IN A RECENT survey of North American chief executives and chief financial officers, nearly 80% listed corporate culture as one of the five most important factors driving their company’s financial performance. A growing body of empirical evidence supports their belief that culture matters—and can boost profitability.
Yet, in the same survey, an even higher number of respondents—84%—said their company’s culture is not where it needs to be. Again the data supports their intuition. The average culture rating of large employers in America on Glassdoor, a website that lets workers rate their employers, is 3.6 out of 5. Few people would be excited to eat in a restaurant or ride with an Uber driver with that kind of rating. Similarly, few employees are likely to relish spending 40 hours each week in an average culture.
Building and maintaining a healthy corporate culture can be even more challenging in organisations where employees work remotely. In an ongoing study, we find that the companies where employees are most effusive about remote work score lower than their peer groups on corporate culture, especially on learning and development opportunities and honest communication.
Leaders cannot improve what they cannot measure. Unfortunately, the most common tool for gauging corporate culture—the engagement survey—suffers from serious limitations. Faced with a long list of questions, employees switch to auto-pilot and assign identical or similar scores to every question. Employers that ask dozens of multiple-choice questions, as many do, might glean only a couple of reliable insights because of respondents zoning out. Even when employees do engage with a question, their score offers little guidance on how to improve things. And what if the topic the employee really wanted to weigh in on wasn’t included?
Recent advances in AI—most notably large language models (LLMs)—allow leaders, for the first time, to glean nuanced insights into their corporate culture from how employees talk about their company in their own words. Rather than answering reams of questions on a five-point scale, workers can now simply explain what is and isn’t working in their organisation and offer suggestions for how it can improve. The AI can do the heavy lifting, providing much more granular classification of comments and assessment of sentiment.
Freed from the shackles of traditional surveys, organisations can use AI to gather and process employee feedback from many sources. The volume of available feedback is staggering. Combining free text from internal surveys, performance feedback provided to managers, online employer reviews and other sources equates to tens of thousands of pages of data each year for a large company. Until recently, organisations had to rely on crude tools such as word clouds or search keywords to gain insights from this trove of information.
Armed with the more numerous and granular measurements that AI brings, executives can more quickly and easily assess whether their company is living up to the values it considers “core”, identify the most important cultural elements driving everything from employee attrition to innovation, diagnose toxic subcultures within the organisation, and plot progress over time. After spotting important patterns, leaders can dive into the raw feedback for more nuanced context and employees’ recommendations on how to improve culture.
Take Amazon, which aspires to be the best employer in the world. We used our AI platform to analyse tens of thousands of employee reviews of the e-commerce giant. This showed that Amazon does well on many of its leadership principles, such as “customer obsession” and “invent and simplify”. But the firm’s culture also contributes to employee burnout, especially among software engineers, who are twice as likely to complain about burnout than warehouse workers or drivers. Raw employee feedback points to ways Amazon could reduce stress for engineers, like fixing a performance-review process widely viewed as brutal or minimising late-night disruptions when technical employees are on call.
Even the largest companies will take their time adopting AI. But cultural analysis is one of the few areas where it can be embraced right now, because it plays to one of AI’s biggest current strengths: understanding natural language at scale.
This does not mean that leaders should blindly trust the output of LLMs. The tools require safeguards to protect against foibles, such as hallucinating made-up answers. Models should measure the elements of culture based on solid evidence, rather than the latest management fad. Leaders need to take a broad view of culture, measuring not only the factors that influence employee satisfaction but also topics that shape a company’s ability to adapt to market shifts and to avoid unethical or illegal behaviour.
Leaders who do adopt AI for cultural insights can use these to make their employees happier, lower the odds of reputational disasters and, ultimately, boost their profits. Measurement is not the only piece of the “successful culture” puzzle, but it is a crucial one. Culture has always been an enigma at the heart of organisational performance: undoubtedly important, but inscrutable. With AI, meaningful progress can be made in deciphering it.
Don Sull is a professor at the MIT Sloan School of Management and a co-founder of CultureX, a research and AI firm. Charlie Sull is a co-founder of CultureX.