those sort of resources or the luxury of waiting five years before that investment starts paying off. 

We’re lucky enough to see and be involved with seriously successful AI projects, and I can tell you that while most of them are transformational to the business, none of them started off as moonshots. Those are for the rich kids. We need to be smarter than that. 

Remember what the moonshot involved  

Let’s reflect on the original “moonshot." The Apollo program was conceived during the height of the Cold War and aimed to leapfrog the Soviets, who achieved most of the early successes in the space race. It was a successful project, but at what cost?

Overall the U.S. spent about $200B in today's dollars (roughly the GDP of Portugal) which was about $135B over original estimates. It was a good, hard problem and we did it. And what was the ROI? Bragging rights? Velcro? Tang? If you're a global superpower with a bottomless budget, I guess you can do that sort of thing.

The business landscape is littered with expensive moonshots. 

It’s easy to think that AI is a new technology that has only emerged in the last couple of years. But it has been around for a long time and has had plenty of time to wreck reputations and careers with high-profile (and presumably many low-profile) moonshot fails. 

Some of the eye-popping ones:

• IBM invested $62 million to develop an AI called Watson for cancer treatment recommendations. The system provided unsafe and incorrect suggestions, such as recommending a drug that could worsen bleeding in a patient with severe bleeding. The project was ultimately cancelled, and the business unit was sold off for parts, resulting in billions in losses.
• Zillow’s AI-driven home-flipping program, iBuy, relied on property valuation models that overestimated market trends, leading to significant financial losses. After losing USD 380 million, the company shut down the program and laid off 2,000 employees (25% of its workforce).

Why moonshots are problematic 

For most businesses, they come with three big problems:

1. No playbook: AI moonshots are uncharted territory. Building something brand new means no one’s done it before, and there’s no playbook. First attempts are notoriously expensive and prone to failure. A moonshot could sink your budget if you’re not ready to absorb the cost of mistakes.
2. Unclear ROI: Big AI projects take years to deliver value, if they deliver at all. That’s like pouring water into a leaky boat and hoping it stays afloat. Most businesses can’t afford to spend millions on a project that might only pay off five years down the road.
3. Shifting Landscape: AI is evolving fast. The tools, techniques, and even the ethics of AI are shifting under our feet. A moonshot might take years to execute—and by the time you finish, the landscape will look completely different.

So, if moonshots aren’t the answer, what is?

Jello shots, the moonshot alternative

I'm joking with the Jello Shot moniker. Here's how we believe you should approach your AI transformation.

• Enlist a can-do team. You need a cross-functional group of people that is hungry, impatient, and empowered to fail. We call these tiger teams.
• Stay focused on value. It's SO easy to pick toy projects that deliver no value to the business. Do the work of identifying use cases that are central to how your company delivers value. Transforming these processes with AI automation can deliver 10x to 100x ROI.
• Start small, think big. Don't make an elaborate plan to transform your entire business. Find one thing that really matters, and attack that first. Then find another thing. Keep going until the business is transformed.

Start small, go deep, and focus on solving problems that are central to your business. To do this, you need to deploy cross-functional teams of a specific nature. We call them Tiger Teams. 

Building your tiger team

We've found that successful AI projects start with what we call Tiger Teams—small, focused groups with clear accountability. Here's the winning recipe:

1. A business SME at the core: Not just any subject matter expert, but someone who lives and breathes the problem you're trying to solve. I've seen too many AI projects fail because they were driven by tech people who didn't truly understand the business context. Your SME ensures every decision ties back to real business value.
2. Technology team with guardrails: Our tech teams are enhanced by working in a new way: they integrate the SME's into the dev cycle. They bring deep technical expertise but also the humility to understand that technology serves the business, not vice versa. 
3. Product and analytics specialists: I'm a product guy, so I'm biased, but having a strong product manager and data analyst on the team is crucial. They keep the project focused on user needs and measurable outcomes. We work in 4-week delivery cycles with clear, value-centric goals and metrics.
4. End Users: Every project is unique, but it's generally a good idea to start testing the product with real end-users as soon as possible. A humble team will incorporate their feedback early for a much stronger product later on.

Identifying valuable use cases

We use two essential frameworks at Machine & Partners to ensure success. They're so important that we discuss them frequently (on our website and in articles like this one), and I'll share the essentials here.

First, use DEW to identify promising AI opportunities:

• Data: Do we have the data to support a successful AI project? 
• Expertise: Is there critical knowledge trapped in people's heads? 
• Workflows: Are there repeatable processes ready for enhancement? Look for tasks that are routine but require human judgment.

Then, run potential projects through the VVVroom test to assess feasibility:

• Viability: Can you build it with your current resources and technology?
• Value: Will it deliver meaningful ROI within 6 months (or whatever ambitious but attainable period you think is appropriate)?
• Velocity: Can you get a working prototype in (example) 8 weeks?

Big vision, narrow focus

The smaller the scope, the greater the impact. When you focus on one high-value use case, you:

• Deliver measurable ROI faster (think weeks or months, not years)
• Earn confidence in AI within your organization through quick wins
• Create a repeatable process for future projects
• Build the foundation of an extendable platform 
• Keep costs under control 

Chaining together small projects will help you learn and update your strategy as you go. It's like navigating in a jungle. On a map it might look like a straight line from A to B, but once you're under the canopy you realized that the fastest way forward is actually much more circuitous.

The good news is that while there are many examples of moonshots failing, there are plenty of success stories for small, deep, and impactful AI projects that scale up. Here are two examples that dreamt big but started with small, practical steps:

• Flipkart’s AI-powered search and bots: Indian e-commerce platform Flipkart began with text and visual search pilot projects. Their approach is notable for cross-functional teams (like our Tiger Teams). These initiatives were expanded to include conversational bots capable of handling multiple languages (initially 11)  and advanced features like user intent detection, language translation, and speech-to-text functions. Now, they’re expanding into computer vision. 
• Proctor & Gamble’s digitalized operations: P&G started with small pilot programs focused on specific business use cases, such as AI, which helps minimize the overpacking of paper towel rolls by better predicting sheet lengths. Those programs have expanded and scaled to digitize operations across over 100 sites globally.  

Choose your next move

If you're considering an AI project, ask yourself: Is it a moonshot or a jello shot? Starting small doesn't mean thinking small. It means being smart about where you place your bets.

Leave the moonshots to the billionaires. Let's focus on what works. Get in touch if you want to get value from your AI projects and see results within weeks or months, not years. We deliver ambitious, transformative AI without sending your bills to the moon and back.