To Get ROI from AI and Other Investments, Use a Scorecard
Did you know…AI is an expense?
Many companies seem to have forgotten that in the race to be first in AI adoption. But the bill is coming due, as the big frontier AI companies are reducing subsidies and raising rates on their most popular products.
AI has been boundary-breaking in many ways, but one way it’s staying the same is how it does (or doesn’t) add value to your business.
There are only three main ways to drive ROI with AI (or any other tool, software, or business investment):
1. Increased productivity. The company has higher throughput and gets more done with the same amount of resources (labor, time, etc.)
2. Greater value. The company delivers a better product or service to customers, which leads to increased satisfaction, higher customer retention, and a competitive advantage in the market.
3. Lower costs. The company delivers its product or service at a lower cost, improving the unit economics and increasing margins.
…That’s kinda it.
The fundamentals of business don’t change much. If you want a return from an LLM—or any other software, platform, piece of technology, or even person—you’ve got to identify which of these three ways that investment will drive ROI for your company.
***
Recent research and reporting shows that AI is not always delivering the hoped-for ROI. There are likely multiple factors that explain those results. But let me ask you a simple question:
If you could do your current job more efficiently—and you were a normal, hard-working team member instead of a slightly unhinged entrepreneur—would you start volunteering to take on extra work? Or would you take the win of personal productivity and slowly, maybe even subconsciously, start to expand your work to fill the available space without adding anything to your plate?
I believe this is one of the challenges of realizing ROI from AI adoption. And I believe we have more power to overcome that challenge than we think.
Part of the agreement with AI for the last few years has been that it is experimental. We’re tinkering, we’re testing, and then we see what happens. R&D is an important business function, but we’re hitting a point where AI needs to prove its worth.
One of the best tools for taking a test and making it part of day-to-day operations? The company scorecard. If you aren’t familiar with the scorecard, I highly recommend checking out this Inc. article.
Here’s the quick guide to operationalizing with a scorecard:
Step 1: Figure out what you’re trying to achieve
Step 2: Determine which measurables will likely help you achieve your goal
Step 3: Set a target metric for weekly or monthly (at most) reporting on the measurable
Step 4: Meet every single week and track your progress
Here’s an example: Suppose you want to use AI to increase productivity at your company, which makes and sells widgets.
Step 1: Your goal is to increase productivity by making more widgets with your existing resources.
Step 2: An obvious measurable is widgets made each week. You typically make 1,000 widgets each week.
Step 3: To justify the cost of your AI investment, you need to increase the number of widgets made by at least 10%. You want to do better than break even, so you set a target that is 20% higher than current widget output: 1,200 widgets each week.
Step 4: Assign an owner to be accountable for widget output, add it to the scorecard, and track it every single week.
Now, keep this in mind: your scorecard might be wrong. Scorecards often are in the beginning, but the iterative process can’t begin until we manage something. You may determine you can only increase widget output by 15%, so you adjust the metrics downward. Or perhaps your team easily meets the 20% increase benchmark, so the next quarter you increase to 25%.
You cannot fully guarantee the outcome, but you can reverse-engineer a process that stacks the odds in your favor.
The most extreme version would go something like this: I want to see 20% increase in efficiency from our AI investment. I’m cutting 20% of our workforce because that will (likely) ensure that our team is 20% more productive.
I’m not recommending that as your strategy. It’s not what I’m doing in my own companies. I am, however, illustrating the point that if you want a particular outcome, you have to set that outcome as the standard.
Otherwise, your strategy to get ROI from AI (or any other investment) is basically one big shrug. 🤷♂️
