Remember back about 10 years ago when virtual agents required you to script each and every reply and possible interaction in advance to work? That was cool, but not practical due to the amount of overhead required to maintain that on most service desks.
Continuing my search for ways to make my desk more efficient, a few years later at an HDI conference, I participated in a vendor demonstration of AI where a virtual agent was able to look up a company vacation policy by request.
That’s great, I thought, but who calls the service desk asking for the company vacation policy? We’re not there yet, I thought. The next year, at another demo, we advanced to leveraging a virtual agent to schedule a meeting. Making progress, I thought, but who calls the service desk to schedule a meeting? The technology was clearly evolving, but I didn’t think that it was practical for supporting my end users who are all in the fast-paced restaurant business.
But two years ago, I thought we had arrived. I participated in a vendor demo that leveraged agentic AI to service knowledge articles to end users on request. That caught my attention because I was focused on reducing handle time internally on my service desk. After taking time to thoroughly research the vendor space last year, we deployed that piece. It’s working well internally, especially for quickly bringing new hires up-to-speed on the desk, after their initial classroom training.
Here are some things that I’ve learned, so far:
- Veteran staff know where to find knowledge articles, so they aren’t using it.
- Our AI vendor is now automatically able to do sentiment analysis and draft work notes. We’ll be testing this within the month. That might get my veteran agents re-engaged, if it’s accurate and saves them time.
- Our team currently interacts with the internal facing bot using Teams. But there’s too much screen switching going on between that, our ITSM and other tools. Soon, our vendor will have this built directly into our ITSM.
- I need to allow the virtual agent to access previous incidents to get better search results.
- Our vendor is releasing automation capabilities in a few months. This part fascinates me with its potential to work incidents without human intervention.However, we’re going to proceed slowly and cautiously here.
Note that we have nothing facing our end users at the moment. Working in the restaurant business, our store management teams want fast, easy service and they aren’t tech savvy. They won’t settle for the agent just serving up knowledge articles to fix their own issues.
In my view, the ultimate play is deploying virtual agents with automation facing our restaurants. Instead of the virtual agent just serving up static knowledge articles to managers, the virtual agent can resolve issues likely faster than the service desk can. I’m hoping to deploy this in 2027.
Some tips from my experience:
- There are a lot of vendors in this space. Take time to research them thoroughly to find one that is a good fit for your needs and budget.
- It’s hard to relaunch a virtual agent to end users if you fail (in their eyes) the first time you release it.
- Virtual agents are just as smart as the information you give them. Make sure your knowledge base is up-to-date and written in a language your end users can understand (or create a separate non-technical knowledge base just for them).
- Involve your IT Security team from the start. They need to get comfortable with the access being provided to virtual agents, especially with automation.
- Partner with your legal counsel to carefully review all proposed agreements so that you know where your data is going.
- This technology is rapidly evolving. If you last surveyed the marketplace last year, it’s time to look again.
This is a journey. The road, at least for me, has had some twists and turns in it, but we’re moving forward. Be open to learning and discovery. Just keep moving forward!