In the working paper “Beyond AI Exposure: Which Tasks are Cost-Effective to Automate with Computer Vision?” [PDF] MIT researchers with funding from MIT-IBM Watson AI Lab looked into the potential for serious labor disruption due to deployment of AI in certain functions that have, up until now, required humans. In less intelligent words, they wanted to figure out if robots are taking your job. Not all jobs, specifically “vision tasks” which in our little corner of the world most closely applies to auditors’ much-loathed inventory counts.
The short (and good) news from the paper: their findings suggest that AI job displacement will be substantial, but also gradual. And for now, mass deployment simply isn’t worth it for most businesses, even massive employers like Walmart.
We find that at today’s costs U.S. businesses would choose not to automate most vision tasks that have “AI Exposure,” and that only 23% of worker wages being paid for vision tasks would be attractive to automate. This slower roll-out of AI can be accelerated if costs falls rapidly or if it is deployed via AI-as-a-service platforms that have greater scale than individual firms, both of which we quantify.
The premise hinges on current costs of deployment, meaning it costs more to replace you with the AI we’ve got now than it does to keep using your eyeballs to count widgets.
One example they offer is a bakery:
Consider a small bakery evaluating whether to automate with computer vision. One task that bakers do is to visually check their ingredients to ensure they are of sufficient quality (e.g. unspoiled). This task could theoretically be replaced with a computer vision system by adding a camera and training the system to detect food that has gone bad. Even if this visual inspection task could be separated from other parts of the production process, would it be cost effective to do so? Bureau of Labor Statistics O*NET data imply that checking food quality comprises roughly 6% of the duties of a baker. A small bakery with five bakers making typical salaries ($48,000 each per year), thus has potential labor savings from automating this task of $14,000 per year. This amount is far less than the cost of developing, deploying and maintaining a computer vision system and so we would conclude that it is not economical to substitute human labor with an AI system at this bakery.
Thus:
The conclusion from this example, that human workers are more economically-attractive for firms (particularly those without scale), turns out to be widespread. We find that only 23% of worker compensation “exposed” to AI computer vision would be cost-effective for firms to automate because of the large upfront costs of AI systems. The economics of AI can be made more attractive, either through decreases in the cost of deployments or by increasing the scale at which deployments are made, for example by rolling-out AI-as-a-service platforms (Borge 2022), which we also explore. Overall, our model shows that the job loss from AI computer vision, even just within the set of vision tasks, will be smaller than the existing job churn seen in the market, suggesting that labor replacement will be more gradual than abrupt.
To determine if similar technology would be economically viable at audit firms, you’d have to calculate how much time it takes to do inventory counts, figure out how much of an auditor’s salary goes toward that activity, multiply that by however many auditors it takes to get all done, and then put that up against the costs of a computer vision replacement. Partners are no doubt doing that math as we speak.
A CPA Canada/AICPA paper entitled “The Data-Driven Audit: How Automation and AI are Changing the Audit and the Role of the Auditor” [PDF] published in 2020 talks about how computer vision can assist auditors on this specific and annoying task:
Inventory counts
With computer vision, an AI-based app can look at millions of pictures taken from cameras (whether statically mounted in a warehouse or mounted on moving drones) and identify articles. Articles that have indexing information (such as bar codes) are even easier to identify and if the “eye sees them all,” then it can count them all, giving the auditor the ability to obtain more coverage.
The risks identified by that paper are the reliability of images “(e.g., are the images being viewed authentic or is there a risk that the image could be manipulated?)” and difficulty accessing the inventory. As in, they can force first years to cram themselves into a cold warehouse among giant stacks of boxes with mere inches between them or shit-smeared barns full of livestock inventory, a machine not so much.
“If task automation of that extent were to happen rapidly, it would represent an enormous disruption to the labor force,” said the MIT researchers in their paper. “Conversely, if that amount of automation were to happen slowly then labor might be able to adapt as it did during other economic transformations (e.g. moving from agriculture to manufacturing). So, making good policy and business decisions depends on understanding how rapidly AI task automation will happen.”
Probably a lot sooner than we think if the cost to do it drops significantly.
