Bots are a category of automation tools that run autonomously to complete work. At a simple level, it could be a script that executes task, as Dawn Sieh, PhD Manager of Finance Talent Development at Verizon, explains in this report automation example:
I was in the office late in December, and one of my coworkers contacted me to explain that it takes most of an uninterrupted day to complete one of her reports and she wanted to discuss if there might be a more efficient way to publish it. We reviewed the current approach and together we identified opportunities to simplify the report and reduce the time to publish it. I suggested she contact a representative of our ‘bot’ team to see if they could assist with our idea. The bot team was able to automate several steps, and now the report runs with the click of the mouse and the bot moves the data into the appropriate tool for the users to access the report online.
Verizon has other bots operating at a higher level of sophistication, such as natural language generation. As Dr. Sieh explained, at the end of the month, “We use a software product called Quill to read our financial results and generate text for us that explains what happened, what variances existed and why they exist. This is preparation for the [human] analysts who read, review and refine as necessary, in preparation for meeting with business partners.”
Practitioners are adopting the practice of personifying bots as team members to whom work is delegated. Dr. Sieh says the attitude she sees is, “I have a new co-worker, it happens to be a bot and I can give it the highly transactional and time consuming tasks to I can focus my time on the more value-add work!” Often, they are given names and sometimes even a dedicated workstation! Sarah Schaus, Assistant Treasurer and AVP for Allianz Life Insurance Company, describes a computer that runs algorithms and reports. It requires the correct passwords and access codes to run its reports. They further “personified” the bot by giving it a body made of cardboard boxes and tubes! Creating an identity will help the team relate to the bot as a partner to whom work is given and results received to the best of the bot’s ability.
As the bots increase in sophistication, they are also transitioning from defined work to defined processes, where the outcome is unknown. How should we accept their input?
Chris Argent, founder of GENCFO, has some suggestions: “First, you need to test the results. The name ‘data science’ may indicate precision, but there is a lot of testing in order to get the algorithm right. Is it performing to your expectation? How do you know if a component of the script is working right?”
Second, Argent counsels knowing how a computer thinks. “A computer will miss nuance and may throw away a piece of data that does not fit pre-defined specifications, say an outlier,” he said. “A content expert may look at that data point and say it is relevant in the context of the business and needs to stay.” Similarly, a bot may push forward with its script where an SME would stop working if the data is fuzzy, results unexpected, or conclusion is illogical. Like a colleague with whom you have one-on-one meetings and coaching sessions, bots need to be retrained on the data to make sure their results remain relevant.