A 4-step model for using AI and automation tools to boost IT efficiency

Almost every day I have a conversation with another IT Director that has leapt to a conclusion about which IT process to automate. My advice? Look before you leap so you don’t leap to the wrong conclusion.

You see, with dozens of different vendors marketing new ways to use AI and automation tools to improve IT efficiency — and the C-suite breathing down IT’s neck to use AI to squeeze more productivity from the team — it can be tempting to just pick a tool, go forth and automate.

However, it’s worth remembering that cautionary tale about the handyman that picked up a hammer. With hammer in hand, everything looks like a nail. But what if a screwdriver (or hand saw or wrench) is really what you need to solve the problem? 

Based on hundreds of conversations I’ve had on this topic, I’ve found the following framework to be effective in helping IT leaders think about how, when and where to deploy AI tools to have the greatest impact on IT operations — especially when it comes to their help desk.

A 4-step model for prioritizing where to start with AI and automation tools in IT operations

Rest easy. I’m not proposing a 10-step, multi-month project. It’s four steps. It should only take 1 to 2 weeks, and the net result is that putting yourself through these paces will let you spend your time and budget in ways that’ll have an outsized impact on your team’s efficiency.

Step 1: Understand your help desk ticket data

Rather than operating on gut feel, it’s important to look at real data in whatever form you can get your hands on. When I talk to IT leaders, without fail they know who’s working the help desk tickets and they have a sense about how busy they are. How they spend their time? That’s when things get murkier.

What help desk data to look for

We offer up a complimentary help desk ticket analysis where we analyze a company’s help desk tickets and then sit down with the IT leader to walk through what we find. Whenever they see their data organized by how much time their team spends on each ticket type, it only takes a couple double-clicks before they say, "Wait! I knew we were spending time on that but I didn't realize <insert epiphany here>!”

That’s why the first and most fundamental step is organizing your data into some high-level categories. Most people get this wrong, and it leads to unhelpful insights and incorrect conclusions. We group tickets into 40 different use cases. If you’re doing this manually, though, try to bucket your tickets into these three categories at a minimum: identity and access management; application issues; and onboarding/offboarding.

Once you’ve gotten that wired up, you can dig into where your team is spending their time. Look at both the volume of tickets and the time spent. We’ve found it helpful to dig into things like:

• Total # of tickets per week (and how that’s growing, if at all)
• Total tickets by use case or problem category
• Time to first touch
• Time to resolution
• Customer satisfaction (CSAT) or user sentiment by use case

If your ticket volume is higher than 20 tickets per day it can also be useful to understand when those tickets are coming in over the course of the day so you can start to forecast future capacity constraints. 

Next, you can move on and look at “how’s it going.” Customer satisfaction (CSAT) surveys can be misleading. If you’ve got them, look at the data. If possible, though, try and assess user sentiment. If you can map user sentiment all the way back to the different ticket categories you’ll start to see some interesting results pretty quickly.  

What to do with your help desk metrics

When we run through this data with prospects some things aren’t surprising (yes … password resets is one big group of tickets). There are, however, also lots of ahas. Here are a few from recent conversations:

• One IT team had gotten proactive and built in some auto-alerting when the tablets in their Zoom rooms went down. It makes sense. We also see these automatic alerts when printers go down or the wifi gets slow. The challenge is those knobs usually get set and never adjusted. In this case, the result was a constant stream of alerts into the help desk that were getting ignored and clogging up the queue. 
• Onboarding is another hot topic. You can see the pain in the data pretty quickly. In one case an IT team was auto-generating about a dozen onboarding tickets when each new employee started. However, whenever there were small adjustments to the requirements tickets would linger. The result was that new employees weren’t getting their equipment configured on time.
• At another company we saw equipment orders auto-generating tickets that required someone on the help desk team to spend 20 minutes closing those out to clear the queue.

Find the things that look weird and the areas where your team is spending the most time. Then, write them down and ask yourself “why?” Ask “why” again and again until you can’t anymore and you’ll realize, they probably shouldn’t be there. But don’t jump to automation just yet. Your next step is: Eliminate.

Step 2: Eliminate help desk tickets

In our experience making a few small tweaks to your tech or processes can slash a wide swath of tickets from your queue. This reduces all the noise in the system that’s often obscuring important insights and can lead you to the wrong conclusions. It’s also putting unnecessary cognitive load on your team. Eliminating tickets gives you a better handle on what’s going on and where you should focus your time when you get to Step 2 (Improve) and Step 4 (Automate).

Usually, with a few small tweaks to rules or processes you can eliminate a sizable batch of tickets. That’s important for two reasons. First, your team is probably spending at least some time clicking on the tickets to clear their queues. More importantly, all those tickets are generating noise that’s getting in the way of what you should really be spending time on.

Which help desk tickets can you eliminate?

In our experience the top candidates for elimination fall into one of these two categories:

• System-generated tickets: Anything that wasn’t created by a person falls in this category. Often a couple configuration changes can eliminate 90% of these tickets. Let’s go back to that example of the tablets in the Zoom rooms. The company had set things up so that the Zoom room tablet would generate a ticket any time it went down. That makes sense. But most of the time it went down it was because the tablet was auto-updating its software. Then it came back up. That doesn’t require a house call from IT.  

• Informational tickets: A lot of apps and third-party vendors are often set up to generate informational tickets. For example, when a server updates or when a new laptop is shipped. You name it. No doubt it was set up for a good reason. Someone wanted to know. But your help desk analysts probably don’t need to know. Those tickets inflate the volume of work that’s actually being done, create work, generate noise and probably skew your response time stats too.

How to eliminate your help desk tickets

Here are common ways to eliminate system-generated and informational tickets. Look at the data and ask yourself if one of the following approaches will work. If so, do it. If not, kick the can to stage 3 — Improve — and we’ll fix it there.

• Turn the tickets off: It sounds obvious but if there’s no value in the system-generated or informational tickets, just turn it off. And you’re done.
• Tweak the threshold for creating a ticket: Going back to that Zoom tablet example. For the 98% of the time that the tablet was simply updating its OS the help desk analyst was still getting pulled away from real work. For tickets like these, changing the threshold – in this case from 0 min to 8 minutes – can eliminate false positives (and the work associated with them).
• Filter tickets out: For informational alerts like those new hire laptop shipments, create a separate category of tickets to get them out of the general queue. Or better yet send them somewhere other than your ticketing system.

Step 3: Improve your IT help desk operations

Now that you’ve got the clutter out of the way it’s easier to get your hands around what can and should be improved. In this phase I recommend that people look for quick wins. If you see something that can be improved in a week or two, do it. If not, move on to Step 4 (Automate) and then come back here later to tackle the long tail.

How to improve your IT help desk operations

In working with dozens of customers here are three of the most common types of tickets that can’t be eliminated but can and should be improved.

• New employee onboarding: At many orgs when a new employee starts it can generate a flurry of tickets to install apps, grant access, ship hardware and more. But most of the time that new engineer who’s starting Monday needs 95% of the same things that the engineer who started last month needed. So, instead of using tickets to manage new hire setups, think about creating a standard build for each department that includes the software and access to the most common apps they’ll need. Then, go one step further and consider hiring a third party to ship equipment.
  
  
• “Birthright access”: There are some application access requests that don’t need approval because they’re honestly not that sensitive and they don’t cost the company more money. I’m thinking about Adobe Acrobat Reader and Notion or Confluence. One IT Director I spoke with calls this “Birthright access.” When requests come for any app on this list, they automatically grant access. No approvals required.

• Remove handoffs: Too often tickets come in to the help desk by default and then have to be reassigned by the help desk analyst. Think about streamlining these to limit handoffs and associated delays. For example, auto-assign all Salesforce requests to your Salesforce admin. This may require a smarter ticketing system.

• Make your help desk team’s job easier: The more context you can give your help desk analysts the faster they can fix issues that pop up. Here’s are a few examples to consider:
  
  • Connect your identity system to your ticketing system so help desk analysts can easily see who’s making the request.
  • Build runbooks/playbooks for common troubleshooting tasks.
  • Integrate your ticketing system with your comms interface (Slack) so your help desk analysts don’t need to swivel between apps

Step 4: Choosing what IT help desk functions to automate

Now that we’ve whittled down the mountain of tickets by eliminating a bunch and improving our process we can focus on what to automate. While it’s tempting to buy an AI or automation tool, you can create more issues if you automate before you think things through. Here are some things to think through as you consider where to apply AI automation tools.

Choosing when AI and automation tools can improve IT efficiency

People often jump to what automation and AI tools could do vs. what it should do. But to figure out what to automate we recommend looking at these things in this order.

• Is the process repetitive? This is kind of obvious so start here. Anything that can be translated into a rule is a good candidate. For example, if someone is locked out of an app, fielding requests with an automated bot and sending password reset emails makes tons of sense. Things like this comprise your universe of potentially automatable things.
• Does it require judgement? Now, look at those potentially automatable things and ask yourself if they require people to make judgements or think through multi-step questions. Anything that requires detailed explanation, or empathy is out. While AI may be good for a lot of things, consoling someone who just lost a week’s worth of work to a corrupted PowerPoint file isn’t one of them. Blue screen on startup? Also, no. Anything that falls in this category — and it’s a lot — isn’t a candidate for automation.
• Does the underlying process change frequently? A common pitfall for AI automation and self-service projects is that the automation works great at first, but then it breaks. Why? It’s usually because something changes. Maybe it’s the required approver, or an app changed or a policy was updated but nobody told the automation tool. If any part of a process is going to change it isn’t a great candidate for automation.
• Can you tap into information you already have? Now that you’ve whittled down your list, think about where you already have information that can help automate the process. Do you have playbooks? How about a rich knowledge base that the AI or automation tool can reference? If so, these are all great candidates to start with.

Next steps

There you have it. Four pretty simple steps.