Culture problems? This AI tool can root them out
When Mike Durland watched Sen. Elizabeth Warren grill then-Wells Fargo CEO John Stumpf about the bank’s phony-accounts scandal at a 2016 congressional hearing, he realized executives need a better way to understand the cultures of their organizations.
“It's hard to manage and it's a very uncomfortable position to be held accountable for something if you can't measure it, observe it, or can't track it,” said Durland, who is CEO of Melancthon Capital, a professor at the University of Toronto and the former group head and CEO of Global Banking and Markets at Scotiabank.
Durland took on his executive role at Scotia Bank in 2008 in the midst of the financial crisis, when culture and organizational flaws at many banks were exposed.
“We as executives are accountable for ensuring that all of the risks of our enterprise are actively and appropriately managed and our stakeholders hold us accountable for that,” he said. “We have to have the tools to do that.”
Durland is on the board of Receptiviti, a fintech that has developed an artificial intelligence engine that evaluates and monitors workplace stress, identifying potential culture and organizational problems, or high-performing workgroups whose practices could be spread throughout a company.
“What we're really trying to understand is, what's going on inside of our organization?” he said. “Where are people excited about what they're doing, where are people collaborating and cooperating, but also, what are the things inside of an organization maybe that are causing resentment or stress or anxiety? Are there areas where silos are being formed? Are there pockets of the organization that need to communicate with one another effectively that aren't, are there these hidden pockets of the organization?”
To be sure, Stumpf publicly touted the bank’s “eight is great” goal — each customer should have eight bank products. He understood that branch employees were under pressure to cross-sell.
“But did he really, fully understand the implications, the effect that this was having?” Durland said. “And did he have a sense of the harm that he was creating because of the way this set of incentives was designed? That's less clear to me and that shade of gray is really important at the end of the day."
Getting insight into potential problems is all part of managing people and managing risk, Durland says.
“Even the healthiest organizations are in harm's way,” he said. “It's nonrecoverable in today's day and age to have something occur on your watch, in your culture.”
How Receptiviti works
Many trading floors and exchanges already use AI software designed to detect signs of collusion, looking for patterns of communications and suspicious behavior such as gloating. Digital Reasoning and IBM Watson are two providers of this kind of technology, which focuses primarily on keywords.
But traders can sometimes figure out what these systems are looking for and therefore what not to say.
Receptiviti is based on the work of James Pennebaker, a professor at the University of Texas at Austin. It doesn't look at keywords, but instead at prepositions and pronouns, to find signals in the language employees use in their emails and instant messages, for instance in Slack and Office 365. For instance, people who are highly depressed tend to use first-person-singular pronouns like “I” more than people who aren't depressed, according to Jonathan Kreindler, Receptiviti's CEO. And they tend to focus on the present.
“Pennebaker's revelation was that it's these words that are not really controlled by us in any conscious way, that that can provide more meaning,” Durland said.
Individuals are not identified or tracked, but branches, workgroups and other segments can be, as well as people of different genders and races within those groups. The software can detect impulsivity and anxiety, Kreindler said. It could see how the stress levels of traders in a Chicago office compare to those in a New York office, for instance.
“You could figure out if they are suffering turmoil in their portfolios or maybe they're up to some kind of collusion,” Kreindler said. “The idea is to try and understand what is the psychological state? Are they manifesting unhealthy tendencies, are they satisfied, are they engaged, are they empowered, are they independent?”
A financial institution that was suffering from the repercussions of questionable sales practices could have a continuous analysis running of all customer-facing inquiries across all branches, for instance, and evaluate trending stress levels.
“You could understand which branches are most likely to engage in the sort of sales tactic that would result from people who were under duress or pressure,” Kreindler said.
An organization might monitor employees’ emotional states to see how they change with the rollout of a new incentive plan, a new strategy, a town hall, a merger or the end of a quarter.
Mastercard may use it
Mastercard is considering using Receptiviti’s technology, according to Walt Macnee, vice chairman of the card network.
He sees a growing need to monitor workplace culture and stress levels.
“The workplace has always been a demanding environment; I feel that somehow it's becoming more so,” Macnee said. “Companies are starting to move beyond traditional engagement surveys, which are static, and often annual.”
Employees often don’t tell the truth in companies surveys for fear of retribution.
“And there's all sorts of human behaviors tied up with that,” Macnee notes. “There can be people who are too mild, there are others who are far too harsh. You're going through all sorts of filters to get out what someone really thinks.”
“There's an interest now in having a more up to date, real-time assessment of the culture of an organization,” he said.
He pointed to the Weinstein Co. and the International Olympic Committee as examples of organizations harmed by the bad behavior of people within.
“Software like Receptiviti’s can show you things in real time and over time,” he said.
Macnee sees the software being useful in the case of a merger, to monitor how it’s going and identify issues or find things within an acquired company that ought to be emulated throughout the organization.
A skeptic’s view
Seth Grimes, president and principal consultant at Alta Plana, has been following this category of software for more than a decade and is skeptical of all of it.
“There's so much money washing around for startups and there are not really high barriers to entry for machine learning and the platforms in particular,” he said.
Some of these programs are solutions looking for problems, Grimes said.
“To try to apply AI to workplace stress — there’s this big market of stressed-out workers and we're going to find a tech solution for it — I’m skeptical,” Grimes said.
And even if you can detect that stress levels in a department are up, how do you know why?
“There's that expression of analysis that just creates more data that doesn't actually do anything for you,” he said.
But Kreindler said companies sometimes bring in a psychologist or analyst to look into the problems his software identifies. Australian regulators, he noted, have already appointed psychologists to sit in on bank boards to identify culture issues. The organizations that are adopting this software, Kreindler said, are forward-thinking.
“They're looking at what's happened to their competitors and they're looking to show the regulators that they are taking a proactive approach to risk as opposed to reactive one,” he said.
“We did a holistic analysis for them,” Kreindler said. “We were able to see that some of the more senior people were risk averse. While that's not bad, it can work against you when you're trying to build an innovative culture, if people are being told no when they bring new ideas to life.”
One financial institution was going through a merger and the Receptiviti software was able to identity a group in the merged organization that was struggling to work together and suffering stress, fatigue and burnout, Kreindler said.
“They were able to start to address those problems before people started to leave,” he said.