The Process Behind Slash

This project constructs an overview of what the ideal messaging app of the next-gen should behave like.

The Final Product →


My aim for this case study was to gain a better understanding of the inner workings of present-day messaging apps and check for avenues to improve this model. During the course of my study, I dove deeper into concepts like artificial intelligence, usability heuristics, utilitarianism, behavioral patterns of people in social groups (large and small), and so forth. I formulated two hypotheses based on my study and I further explored these ideas empirically in order to find supporting data. The end goal of the case study was to build an actual product based on the culmination of my findings. 


• By building a messaging app around the core idea of connecting large groups of people more intimately, we’d achieve greater success at delivering a product that truly enables people to be social. 

• Hybrid interfaces are the way forward for conversational UI. Chatbots specifically have to adapt to the chaotic structure of conversations humans follow as opposed to machines. This is also a design problem as much as it is a linguistic and technological problem. 


The Process - TL;DR

For this case study, I used Google's framework that supports divergent and convergent thinking based on the IDEO model in attaining intuitive design solutions to end-user problems. This process provided me with an efficient pipeline to build my solutions around user findings.

Research & Analysis

Social media and messaging apps as an industry seem to be pivoting towards instant gratification being the key underlying motive for every app. Recent studies have found that despite being more connected than ever, more people feel more alone than ever. I wanted to understand the root cause for this. Turns out that most social apps today, by design, are meant to connect people who already know each other personally in an online space. While this is surely helpful, it gives rise to a few problems. 

“Only 7 percent of Facebook friends are people a user knows only in an online context.” - A Pew Study

The warning that pops up if you add people you may not know.

Most social apps involuntarily tend to create feelings of loneliness and envy. They also promote broadcasting our lives in the form of highlights on a day-to-day basis. This moves away emphasis from the primary need for all of us, a feeling of belongingness and intimate conversations. “It doesn’t mean Facebook causes depression, but that depressed feelings and lots of time on Facebook and comparing oneself to others tend to go hand in hand,” said Steers, in a study researching the links between Facebook and depression symptoms. 
There is a need, more than ever to create a social network built on the premise of connecting people based on interests. A platform where we don’t collect friends as stamps and value meaningful conversations over momentary satisfaction. 

Taking a step back from this, I also wanted to examine the role of social apps in our broad behavior over the internet. How do we use the internet and how do they play a role in it?

Over the years, we’ve seen the capabilities of social apps broaden. What was once tools meant for socializing are today so much more. We use social apps for everything ranging from buying products and playing games to keeping up with the news and discovering new interests. 

With the advent of messaging apps, we’re noticing a shift in the number of active users on traditional social networks. Messaging apps now have more monthly active users and this figure is steadily growing. For businesses, it’s a matter of staying where your consumers are and this influx has caused brands to rethink how they talk to their audience. 

Bots are a great example of how brands have made their way into chat apps. Although their implementation in the current form has caused a fair amount backlash, the bot industry is growing faster than ever with its predictable hiccups. Most bots today follow a flow based model, where the flow of the conversation is predefined by the developer and it has a rigid structure that doesn’t support breaking away from the script as this requires the bot to have some level of artificial intelligence. Flow based bots are as good as the instructions set for them. Sadly, not all businesses are good at programming their bots, which can lead to immense frustration among users and give rise to broken product experiences.

Businesses can’t be blamed for their incompetence at building complex conversational flows as it’s naive to expect all brands to turn into conversation design specialists overnight. This change for all we know may never come. A model like this puts the onus on the businesses towards delivering a chatbot that is functional. Thus it creates an unfair platform where only brands with the knowledge of creating such extensive flows might achieve better success. Nevertheless, the idea of chatbots and the benefits it brings with it is extremely exciting. 


1. Consolidation: Our usage of the internet is largely fragmented and an individual launches an average of 9 apps a day and up to 30 a month, to either stay connected with people or consume a service. Is it possible to collate many of these experiences within a single app?  

2. Support for large communities: Is it possible to create a messaging platform that supports connecting large scale communities intimately?

3. Level playing field for businesses: Bots are the new websites. Is it possible to do with bots what we did with websites? Make it truly simple for everyone to own one.

4. Stability: Is it possible to create these advanced experiences for users in a way that they consistently deliver positive experiences for users? 

5. Familiarity: Is it possible to craft these experiences in a way that they still remain familiar and do not introduce a steep learning curve for the users?

Exploring the problem space

In order to understand the scale and variety of the problems that exist in the space, it was necessary to define a few key things - 

1. User Types, Expectations & Pain Points
2. Business Types, Expectations & Pain Points


Key points on Users

- Users want fast ways to stay in touch with people they know and also meet new people. 
- Users want to use messaging to interact with businesses.
- Users are frustrated with broken bots and want them to be more reliable. 
- They aren’t comfortable with the idea of brands invading their privacy inside messaging apps. 
- They do want AI to solve a lot of their problems. 
- Users want features that’ll enable them to share their lives and thoughts better. 


Key points on Businesses

- They believe that automation will improve the ease of the customer.
- For many businesses, messaging apps are a great point of contact to potential customers.
- They are careful not to cause any fear of a breach in privacy and security.
- They want an easy-to-use platform to dive into this market. 
- They believe that chatbots are the future and 80% of companies would like to integrate them by 2020. 



I started penning down and sketching as many abstract ideas as possible. Ideas that could add up to a solution that would satisfy both the groups - users and businesses. Some of them being: What would an interface like this look like? How should chatting within large groups and bots work? How would people talk to bots? How would people call bots into conversations? Based on the pre-defined goals, I circled in on some interesting ideas  

Groups and Threads

Carrying forward on the goal to create a messaging platform that supports connecting large communities more intimately, I dug deeper into finding solutions for it. Present messaging apps do not have the capacity to conduct focused conversations for a group of say 10,000+ people. A group like that would consist of multiple clusters of people with varied intents in a conversation. For example, in a group for an entire city, cluster A might be talking about cats while cluster B might be talking about meeting up for drinks, and so forth. In a group with many such clusters with parallel conversations, there is little scope for meaningful conversations to occur without being interrupted.


Threads seemed to be the way forward. The thought behind threads was to create parallel spaces for micro-conversations to happen while still being a part of the core group. Similar to a large party where people form clusters with each other to have conversations based on common interests but are still a part of the large party. The switching between threads and groups had to be seamless.

How is this any different from smaller groups? Having larger groups have their benefits. You have a wider reach and pool of individuals to speak to. It enables better discoverability of like-minded individuals and new threads. This is also great for businesses and consumers since people naturally segregate themselves into interest based groups and hence results in better targeting for brands and puts less stress on tracking users. For example in a group for a neighborhood, it’s possible to instantly reach out to food lovers or people who own pets. 

Bots with hybrid interfaces

With the options available for creating flow based bots, AI-based bots and hybrid bots, I chose to go with hybrid as a direction for Slash. Hybrid bots are a bit of both worlds. They do have the power of artificial intelligence and can be interacted with not just by voice or text but also with touch. This was a decision made solely based on the goal to make bots on slash extremely reliable and intelligent. Bots on Slash had to not just be capable of comprehending natural language but also act as fully functional apps. This would mean that it had to have the reliability of an app. Where users could simply launch a bot and not have to deal with the stress of their intent not being understood or being forced to talk to a bot for simple tasks.


This was a huge challenge as it meant that I had to create a framework that worked with almost every brand across the globe and had to emulate the experience of their entire app within a chatbot interface. I sketched and wireframed multiple variations of hybrid interfaces that would support such a model. Eventually, I achieved the best success by actually designing and rapidly prototyping bots for real-life brands. Also, I made sure to use real content and imagery from the brands in order to create an experience that would be more realistic to implement.

The simplicity of ‘Slash’ 

It became clear from the start that there had to be a common logic that governed how bots, groups, threads, and users within Slash could be searchable within chat screens. Apart from just being able to search for them by names, they had to have specific IDs. 

Users, Bots & Groups - /JonBatiste/Starbucks & /Kittens
Threads - /Dribbble/AmsterdamMeetup


The structure of IDs on Slash is quite similar to URLs. ‘/’ proved to be the most intuitive solution as this brought in the affordance of paths. Instead of trying to have the user remember multiple syntaxes inside of an app such as @, / and +, that all do different things, on Slash it’s all just minimised to one. 

A simplified framework to build AI-based bots

Building bots even with simplified flow based tools often include having an extensive knowledge of communication design and user psyche. While this may work for large companies with the capability of hiring the best minds to build their bot, it doesn’t work with small-scale companies and brands that do not have the capital or expertise to build such a bot. The bots are often confined to instructions laid out for them and are rather primitive in nature. 

In order to build a truly intelligent bot, we need to master the art of conversation. It’s difficult and nearly impossible for all businesses to suddenly acquaint themselves with such a platform. Which is why it’s important to build an AI framework with a wide range of capabilities that support almost every type of business. Consider the job of a waiter — what makes the waiter from a Chinese restaurant different from that of an Italian restaurant? It’s the knowledge and skills pertaining to the specific restaurants. On the contrary, what are the commonalities between the two? It’s traits like empathy towards customers, communication skills, knowledge of how to serve, and so forth. Our job is to emulate these commonalities. 


Imagine a tool that asked you for the type of business you owned and then only required you to personalize the bot according to your brand. So, if you’re a restaurant owner on Slash, you’d never have to draft complex flows of how your bot ought to respond to your customer. You simply would have to upload information about the food you serve, your timings, neighborhoods you deliver to, etc. The bot would learn from this newly added information and figure out how to respond to customers by itself. You can even edit the personality of your bot on granular levels to suit your brand. 

Prototyping, Iterating & Validating

Apart from following the conventional method of rapidly prototyping user flows between screens and navigating through the app, for Slash, it was important for me to simulate real life conversations and test how my ideas were holding up. Some of the methods I used were to simulate real life conversations from groups and forums, personal interactions between pairs and conversations with existing bots, within Slash. I built detailed chat scripts which I translated into prototypes to play out the whole conversation. The goal was to make it as real as it can get and to understand how people truly communicate. 

Some of my findings include people actually sharing a personal connection with most bots. Many individuals greeted their bots with a “hello :)” or sometimes saying things like “I love you. Thanks!”. It’s surprising how little bots are equipped to respond to these messages and often reciprocate with a robotic “I do not understand” or remind you that their abilities are only confined to executing a narrow set of tasks.

Prototyping these included studying individuals with specific intents in mind and how they would potentially interact with bots or other people. For example, for a person buying coffee beans from a local store and have it delivered to her/his home. How would they search, locate, customise and pay for their order? How would they be able to track it once the order is placed? What are the different states that could exist through the entire transaction? etc. 

All work copyright Sreekar Pradyumna © 2017.