The financial world’s computer-loving crowd is preparing for the dawn of a new AI-powered era — but that doesn’t mean they’re ready to fully embrace the technology just yet.

In an Invesco survey of systematic investors with $22.5 trillion under management, 62% said artificial intelligence is going to be just as important as traditional analysis in a decade’s time, while 13% reckoned it will be even more significant.

Yet when asked about their own current use of the technology, respondents were ambivalent: Only 9% said they use it extensively, 38% said to a limited extent, and the rest said not for now.

The contrast paints a sobering picture of the pace of AI adoption on Wall Street as hype surrounding the technology hits fever pitch. Investors have turned to machines for tasks such as scanning the news for trading signals or dissecting market patterns, but they’ve largely held off from deploying them directly for actual allocation decisions.

“People don’t believe this is an easy thing,” said Bernhard Langer, chief investment officer of Invesco Quantitative Strategies. “Yes, AI is a huge toolbox. Big data is opening new horizons. But I have to be careful and understand what I’m doing.”

Even as they demur, the AI use case suggested by most respondents in the annual quant survey was identifying patterns and trends in markets — underscoring its immense potential to boost portfolio performance. Proponents say that because AI — or its data-driven offshoot, machine learning — is better at detecting complex relationships between multiple variables, it will prove more adaptive to changing markets.

The biggest perceived challenge of AI was the complexity and interpretability of the models, followed by the quality of available data, the poll showed.

Meanwhile, nearly all respondents said they invest based on factors — a traditional way of picking stocks based on the characteristics of a security. But the majority said they adjust their exposures to these trades as the market environment changes and expect to do so even more in coming years.

That might sound intuitive, but it’s still a divisive approach among the cohort given how hard it is to predict factor performance.

“We are living in difficult times,” Langer said, citing the political environment and rapid rate increases. “People are looking for ways to weather the storm and to be more dynamic is an answer — if this answer is successful.”

Here are the other highlights from the survey:

• 78% of respondents said they use dynamic asset allocation — changing how much they allocate to each asset class based on the market regime — in putting portfolios together
• 75% said they adjust factor weights over time, with 52% saying they expect to become more dynamic in the next two years
• 41% use natural language processing — where machines are trained to extract information from reams of text — for sentiment analysis
• Other than equities, where quants are most dominant, 73% said they are using a systematic approach in fixed income while 91% saw that asset class as a potential area of application
• 32% said they are adopting these methods in private equity, while 31% use them in real estate
• 63% saw value as the most attractive factor in the coming year, followed by quality at 62% and momentum at 42%

The survey, conducted in May and June, covered institutional and wholesale investors such as pensions, sovereign funds, wealth managers, model portfolio constructors and the like. It mostly excludes fund managers.