The bot land grab is officially under way and everyone is rushing to recreate successful app ideas in chat interfaces. Somehow the old pitch by analogy — we’re “AirBnB for pets” — has found a way to be even less original. Now any aspiring entrepreneur need only take the name of a successful app and add “as a chatbot” at the end.

Of course, this is nothing new. The exact same thing happened at the start of the mobile revolution. The top tech minds were busy jamming interfaces designed for 18-inch monitors into iPhone screens and calling it good. It was only later that people realized, rather than just cramming successful website ideas onto mobile devices, the real opportunity was in leveraging the fact that these new screens were, in fact, mobile. A taxi-hailing website would have only been minimally useful. Sure, I could order a cab at the beginning of my night, but an app comes into play when I’m out and about. Uber was successful because it leveraged the unique features of a mobile platform.

Similarly, the winners from this next interface revolution will be the startups that leverage the fact that conversational interfaces are, well, conversational. Just like the shift from desktop to mobile didn’t mean everything was necessarily better on mobile, not everything that was great as an app will be great on conversational interfaces. The idea of a “Tinder you talk to” may seem fashionable but ignores the fact that people actually like using Tinder (although few will admit to it). Far from being some arduous chore, Tinder is a game, a delightful distraction from engaging with real-life strangers.

Sometimes visual interfaces are just an inherently better way of solving a user’s problem. A waiter could tell me all of the beer options, but if we were in Portland that monologue might drag on for the better part of an hour. What’s worse is that by the time he got to the end, I would have fully forgotten the first options. Sometimes it’s just easier to look at a beer menu. Furthermore, many of the most successful apps were successful because their interactions worked well on visual interfaces. Trying to recreate it in a voice interface may leave many well-intentioned startup teams ice-skating uphill.

This leads me to my obvious conclusion for what will be the most successful bots. The most successful new ideas will be the ones that leverage the unique feature of conversational interfaces — that is, conversation. The low hanging fruit that so many in the tech world seem content to ignore is in recreating real conversations that people are forced to have, and making them better with bots. In this way, the easy choice for disruption is not replacing apps but replacing people. If you’re looking for a place to start, that’s it.

I can appreciate that anyone who has used one of the early versions of chatbot might balk at the idea that a bot’s conversation could outperform a person. After all, even the densest of customer service representatives can generally get the gist of what you are talking about without falling into a “sorry I don’t understand” loop the moment your questions deviate even slightly from their training material.

It is worth remembering that these technologies are still very new. If you look back, it’s not like the apps from 2007 were blowing anyone away either. Also, thanks to the collective training power of neural nets, it is not hard to imagine that these systems will improve at a much faster rate than apps did.

Ignoring the current technical shortcomings, bots also offer many advantages over their current human competitors. Namely, they are hardwired to care, which is far more than I can say for the vast majority of customer service conversations I have with people. A bot doesn’t have a fight with its girlfriend and then vent its frustration on users. A bot isn’t distracted by the nagging feeling that it should have stayed in school to avoid the bleakness of its current reality, in the middle of me talking to it. Users have a bot’s full attention, free of all the bias, pettiness, and forgetfulness that plagues conversations with humans.

Furthermore, it is just easier to replace a conversation with a conversational interface. When tech teams were stuck with the unenviable task of converting from human-to-human interactions to human-to-screen interactions, there was a lot of guesswork that went into figuring out what would work. It was like needing to bike across town with a blindfold on — you had a general sense of what direction you needed to go, but the only way to progress was by hitting a wall.

That’s not the case anymore. Recreating conversations is much easier, although it does involve an underused skill set — e.g., listening. Want to learn how the conversational interface for your paralegal bot should work? Try shadowing a paralegal. They have already spent years developing the perfect conversational interface, and you can just borrow it.

Perhaps the main reason the biggest opportunities in the bot market is in replacing people and not apps is simply a matter of economics. If an app is solving users’ problems, it is likely doing it for virtually nothing, whereas there are endless examples of humans that are being paid huge monthly sums for their half-baked efforts. Even a small increase in inefficiency can save businesses millions of dollars.

In a recent conversation with David Beisel of Nextview Ventures, I was trying to glean some insights on where the real opportunities are for conversational interface technologies. Even though he knows the voice computing startup market, he noted it was still too early to tell who will be the unicorns of bots. His savvy for investing revealed itself when he told me about one of the startups they recently invested in, called Troops.AI, which is a bot that improves the productivity of sales teams. He said, “It was an easy decision because it helps salespeople sell more and do it more easily, and that anything that makes businesses money is bound to do well.”

I would add any startup that saves businesses money is an obvious bet too. Virtually every large company has a dedicated human resources department, with each employee making on average $57,420 (BLS) in exchange for their core skillset of filtering hundreds of resumes to a couple dozen appropriate resumes. The repetitiveness of this job makes it feel like a website should have already replaced this job. However, using any of the websites that have tried to replace HR workers shows why companies still use HR employees. First you upload your resume, then you have to type out everything written in your resume into these obtusely organized text fields that are meant to serve every potential employee from CMO to janitor. Applying for a job online is enough to make a user think being unemployed and living under a bridge might not be such a bad alternative.

It is exponentially easier to apply directly through a conversation that can just ask about your experience as it pertains to the job you are applying for. This is exactly what the conversational AI startup Wade & Wendy created. Instead of sterile, life-draining online forms, their bots ask job seekers to elaborate on specific parts of their job experience to better filter applicants for any given job listing. Since this is the bulk of HR employees’ workload, this can save businesses upwards of 50 percent on their current HR expenses.

Ultimately, trying to make apps more conversational may seem like the obvious answer for entrepreneurs looking to cash in early on this new tech revolution, but the juice may not be worth the squeeze. First off the solution may just be better suited for a visual interface, which could explain why it became such a popular app. Additionally, the efficiency of apps mean they are probably solving any problem for virtually nothing, so saving users $.99 per download is hardly life changing. By comparison, automating half of an entire HR department’s workload is a very big deal for any Fortune 500 company.

Furthermore, you can build a bot that fits so naturally into people’s current workflow that users may not even realize they’re talking with a bot and not a person, much like X.ai was able to do with their call-scheduling bot Amy Ingram. Amy is so good at communicating like a helpful assistant that most people don’t even realize that she’s a bot working for $39 a month and not an executive assistant making $51,725 a year (according to payscale.com). The real winners in the conversational interface revolution will be the ones competing with highly paid employees, not freemium business models.