C3PO head on a desk
PHOTO: jens johnsson

Chatbots, or conversational user interfaces, allow people to interact with systems in more human ways.

In the workplace they promise to tackle information overload, simplify services and lighten the load of repetitive tasks using rules and artificial intelligence (AI).

The rapid development of AI promises a future where bots will run routine tasks better and cheaper than humans. 

This, in turn, will bring changes in the way we look at organizational structures, management and hierarchy.

But while 2017 was the year of chatbot hype, in 2018 they’re falling out of favor.

Related Article: Workplace Chatbots: Too Little, Too Soon?

The Winter of Our Chatbot Discontent

This month FinTech startup Digit ditched its chatbot, announcing “we think chatbots haven’t lived up to their promise — and we’re done believing that they will.”

A survey by chatbots.org found consumers are lukewarm about bots. Of those surveyed, 53 percent said they found chatbots to be “not effective” or “somewhat effective,” while 22 percent felt bots weren’t smart enough.

But we shouldn’t lose sight of the potential of chatbots in the intelligent workplace, because it’s at work where a step-change in experience is badly needed.

Organizations have introduced a lot of technology to help people be productive and engaged. But this is often a hot mess of user experiences and designs, increasing cognitive load and damaging productivity.

Chatbots can streamline and unify the employee experience, but only if they’re well-designed.  

The problem is that few have been. Instead, vendors have rushed to incorporate chatbots without considering what the user value proposition is. Organizations need to start with user needs, and determine if a chatbot is the most effective way to meet these.

Finding Your Chatbot's Voice

If a chatbot is the best way to deliver a service or information, it should ideally be built into a user’s existing habits — for example in Slack or Facebook’s Work Chat — rather than as standalone apps, so that it doesn’t interrupt the flow of work.

Chatbots also need to work like good online services. They should walk people step-by-step through a journey or transaction, and give reassurance on progress toward the goal. They have to behave like a graphical user interface, albeit one that sits within a messenger app.

But chatbots need to do more than provide information. To work, they need to feel like real conversations. The chatbot needs to project a personality that has the qualities of a real person. It’s an extension of your brand, so it needs to talk in your brand voice.

Crafting a personality — with a purpose, matching characteristics and suitable tonality — is a critical part of chatbot development. You’ll need to invest time in extending existing tone of voice guidelines.

Related Article: The Rise of the Digital Workplace Chatbot

Tailoring Content for Chatbot Delivery

Remember that getting mechanics such as conversation paths, keywords, and call-to-actions right is only a part of the chatbot experience.

For users, content is the user experience. For that experience to be a good one, the content returned needs to be optimized for the channel, which is predominantly a mobile one. Think Google Now-like cards, not PDFs or long chunks of text.

If possible, content should be designed for the bot, just as media producers have begun producing short versions of content for Alexa.

Chatbots provide an alternative interface for search much of the time. But enterprises have long struggled to get search right — chatbots aren’t going to fix this out of the box.

A bot framework can work wonders at parsing badly-spelled queries. But, just like enterprise search, they need training on the specifics of your company such as names of people and products. On the plus side, chat interfaces tend to produce longer query strings, increasing the likelihood of relevant results.

The content management system of the (near) future will transform search results into chat or voice summaries. Until then, consider content-managing snippets for chat or voice search for common queries to improve user satisfaction.

An Integrated Approach to Virtual Assistants

The chatbots.org survey found users are frustrated by the fractured experience virtual assistants provide. This is because many have been built as a solution for single broken user journeys, rather than a coherent product, forcing users to go to multiple bots for individual tasks.

Insurance startup Lemonade took an integrated approach. It has an ambition to build an autonomous organisation. A single bot, Cooper, sits at the heart of this. It acts as a scrum master, managing the workflow between Product, Engineering, Testing and QA, and running automated tests.

Lemonade recently expanded Cooper to answer knowledge base queries, measure KPIs and more. This is a neat implementation as it sits within the existing workflow (Slack), and doesn’t force users to remember the functionality and competence of multiple siloed bots.

Part of a successful deployment involves anticipating when and where issues may happen. For example, any business thinking of using a chatbot needs to look at the end-to-end journey, including making sure when a bot hands off to a human it doesn't force the user to repeat information.

The Promise of Enterprise Chatbots 

The buzz around chatbots is how they’re used in the consumer space, but I think enterprise is where it's at.  

First, this is a safer space. The tech isn't mature and sometimes disappoints, but mistakes within a firewall don't carry the same reputational risk. It's a perfect place to experiment and help stakeholders get comfortable with bots.

Second, the growth in consumer use of messenger is being mirrored in the workplace, thanks in particular to Facebook’s Workplace, making it easier than ever to buy, develop and deploy bots at work. For mobile-oriented tasks and workforces, they can put content and services into the apps people love using, so they’re accessible at the point of need.

But to take conversational interfaces to the next level, it’s essential we get the basics right — a value proposition, good content, clear user journeys and wayfinding — and ensure we have robust governance to avoid virtual assistants recreating the problems we’re trying to fix.

Get them right, and bots do the hard work to keep things simple for your users, delivering productivity gains and an improved employee experience in the process. So while chatbots may have their naysayers, I believe bots have a role to play in the intelligent workplace.