According to a study by Oracle, almost 80 percent of businesses have already implemented or are planning to adopt AI as a customer service solution by 2020, and a recent report by Deloitte and Oxford University suggests telesales could be the next to go.

Other experts are ambivalent about whether AI is really advanced enough to take over the role of a talented sales representative. AI is already being used in the sales field to respond to basic email or chat inquiries, organize sales interactions, and follow up with leads. However, recent studies show that even the most advanced uses of AI bots still struggle when faced with complex user queries, and experts argue that if rolled out too early, bots may frustrate users and create more problems than solutions.

Here’s why it’s too early for AI to take over the role of sales representative just yet.

1. People can read between the lines

As motivation for his controversial robot tax, Microsoft cofounder Bill Gates proposed taxing automated enterprises partly to provide more funding for “human” roles such as caretakers, teachers, or nurses, who require compassion, tact, and sensitivity to interact with those in their care.

And while sales might not be as compassionate a role as nursing, as the famous business mantra goes, people do business with people they know, trust, and like. While AI has been successfully been rolled out for a number of customer service and sales tasks, the technology still cannot truly develop relationships with clients in the same ways human sales reps do.

Great sales reps have emotional intelligence that is nearly impossible for AI to replicate right now. Being a sales A-player means knowing how to interact with customers, listening and asking the right questions to show the client that this product or service could make their lives easier. Thanks to advances in machine learning, the technology has moved past programmed responses. But in the fast-paced world of sales, sometimes what isn’t said is just as important as what is said.

Experienced sales reps need to read between the lines to understand if a sales lead has any reservations they aren’t making readily apparent on the call. Sometimes a potential client will seem extremely interested on first appearances, but then “go dark” post-call. This is especially important when dealing with potential clients from different cultures. In China, for example, it is seen as improper to refuse an offer or say no immediately, with Chinese business people preferring “indirect rejections.”

Seasoned sales reps detect hesitation and reservation, even when a client may be nodding their head, and then offer up the right information to get them back on their side. However, AI in its current form would most likely be tricked by this type of human behavior.

2. Human sales teams are more accurate and agile

In a recent Gartner survey, customers say interactions with salespeople are the most important factor in influencing their buying decisions. Sales reps have a limited amount of time to woo a potential client, and first impressions count for everything. Clients want to know the person on the other end of the line is listening to them, and that their time is being well spent. There is nothing more frustrating than being asked the same question twice or being forced to repeat yourself due to a misunderstanding, and these small slip-ups could lose a potential client for good.

Microsoft reported 93.9 percent accuracy for conversational speech recognition on an assigned topic, which is on par with human capabilities. However, it would be hard to reproduce this level of accuracy in a real-world sales environment. Microsoft’s impressive results were based on offline speech recognition from recorded transcripts, which are not as challenging as online, real-time conversations, which can shift direction and be affected by external influences from one second to the next.

Sales calls are predominantly made via phone, conference call, or communication tools like Skype, and reps need to deal with background noise and overlapping speech when speaking with more than one person, which increase the risk of error for AI sales tools.

3. Human sales teams can look at the bigger picture

Especially when dealing with large B2B contracts, sales conversations can involve multiple calls, emails, and pitches to various people due to the rigid hierarchies of large corporations. As such, it is essential that sales teams piece together all of the information from various interactions to provide a fluid conversation that is moving in the right direction: toward making the big sale.

A salesperson may need to speak to stakeholders in different roles and responsibilities at a company, and needs to understand their distinct needs and how their product or service can solve their unique problems. One missing piece of the puzzle could blow a deal out of the water.

During complex sales, representatives need to stay on top of multiple conversations, and be able to refer back to comments made on other calls which may be relevant in the current call. As it stands, there is no evidence to show that AI has the capacity to fully piece together separate multi-threaded conversations in a complex sale to get a whole overview of the solution a company needs, and sell the right parts of a product or service to the right people.

However, while it appears that AI is simply not advanced enough to entirely take over the role of human sales representatives just yet, there are plenty of ways the technology can improve efficiency and productivity for human teams. For example, by using big data and machine learning technology, AI could help human sales reps by triangulating data from sales calls to offer real-time prompts for the best wording for questions, to obtain the best results.

AI could also be used to train new sales reps who don’t have complete product knowledge by offering up the right information to deal with detail-orientated questions and by linking a product’s benefits to a client’s pain points. To date, AI trumps humans in its ability to present exact data, figures, and financials in response to a question, which a human salesperson may not have at their fingertips or remember immediately.

Like yin and yang, humans and AI both bring something to the table. While AI can provide real-time information on technical aspects of the product so that sales reps don’t get tripped up, sales teams can build rapport and focus on the emotional aspects driving the buyer to buy or not buy.

It seems that at least for the time being, rather than replacing human sales teams, the power of AI should be harnessed to empower them.

Sabrina Atienza is the cofounder and CEO of Qurious, the real-time voice AI platform.