Why Automation ROI Looks Worse Than It Actually Is

I’m in a business-y mood this week, so here’s another piece of the puzzle that sometimes gets overlooked. This time, it’s a mistake that pushes us into not going for an AI project even if it would make total sense.

The mistake? Looking at the ROI of time saved purely through the lens of salary fraction. Let’s look at an example with simple numbers so we don’t get distracted by math.

• Let’s say you’re a company with a viable business model and you have good economic sense

• You pay someone a $100k annual salary

• If your company makes any sense, that person must create annual value in excess of their annual salary. Multipliers vary by role and industry. Let’s use 2x for easy math: A $100k annual salary gives you $200k of economic benefit

Now let’s assume that part of their current job is an annoying menial administrative task that, for some reason, only they can do, even though it isn’t part of their true value-creating activity. Let’s assume that this takes up a quarter of their working hours:

• 25% of their work goes towards something that creates no direct value

• Only 75% of their work goes towards the value creation.

• That means they only create $150k of economic benefit to the company (2x value multiplier with a 25% penalty multiplier)

Next, we imagine that we could wave a magic wand (AI-powered, no doubt) to make the annoying task go away. How much should that be worth to us?

• The simplistic calculation says: 25% of their time costs us 25% of $100k, so that’s $25k.

• The better calculation says: 25% of their maximum value creation potential is 25% of $200k, so that’s $50k.

So with these simple numbers, we see that the true ROI of business process automation can be much higher than pure salary cost.

Caveats

• These gains can of course only be realized if the worker actually has something better, high-value, to do with the freed up time. For today’s knowledge workers that’s almost certainly true, but needs to be taken into account.

• Can the rest of the system absorb their increased productive output (see yesterday’s post on contraints)

The difference between a “cost savings” versus “value unlock” ROI calculation can be big. Miss this distinction and you’ll systematically underinvest in automation that would actually move the needle.