âSoftware is eating the worldâ â Marc Andreessen declared in 2011 to emphasize the massive disruption caused by software companies. Then, in 2019, Tarry Singh wrote a Forbes article titled âSoftware ate the world, now AI is eating software.â This article from Singh had some highlights, and it got me thinking about the greater implications of AI on workflow automation.
We are entering a very interesting time in the world of software development as more and more people jump on the AI bandwagon. Bill Gates relates the AI revolution to the microprocessor, internet and phone. However, it has become abundantly clear to me that not all problems can be easily automated away. And if they can, whatâs not clear is that all customers have the appetite to blindly trust the automation.
The trust problem
Skepticism is the largest barrier to entry in the willingness to accept end-to-end automation of a process or job function. While some automation and AI assistance is fantastic, even in the most sensitive roles, it is important to understand that there will be humans in the loop for a long time.
Companies like Numeric and Doss (in our portfolio), are improving otherwise painstaking, tedious, and error-prone processes. Numeric helps accounting teams close the books faster and with more clarity, and Dossâs âAdaptive Resource Platform (ARP) combines the system of record of a traditional ERP with artificial intelligence and machine learning to create a solution for all of your business applications.â However, these products are not making all decisions for teams; rather, they are making it more efficient for teams to achieve a desired result and drive outcomes. As they gain the trust of their users, they will earn the right to automate more and more.
A case study: the autonomous close
ââLetâs take the example of the autonomous quarter close â a game-changing accounting product that was supposed to make Workday financials irreplaceable. Rumor had it that ServiceNow was also racing to market, and whoever figured this out was going to be the household name for large finance teams globally. However, it is now 2023, and no one has actually solved this problem and automated their accounting teams out of a job. Financial data is too important, and there is an art to accounting that cannot be easily replaced by machines. It will take time and good will to earn enough customer trust for this kind of end-to-end automation. In the end, both of these companies decided to go back to walking, instead of sprinting.
Someone will figure it outâŠit might just take a whileÂ
Some problems may be too hard to automate with 99.9% or 100% accuracyâŠand some problems cannot sacrifice the .1% error rate. Can you imagine the potential impact of a .1% error rate on Googleâs bookkeeping? That could have amounted to a $280 Million error to close out 2022.
The worldâs most ambitious entrepreneurs want to solve meaty problems that scare the majority of us away. I canât begin to emphasize my admiration for those people. But, itâs important to remember that sometimes, we are better suited to being a facilitator rather than a decision-maker and take incremental steps towards a vision. Who knows, eventually you may gain enough trust as the facilitator (and source of training data) to own the process end-to-end. That journey is one we want to bet on.