Bots for ECommerce Business: Improving Your Customer Service

Bots for ECommerce Business: Improving Your Customer Service

From the unaffordable to the affordable.

Providing 24 hours a day, 7 days a week Customer Service used to be the preserve of the large corporate with offices across the globe. They could ‘follow the sun’ and like Arkwright: “open all hours”.

Whilst eCommerce sales were 24/7, Customer Service was stuck at 8/5. Quite an odd mismatch really.

The only viable solutions were email, the trusty FAQ page and other customers on Community pages. Workable, but not great.

In the last 24 months, chatbots have started to provide that 24/7 customer service. Whilst not perfect, they are an improvement on what’s gone before. Especially as handing first interaction.

They have though been expensive. And the preserve of those with deeper pockets. Less so now.

Now, using Microsoft Cognitive Services and the Bot Framework, Bots are affordable and within the reach off all (eCommerce) firms.

How so?

There are two key developments from Microsoft that make the difference to affordability.

First is LUIS, or Language Understanding Intelligent Service. This is a machine learning-based service that can be integrated into Bots and provides a natural language interaction with users. Customers and employees can run price queries, stock queries, ordering, account status checking, contact updates and preference setting by emulating human to human interaction.

The second is the Microsoft Bot Framework. This reduce the time needed for the Bot development process quite dramatically, as it provides a rich and full feature set to make interactions between bots and users much simpler.

It also provides access to the other Azure AI and Cognitive Services (e.g. image recognition) as well as the Cortana engine (speech).

The last element is Microsoft’s FAQ Maker. We can point it at your existing FAQ page to help short cut the learning process a lot.

None of this works work out of the box though. The Bot needs training to understand the customer’s intent and the desired response. Once given examples, it predicts intent and desired response more accurately.

So, it is important that the development is supported by a thorough analysis to ensure that the business needs are understood so the Bots can be trained effectively.

We recommend first understanding if a Bot works or not for your business. We run a simple 2 hours workshop to do so.

Then, using the output from the same session, create a Bot training approach.

Bot or Not? The workshop phase

What does that look like?

First, we spend a few hours creating a Proof-of-Concept that demonstrates the potential and limitation of a Bot. We then demo it you and any stakeholders. This helps everyone grasp the concept and a little bit of the technical approach.

In other words, it brings to live what is possible. We do this at the start of the workshop.

Then the workshop session moves on to determine if we, as a group, can generate enough “Bot friendly” and questions to justify the development of a Bot.

We do this by asking your teams to generate the questions they get asked frequently – or, if it is an employee services bot, they would like to ask frequently. We then work through those questions with them to determine if the structure and content of those questions were answer-able (easily) by a bot using LUIS.

Then we map out if those questions were of enough value (to your customers and/or your business). It provides a high-level business case. This tells us whether it makes sense commercially to proceed.

The development phase

At the end of the workshop phase we have a business case and a set of test questions to kick-start the development. We can pause and develop a robust business case as needed.

We then move into the build phase, which is probably 2-4 weeks.

During the build phase, we factor in “Refinement questions” to focus the results and an in-Bot feedback process, which provides structured feedback from the test user base. This feedback is structured and allows us to train LUIS further on the “utterances” and the “intent”.

Finally, the bot is then run in Test mode for a number of weeks to help refine its understanding and “intelligence” before being released in Beta.

By following these steps, we ensure that once the bot goes live, it is able to interact with employees and customers using natural language and solving queries that otherwise would require human input and thus, reducing costs for the business.

The process of refinement doesn’t stop there. Each month we help train the Bot adding further refinements and functionality. It doesn’t take long, a couple of hours a month, but it helps improve the offering.

So, in fewer than six weeks, you can go from 8/5 customer services to 24/7 – and – either cut the cost of existing Customer Service or, more likely, offer improved Customer Service.


Leave a reply

Your email address will not be published. Required fields are marked *