On bots 2/3: In search of a bot-native UX — New levels of personalization in enterprise software based on your job and preferences

My co-founder Alan and I were pretty skeptical about bots when we first started thinking about them last year. In this series of posts, we want to share what changed our mind and what we’ve learned exploring bot-native UX. In part 1, we argued that bots shouldn’t be scaled-down versions of existing products, mentioning two potential avenues for a bot-native UX: personalization and collaboration. This is part 2, exploring personalization in more detail.


Software with a consistent GUI will soon seem stale. In a few years, it will sound archaic for an executive to click lots of buttons in a cluttered dashboard to answer a question like: “How is my business doing?”. Ideally, the software should be able to figure out what she wants and present the answer in her preferred format right away (see the example below). In the last couple of months, we’ve found that bots have the potential to deliver this level of personalization in enterprise software, adapting to your job and preferences over time.

An example of a bot-native UX — personalization in enterprise software:

Two users ask the same question but get a personalized answer based on their different jobs and preferences. Alex (CEO) gets an Executive Summary while Mary (CMO) gets an overview of channel performance.

Before bots: Personalization for somebody means more clutter and friction for everyone else in a company

The problem with GUIs is that there is a trade-off: As you reduce friction for somebody by personalizing a GUI, you usually make it harder for everybody else to use.

As an example, take the common case of deployment of a business intelligence tool such as Tableau (dashboards) in a 100-person company. Tableau often gets introduced with the promise of “democratizing data” and making it accessible across all departments.

Yet, within the company’s departments people have different needs. A campaign manager for Facebook has totally different needs from someone who manages Google Adwords. A product manager wants to take a look at different KPIs. So the BI team creates custom dashboards for marketing, product, and sales. Very quickly there’s a huge backlog of dashboard requests (some people wait for months!) and an ever increasing number of dashboards (hundreds!). When the effort for finding the right dashboard becomes too high, many people stop using Tableau and go back to their own Excel sheet. With hundreds of dashboards floating around, maintenance costs get out of hand quickly. In other words, the costs of “democratizing data”, until today, were too high.

With bots: Each user experience is personalized without everyone else suffering

Personalization is native to a conversational user interface (CUI) as it “removes abstraction” from a GUI. The great advantage is that the interaction is always the same (you just write), and the user is only presented with a small piece at a time.

So if a developer wants to personalize a bot for a specific user, she can build this additional functionality without making it harder to use for everybody else. In the example above, a bot can deliver different experiences (“dashboards”) to every employee, without cluttering the interface for the other 99.

In our work with databot we’ve been exploring two ways to personalize enterprise software: (i) based on somebody’s job (e.g. a specific tool to manage Facebook campaigns) and (ii) based on somebody’s preferences (e.g. you like numbers more than graphs).

Level 1: Personalization based on your job — How bots can reduce friction to use general purpose enterprise software

Most enterprise software tools get used for more than one job. Bots will allow for a hyper-personalized experience built on top of general purpose tools and infrastructure. A user gets only what they need for their job with the lowest friction possible.

Let’s take a look at another dashboard, this time with a very smart underlying data warehouse: Tenjin is a great piece of software for streamlining mobile marketing and investing more effectively in campaigns. Here are some jobs people use it for, all of which happen in the same GUI:

  • Managing and optimising spend on Facebook campaigns (growth manager)
  • Analysing the impact of a new onboarding flow (product manager)
  • Getting a high level overview of the numbers (CEO)
Same interface on top of the same infrastructure for everybody.

The challenge for a GUI is that each persona — the growth manager, the product manager and the CEO — is interested in different data and has a different depth of understanding.

A bot can present a personalized view for a specific job without anyone learning a new interface. If a CEO only wants a high-level view of “How is my business doing?”, just asking a bot dramatically reduces the signal-to-noise ratio.

Personalized interface on top of the same infrastructure for everybody.

Level 2: Personalization based on your preferences — How bots will learn over time what you like

Everyone has their own preferences. Some people like numbers, other like visuals or wordy explanations. People have different routines, ways to organise their day and prior knowledge. These are very human things that most of our current software doesn’t take into account. We think that bots will get very good at delivering an output with a personal flavour based on our preferences and will be learning your habits over time. Here are some examples.

Example 1 — Visualisation: 
People have different preferences regarding the output format. Some people may prefer a graph, others a table. Most software offers a list of “export” options — but it’s even better to skip that step entirely. Bots can learn your preferences from previous conversations and continually improve the user experience.

Same question, personalized formatting of answer depending on who is asking the question.

Example 2 — Filters:
Let’s say you’re in charge of Facebook campaigns in France. In a GUI like the tenjin dashboard above, you would need to set filters manually to drill down the numbers to only include France. Based on your previous questions and conversations, bots will figure out automatically that you’re the Facebook campaign manager for France and personalize the responses based on your preference.

Example 3 — Routines:
As a campaign manager, you aim to optimise your campaigns every day but might not always get around to it. Because you’ve developed a habit of asking your bot about the campaigns, it will proactively send you the numbers every day.

These might sound like small improvements, but taken together, personalization based on your preferences will have a huge impact on improving the UX.

Bot-native UX: New levels of personalization for everyone

Our early experiments indicate that bots enable new levels of personalized UX. We expect the first wave of bots to mainly personalize based on a job level, partly because it’s a good idea to start with a narrow domain (as we pointed out in our last post). Over time, personalization will become more sophisticated and take your preferences into account as well. But while conversational interfaces make it practical to present personalized software, it’s not at all obvious how these bots will be built.

Thanks to Alan and Matthäus for your great input on this post!


Interested to join us building more human-like bots? We’re hiring!

One clap, two clap, three clap, forty?

By clapping more or less, you can signal to us which stories really stand out.