Elevating Roles in the Face of Automation

A firm implements software that auto-fills tax returns from scanned client documents. The staff accountant simply opens the tax software, manually inputs a few deductions that the software didn’t pick up on, and forwards the task on to the reviewer. The reviewer finds the return riddled with errors, numbers that were misread by the software or perhaps missed entirely. She sends it back to the preparer with a list of review notes. Both are frustrated that the technology didn’t work.

Was this really a failure of technology? Or a failure of people?

In an article for Harvard Business Review, Andrew Ng, a founding lead of the Google Brain team and former director of the Stanford Artificial Intelligence Laboratory, explained what AI can and can’t do.

“Almost all of AI’s recent progress is through one type, in which some input data (A) is used to quickly generate some simple response (B).”

In the context of tax return preparation, the software recognizes a 1099-INT and responds by auto-filling the amounts from Box 1 onto Schedule B. But anyone who prepares tax returns knows that very few returns, as least those prepared by a CPA firm, are as simple as AàB.

In the real world, a preparer interprets information provided by the client, applies tax law and makes decisions that AI simply isn’t capable of handling yet. The problem in our example above is not that the machine failed to prepare the return correctly, but that the staff accountant wasn’t aware that his role had been elevated. He’s no longer simply a tax preparer. He’s now the first reviewer, but nobody in the firm told him this was his new role or trained him to review a return.

Consider another scenario we often see in our Lean process improvement work. A firm is considering implementing software that will assemble and deliver tax returns and obtain signed e-file authorizations. This role has traditionally been handled by a member of the Admin team. She’s been with the firm for 20 years, and this is her only responsibility. Firm leaders know that technology could significantly improve the process, but fear it would make this team member obsolete. So they resist change.

Typically, this fear is not expressly stated but communicated through body language and hesitation. However, the leaders of this firm are not protecting their team member so much as they are holding her back from elevating her role. In these cases, we ask three questions:

1. How much time would be freed up from automating this role?
2. What is this team member’s skill set?
3. What gaps exist now that could use these skills?

One firm we worked with realized that this person’s skills could be employed in setting up meetings to deliver business returns to clients. The resolution will vary by firm, but in most situations we’ve encountered, automating a routine task didn’t make a team member redundant. It gave them an opportunity to elevate their role and deploy their skill in a way that leveraged their thinking capabilities – something that AI is not able to duplicate.

Writing for Harvard Business Review, Julia Kirby and Thomas H. Davenport wrote:

“When we talk about how smart machines should be deployed in workplaces, we constantly emphasize the importance of augmentation rather than automation. Employers, we insist, should implement cognitive computing solutions not so that they can make due with fewer people, but to enable their people to take on bigger challenges and have greater impact than they did before.”

As we ready our firms for the future, don’t neglect to think about the roles people play today and how we can align people with technology. Your people shouldn’t be left wondering what they are supposed to do. Figuring out how your people can best work with machines is better than resisting change or not addressing it at all. Instead of letting automation happen to them, train them for the next level of thinking.