Principles for Starting High-ROI AI Initiatives in Your Company

Looking at the news headlines about AI makes you feel FOMO about your company. It makes you think you really should be implementing this new tech, or your business will be left behind and your clients will leave you for more competitive companies that can do more for less.

You’ve tried a few things, but they don’t seem like they will change everything as everyone says. And you don’t want to make a big investment that won’t deliver any meaningful results.

This is actually the reality for most companies:

• A 2025 MIT research report found that 95% of enterprise GenAI projects showed no measurable ROI.
• A recent report from McKinsey found that while more businesses are deploying AI hoping their investments will start paying off, the number of people reporting increased earnings from their AI initiatives hasn’t changed.

You might be thinking, ‘If it didn’t work for them, why would it work for me when I know nothing about AI?’

Well, the reason most companies can’t get consistent ROI from AI is that its results are not deterministic. This means that even if you give it the exact same instruction, you will get slightly different results.

While there are techniques for getting more deterministic outputs, you will still face tons of uncertainty. And under uncertainty, the best approach is to run projects like a scientist runs experiments.

Follow these principles to start implementing high-ROI projects in your company:

• Choose 1–2 functions that can have the most impact. MIT Sloan Management Review recommends implementing solutions in just one or a small number of business functions. If your company is small, don’t run more than one experiment at a time.

• Target the right process. Choose a repeatable process where you waste too much time, spend too much money, or where you can increase your revenue the most.

• Set a small and clear goal. What you try must be defined and narrow, but impactful enough to prove results:

  • 1 specific problem

  • 1 single process

  • 1 solution to try

• Set your limits. It’s best to establish a narrow budget and commit to it. Define the maximum amount of resources you are willing to spend: time, money, people, and other assets.

• Choose clear success metrics. Ensure your metrics are clear, meaningful, and easy to measure, and make sure to measure both before and after the initiative.

MIT notes that these function-focused, narrow AI initiatives often deliver some of the most tangible ROI, and they are particularly effective at building confidence when starting to invest in AI.

By limiting your scope, you create proof points and learnings that allow you to replicate results across other processes and areas. Although results might differ at a larger scale, this approach is still super useful for learning the core principles of implementing AI.

You’ll start accumulating knowledge on the best ways to test solutions, understand limitations, set realistic timelines, and prepare your team.

Try to maintain a structured process with each iteration, building your AI playbook little by little. Before long, you’ll join the top 5% of companies getting positive ROI.