To Bot or not to Bot?

These days everyone, myself included, seems to be talking about AI (Artificial Intelligence), Machine Learning and Bots (Robots, Chat Bots, Cognitive Agents and Voice-First technologies, to name a few). I've been lucky enough to be involved in a few AI projects over the past couple of years, working alongside tech firms to try and create Bots for local authorities, and even building a few personally. What follows are a few of the lessons I have learnt along the way. But first, a short preamble...

It is sometimes difficult to distinguish between the marketing hyperbole that accompanies new technologies and their actual maturity. This is no different for the range of solutions that fall under the umbrella of AI. In some cases, the technology is already mature. Machine Learning predictive analytic tools are one example; Robotic Process Automation is another. Coming up close behind are speech, vision and sentiment analysis.

For other forms of AI the dilemma isn’t in the technology itself, rather it is about finding solutions that work out of the box, without a lot of bespoke “coding and configuration”. By some described as “solutions looking for problems”, Bots can be designed to work well when focused on a single, specific service. Applying them to a wide range of services is however almost exponentially more complex.

With these considerations in mind, let's look at a few of the lessons I've learnt along the way...

Bots don't do anything unless you programme them to do so. The simpler ones use drag and drop techniques. For anything more sophisticated you need developers to write reams of code, business process experts to map and redesign process workflows, language experts to build the library of utterances that people will use to engage with the Bot, and many, many iterations of testing and refining the process and the solution.

There aren't any off-the-shelf Bot solutions (to the best of my knowledge) available for complex organisations with complex needs. This is a area of AI that is still maturing. Today, Bots are ideally suited to single service solutions. The more services you add the exponentially more complex it becomes. However, existing solutions may give you the same, with less cost and less development time.

Consider the demographic you are targeting. Are your target audience likely to have access to the platforms you will be using? Are they more or less likely to choose this channel over others? 

Is the customer journey likely to be easier using a Bot? A “series of yes/no questions is likely to be far simpler to implement using an electronic form.

Not all marketing reflects the current state of play, rather a vision of things to come. Look for evidence of solutions that are in place and working, rather than relying on marketing material or press releases.

Consider co-development with other authorities rather than going it alone. Shared development and support costs will be individually lower, and you will have a common platform that can be shared more widely.

Don’t give away your intellectual property. If you’re offered something for free, you’re probably giving away your IP. 

Whilst these points may give you pause for thought, I for one am positive about the future potential of Bots, in particular voice-first solutions. Over the next couple of years I expect these technologies will mature into more rounded tools, easier to develop, with less need for specialist development. For me it’s not a case of if we should consider using Bots, rather how and when we choose to engage with the experts in this field. Ideally we will come together, so we gain the economies of scale that come from doing it once and doing it well. Here's hoping!