AI in the workplace - Spoke the help bot

A short article about how the AI help bot, “Spoke", can be used in workplace and hereby the HR department in a company.

Sources:

askspoke.com - 21 Ways AI Is Transforming the Workplace in 2019

askspoke.com - How automation is changing help desk support

askspoke.com - Self-service AI

Background for the article

You have proably seen or heard how AI can help companies by applying AI to classify for example certain customers, reviews or how well trained chatbots that can help customers.

Maybe does people often think about how to use AI in their business units and to their external customers, but today one of the key success factors for a company can be their employees and how they are treated.

This will often point to the HR departments, where these HR people are helping the employees with questions about their employment, contract, salary, vacation, absence, etc.

I work in a HR department myself and there I often see how the administrative colleagues get several questions every day, all from questions regarding their latest pay check, maternity or how to register vacation.

Many of the questions are often some sort of general questions and therefor also have some sort of standardized answers or may be directed to specific pages on the intranet. But even the questions are easy to answer, it takes a lot time for the HR people to answer each employee, and the questions are just informative for the employees, but probably not value generating.

So it would be nice to use AI to develop a machine that can help answering all these more common questions from the employees, so the HR can focus more on the impact jobs and these that can improve the employee satisfaction and value.

Fortunately, there are already developed AI models often called help bots, where I found a interesting help bot called Spoke, which I will talk about in this article.

Why use AI and deep learning?

A help bot without AI and deep learning, will demand a lot of programming and changing versions, when new questions can often come up regarding workplace. For example, the new Danish vacation law that soon takes effect but where there are still a lot new and uncertain answers, would demand allot of retraining the model.

With AI and deep learning, you can instead “feed” (train) the help bot with new data in form of new questions and answers, so the model kan adapt to changes, instead of creating new models/systems when change occurs.

So AI and machine learning help creating more efficient and effective models, because you only create one model, that you can then adapt to changes by “simply” add new data and train it with this data.

What is Spoke?

Spoke is a help bot that can help the HR people by answering questions from the employees over Slack, email, SMS or the web, where Spoke will answer the employee prompt with a friendly response.

Spoke can answer up to 50% of the questions on its own, before the question reach a HR member and therefor it can save a lot of the HR peoples time. Spoke can respond in several ways, for example a quick instruction, link to HR system or intranet or attach a PDF file, depending on which method it is trained best to.

Below you can see an example how the Spoke bot answer an employee’s question through Slack.

Not every request is a simple question, but here will Spoke identify questions which require a human touch from the HR people. Spoke will file them under the right team, assign to a team member, and if needed, serve up a simple form based on the type of request – see picture below.

How is Spoke trained?

Spoke is trained by machine learning and NLP algorithms which we know from M2, where the bot has managed to identify certain words and matched them by relevant knowledge, for example some instructions or a link to a tutorial.

Some of the questions that Spoke can’t answer and perhaps are assigned to a HR member, for example regarding to the new vacation law, can easily be added as new knowledge (data) to the help bot, which should improve its performance (the percent of answered questions).

Below you can see how a question about the wifi password where assigned to a HR member, where she afterwards add the question and answer to Spoke’s knowledge base (KB).

So, this function should help the Spoke bot getting better and better when new questions and answers are added to the bot’s knowledge and this makes sense, where we know that more data can improve a model’s performance.

What about when knowledge (Q&A) changes?

An obvious question is what about when the answer to a question changes, for example the above picture with the wifi password.

In the knowledge base which contain all the data for the bot, can you easily set review dates for when its time to review the data or just go through the knowledge and see if some of the data needs to be edited or deleted.

Below you can see how a knowledge about holiday schedule can be flagged for review, edited or deleted.

Last words

Spoke is a great example of how AI can be used in HR to minimize their workload from some of the many regular questions from employees and how it can be trained and fitted continuously, so it can improve, and resist avoid outdated answers.

A system like this can help companies and their HR people by saving time on general questions, so they can focus on more value generating operations for the employees.

I Hope you enjoyed this short article and could use it see how AI can be used in the workplace.

Sincerely.

Michael