As we enter 2018, it’s clear the expectations of intelligent assistance providers and enterprise practitioners are more in balance than ever before. It has also become evident that the number of “known knowns” is growing rapidly.

Here are five examples of how we will learn more about the complex workings of conversational commerce  in the coming year.

1. Intelligent assistants (IAs) can do a lot right out of the box

Solution providers have access to a history of FAQs, recordings, call detail records (CDRs), chat transcripts, product literature, CRM records, and even interactive voice response (IVR) scripts to inform IAs on the basic intents of customers in the top vertical categories. Even if you are not Google, Facebook, or another one of the tech giants that benefit from exposure to inconceivably large data sets, it’s possible to access quality insights about your consumers. Big data is no longer required to get started in conversational commerce.

2. Humans still have a place in training and tuning

IAs can result in nearly immediate reduction in operating costs and enhanced customer satisfaction, but their ability to replace live agents and assistants is vastly overrated. On the contrary, the need to monitor and constantly tune responses to keep them relevant for prospects and satisfactory for customers is growing rapidly. Meeting this need will prove a source of new of jobs and reliable employment for the foreseeable future.

3. Managing true conversations comes next

Categorization? √ Understanding? √ Conversation? Nope! Some solution providers are better than others at recognizing context, promoting turn-taking, and allowing customers’ minds to wander and jump between topics, but the game board is still largely governed by statistics and branches on logic trees that can come up empty. When that happens, IAs are no better than poorly designed, frustrating IVR apps. Watch for some real advancements to correct this situation in 2018.

4. We will be able to hurdle the walled gardens of #ConvComm

It’s a shame that the branded Big Four (and other heavy hitters in the tech industry) have elected to create separate and unequal “platforms” for messaging bots and voice first agents. Companies would like to see intelligent assistance infrastructure that lets them develop a service once and have it render correctly on all devices and in all modes. Instead, in the short to medium term, we are stuck with the same duplicated efforts that plague mobile apps (iOS versus Android, for instance). Separate staffs maintain the code base for Alexa skills, Google actions, Facebook Messenger bots, and a few others.

5. Our reach will exceed our grasp

Now that we’ve tackled the “known knowns,” we can expect startups and stealthy entrepreneurs to emerge in 2018 ready to tackle “known unknowns.” They will do this by blending machine learning, natural language processing, knowledge management, and other elements of AI with the human touch. Some might even expose and exploit the “unknown knowns.” Their nature is to be determined, but there is always room for breakthrough technologies. At the very minimum, solutions will span the traditional lines of demarcation between customer care, marketing, user ID, and authentication and relationship management platforms.

Adoption and use of conversational commerce will accelerate in 2018. Forces are aligning for people to employ conversational technologies at home, in their cars, and in their offices. It’s up to solution providers to manage the complexity of underlying processes to keep things simple for end users and enterprises alike.

This article originally appeared on the Opus Research blog. Copyright 2017.

Dan Miller is founder and lead analyst at Opus Research, a market research firm focused on conversational commerce.