The ‘Christmas Present of the Year (2022)’ goes to OpenAI for ChatGPT! If you haven’t heard of this AI utility yet, you either aren’t reading the news or social media – I may even go as far as suggesting, you are purposely existing inside a vacuum.
Riding the AI/ChatGPT bandwagon should be on everyone’s to do list for 2023, if not already being undertaken within the first few months. Within our network, we’ve already seen it implemented in a number of ways: website chatbot (with crazy accuracy), bug-checking for programming code, even as automated support ticket responses with a little human oversight.
Two telling-tales of how AI is going to play out: 1) when Microsoft offers billions of dollars to augment the service, and 2) right-click, ‘Add to Dictionary’.
IMS is pressing this tool into immediately action. Here are three areas where we are adopting AI or planning on using it.
HELP THE HELP DESK WITH EMAIL RESPONSES TO TICKETS
No, you aren’t getting an automated robotic response when you create a ticket within the IMS portal. You should be getting an empathetic and thorough response. This format comes from AI. This is more than a standardized template in action; this is the ability to recognize content, context, and emotion and create an outline and action plan to help resolve issues more quickly.
Here is how we see this capability playing out within our environment. We feed the AI model our Knowledge Base articles as well as some default resolution plans for common issues. As a ticket is received, we ask our AI for a response to include references to Knowledge Base articles and possible resolution plans. From here, we could even inject these likely resolutions and steps directly into the email acknowledgment, allowing for self-service resolutions. However, we feel our support gurus’ personal touch is what best adds context and compassion to the responses. We can leverage AI to more quickly identify top culprits, include actions to resolve and improve customer satisfaction with shorter response and resolution times.
ChatGPT distributed a 60-plus page slideshow (on LinkedIn) with some great examples of how you can feed your various email and response templates into your model. Then, you can feed a ticket request into ChatGPT, and it will return an appropriate response. The more content you can feed into the system, the more accurate the responses can be. I have heard of other businesses using this method, followed by a quick human review, and getting great responses that don’t need to be altered. The first few steps can be automated using the AI (https://beta.openai.com/examples).
ALERT MONITORING AND ANALYSIS
A quick search on whatever engine you prefer will share that technologists can be inundated with alert notifications. It is critical to receive alerts, but getting them tuned to eliminate all false positives is not practical, nor is it desired. You want the notifications so that you can gleam the possible sick tree within the forest.
Technology companies will filter notifications into separate mail folders and write email rules to trigger human activity – to double-check the service is active during the next poll (between 5 and 15 minutes). If the service is live, the original notification is then marked as completed with no need to follow up. Reviews and responses only take 15 to 30 seconds, but that time adds up as your environment scales.
Let’s automate that process entirely by also feeding the alert notifications into your AI model and having the AI model trigger an action plan. You could program various notifications or even script actions to take when the AI recognizes two consecutive alerts of the same type. Going one step further, analyze problematic areas within your environment by feeding all your alerts into the AI model (please make certain you scrub any access identification or credentials from your data) and asking how often events happen or for a list of event types by frequency.
As mentioned in the introduction of this article, Microsoft is investing billions of dollars into OpenAI (the owners of ChatGPT). Their idea is to leverage this technology to better integrate their response actions within their platforms and software (cloud and other). Microsoft Lighthouse could then really take over as the dominant fully integrated support and monitoring tool in the business space as it could not only detect issues within Azure, Servers, Active Directory, Security, Desktops, etc., but also take immediate actions to rectify any issues. It could even go as far as noticing if something was developing as an issue and providing the action plan to resolve (and even do it with a simple acknowledgment from an admin).
Depending on the form and function of your blog content, you may be able to rely directly on ChatGPT or other AI tools to simply knock out content (500 to 1000 words at a time) in blazing speeds. In December, while technically on vacation, a couple of us tested the output from ChatGPT in writing blog content for us. It was great, but it isn’t perfect for our needs for a few reasons.
The first issue we found is that the technical content isn’t timely and modern, and there is a delay in uptake. We used ChatGPT to write the Business Premium article, and specific details about the capability of the license were a year old. Some technology changes with greater frequency, and without the model being fed the most recent information, which you can do in other scenarios, the output isn’t 100% accurate.
If your purpose was blog-stuffing for SEO-based writing, it’s a gem. However, once pricing is nailed down, how competitive will it be versus paying $5-$25 for technical SEO articles from fiverr. As a quick side note, I think these services may also start baselining cost at the lower end as these writers are also going to adopt the use of ChatGPT.
We’re using ChatGPT to do our research for articles and develop outlines for content. The responses typically lack a reflection of the information into practical use – what we call experience. Often, readers are looking for examples of experience to qualify or quantify their desired outcomes. Using ChatGPT saves us an easy four to eight hours of research and organizational time per blog post. We can develop and deploy content within one day versus multiple days. We’re technologists, not marketers – your results may vary.