Text-Based Assistants: A New Frontier for Digital Shopping?

Text-Based Assistants: A New Frontier for Digital Shopping?

Who would you turn to if you wanted to buy an live tiger online and have it delivered to your door?

Not a toy “Tigger” or a stuffed animal that looks like a tiger, but a real, alive, breathing tiger.

Probably Google, right?

I would. It’s the logical choice, really.

I’m also sure that it would take me hours upon hours of research just to find out if there are any sellers. Plus, another couple of hours at least to find out out if it’s even legal.

Another option is to use one the emerging SMS Text-based assistants that promise to deliver everything. Just send them a text starting with “Hi, I would like a tiger delivered.”

That’s what TechCrunch did. They asked a tool called Magic to help them find a tiger. After all, Magic promises to “get whatever you want on-demand with no hassle.” They didn’t succeed.

It wasn’t for a lack of trying though. What stopped them in the end was the fact that it’s illegal to sell wild animals in the State of California. To be fair to Magic, the company was able to find a “source” who would rent a tiger for $13,400 plus delivery charges.

Text based assistant used to try and get a real life tiger delivered - TechCrunch

Image via TechCrunch

With companies like Magic gaining prominence in the past year or so, we’ll soon discover how these “on-demand, deliver everything companies” work, how the technology that makes the “magic” happen (pun intended) and how, if at all, this can be used by ecommerce companies.

What Are Text-based Assistants?

The basic idea of text-based assistants is that whenever you want something, like to buy a Tiger like TechCrunch did, all you need to do is send an SMS message and it will be resolved.

No need to download any new apps or waste time on research. All these services functionally act like an app, without actually being one – an invisible app if you will.

It all started with companies offering to find and order stuff for you, according to the instructions that you sent them.

Want to send your girlfriend flowers? Done.

Want someone to book a table at your favourite restaurant? Done.

Having a dinner party and don’t have time to go shopping? Have someone else go shopping for you.

Getting flowers delivered using a text based assistant

Image via Magic

On its own, this is already looking pretty sweet. Now what would happen if you added Siri and Google Now-like capabilities to the mix?

You would get Cloe, a text-based assistant that not only helps out with everything but also knows a lot about you, and so is able to offer comments and suggestions based on that information.

For example, suggesting that after you’re done with your formal dinner, why don’t you go to the local sports bar to catch the latest hockey match?

Cloe, an example of contextual assistant that knows deeply about you and suggests activities

Image via Product Hunt

While Magic and Cloe are examples of “do everything” assistants, there are a variety of text-based assistants that are more specialised in specific tasks/fields.

Riley, for example, will help find you an apartment. Unload will help you get rid of anything by offering a fair price for whatever you’re selling.

And then we have Kit.

Kit is your automated marketing manager that can run and manage Facebook ads, send email marketing to lists and report on the previous night’s sales among other things.

It works in any country, as long as you’re using ecommerce software that Kit understands. Shopify is among them, naturally.

It’s not only startups that are in this space. In August of last year,
Facebook announced its own personal assistant that lives inside the messenger app and is simply called Facebook M.

At first glance, text based assistants appear to be on the fringes, but upon closer inspection, could actually signal the beginning of the “Post-App” era.  

With Gartner reporting that app adoption is mellowing, with users downloading fewer new apps, and there being research that suggests more than half of internet users are experiencing content burnout,  it’s not a stretch to think that text based assistants who deliver what you want only when you ask for it wouldn’t be a breath of fresh air for casual internet users and business owners alike.  

How Text-based Assistants Work

Just because an app is “invisible” doesn’t of course mean that there isn’t work going on in the background.

To the user, the experience is that I ask a question over a text message and, after a while, I get an answer.

On the backend, receiving and answering a question is actually a three part process. You need a messaging platform, a human or AI interpreter and a way to fulfill the task.

Graph representing the 3 pillars of text-based assistants - messaging platforms, AI or human interpreters and fulfilment

Image via Sean Everett

A Messaging Platform

The first step is to receive the message that is about to be interpreted. For that, we need a messaging platform. By far, the most popular app for that so far has been your phone's built-in SMS capabilities. Then again, most of these apps still only work in the U.S.. So, international coverage has been a problem.

As a solution, some of the available assistants have turned to popular messaging apps to get coverage around the world.

Cloe, for example, uses Telegram in markets where it hasn’t been able to secure SMS-ready telephone numbers; Facebook M uses, well, Facebook Messenger.

The Interpreter

Graph about how Facebook M "reads" and understands questions asked.

Image via Sean Everett

The next step is for the assistant app to understand what is being asked and find the answer.

The image above shows how Facebook M gets asked a question, what information it gathers from it and what its answer looks like. The interpreter part can either be powered by AI (artificial intelligence) with machine learning algorithms, a human, or a combination of both.

For now, most companies use a system whereby AI does most of the work. But for validation, there’s always a human that makes sure that everything is alright. As the technology gets better, and AI powered by machine learning learns more and more from the work that humans are doing, there will come a time when all of this will be fully automatic, without any need for human interaction. But that time is still a little ways off.


The last part of the equation is fulfillment, or in other words, making sure that whatever the customer was promised, actually gets delivered.

With requests like booking flights and restaurants, the solution is simple. The app simply buys the tickets and calls the restaurant. When requests involve buying physical objects and getting them delivered, things get a bit more complicated.

Today, none of the assistant apps have a delivery network of their own and they all piggyback off other on-demand services like PostMate, Task Rabbit, Uber, Instacart and others to go buy and deliver stuff to customers.

So far so good, right?  People want stuff done, they text an assistant and it gets magically done. What can possibly be wrong with that?

Connecting the Dots

The reason why we’re bringing this up, is again, these assistants are a signal that we’re moving into the “Post-App” era.

To remain competitive with larger retailers, and more importantly, deliver more personal experiences with your customers, building your own text-based assistant may be a way to differentiate yourself in the market.

Imagine this scenario: You wake up and remember that you have this big fundraiser event in the evening. You need fancy clothes but you’ve just moved to a new city, most of your stuff hasn’t arrived yet, and you don’t have time to go shopping.

Panic begins to set in.

Then, you remember your favorite online store released a text-based assistant that sends personalized recommendations based on your purchase history., After a couple of questions, you’ve selected your outfit for the evening, and have been notified it will be delivered by UberRush to your office in just a few hours.

You want to know the crazy part? Most of what I just described is possible with today's technology. The future is now. We don’t have to wait anymore.

The text-based assistant technology that understands what I’m looking for and can ask clarifying questions? We have that today.

Ability to send and share product images and links from any messaging platform? Shopkey takes care of that.

Inputting your measurements and shoe size? Nothing hard about that.

Analyze the fabrics used and make sure that whatever I will choose will actually fit me well while looking good? Yep, that’s possible today and something that Fits.me has been working on for years.

Speedy delivery? Done. Services like UberRUSH and equivalents will deliver orders within hours of purchase and with the coming age of drone deliveries, things will only get better for delivery speed.


What seemed like science fiction only a few years ago will be the norm way sooner than you think. While a service like the highly personalized fashion assistant described above, that can guarantee well-fitting clothes with same-day delivery might seem like a pipe dream, the reality is that most, if not all, the pieces of the puzzle are here today to make it a reality.

While you may have thought that I was crazy when reading the first part of the fashion assistant story, after reading how the technology is available today, do you still think so?

If not, then maybe it’s time for you to start building these apps to your advantage.



About The Author

Ott Niggulis is a chef/paramedic/freelance writer who focuses on marketing and CRO. Marketing is a numbers game and he loves numbers. Follow him on Twitter.