10 Random Things I’ve Learned About AI

Not in priority order, but as I step back and examine the AI space and my own journey to date, some things come to mind that I’d thought I’d share that may help you on your own journey:

1. There are a TON of startup companies in our space. It pays to do some independent research to understand the backgrounds, financial stability and capabilities of these companies before entering into a long-term commitment.
2. Agentic AI has opened the door to easily serve up existing knowledge to end users via a chatbot. The bot can either search your custom knowledge base, previously resolved incidents or the Internet for possible resolutions.
   1. If you don’t have a knowledge base or your knowledge base is only meant for a technical audience while your end users are not, this may not work well.
   2. If your knowledge base is out of date, this may not work well.
   3. If the resolution entered for incidents is ,“I fixed it” that will not be useful to the bot.
3. In the past 12 months, I’m seeing more emphasis on AI tools that can resolve incidents and requests so they never even reach an agent. From password resets to system access requests, these systems can free up staff to handle more complex issues. Before implementing, be sure to partner with your IT Security and/or Legal teams so that they understand the risks and rewards before potentially granting access to your most sensitive systems and data.
4. The value that you’re going to get from deploying AI that can leverage automation for routine incidents is relative to how much you’ve focused on “shift left” initiatives to date. For example, if a high number of requests on the service desk are for password resets, account provisioning, access requests, etc., there are opportunities. However, if the vast majority of your incidents involve technical troubleshooting where the resolutions are complicated and can’t generally be resolved with scripts, these will be initially harder to program using AI.
5. Many companies offer a free evaluation process where they will look at a sample of your incidents and come back to you with how they can potentially help through automation, deflection, etc. Be sure to check with your legal counsel before sending data (they may require a Non-Disclosure Agreement) and carefully review what you’re sending to ensure that it does not contain Social Security numbers, passwords or other sensitive company or personal information.
6. The time to build ITSM system integrations and workflows is dramatically decreasing (may help with No. 3). The latest systems now can automatically build their own integrations to other systems and also build their own workflows just by watching what agents do. Expect this to become even easier in time.
7. Tracking incident trends has been one of the hardest things to do in my ITSM. AI promises assistance with this. I’m excited to try it out as it could streamline efforts with problem management.
8. There’s a lot of hype out there in terms of deflection rates, automation potential, etc.It pays to do your research to completely understand how a vendor can help you and ensure that the payback is realistic for your specific use cases and system setup.
9. This field is rapidly evolving as companies find new ways to bring value in the IT Operations space. This article is likely going to be out of date in six months and laughable in its antiquity in two years.
10. If you’re not spending time looking at AI solutions on the service desk, you are likely doing a disservice to your employer. AI will not go away. It’s important that we all stay on top of what’s ahead.