We are moving aggressively into a period of automation across IT services. Yet, the industry has to adopt anything like the automotive industry to talk comparatively about what will eventually be an IT artificial intelligence (AI) hierarchy.
Given that this process has advanced significantly with autonomous driving, why not use that framework to create a similar framework for autonomous IT operations.
The danger with any discussion on IT automation is the concern about job losses that can destroy almost any automation effort, because the status of critical decision makers is often connected to the number of people they manage. But with today’s massive shortage of qualified and particularly diverse candidates, the bigger problem, which automation can address with less controversy, is staffing shortages that have become problematic for most firms.
We aren’t anywhere near full automation yet, which will have staffing implications, but the industry will have years to get there. Since IT is generally understaffed, it will be a while before this effort results in downsizing. But this trend does suggest that retraining be considered as part of any progressive automation process.
Automotive Levels of Automation
Autonomous cars have six levels.
Level 0 is where we started, where the driver does everything.
Level 1 is cruise control, where only the throttle is handled somewhat automatically to avoid hitting the car in front of you.
Level 2 is cruise control plus lane keeping, and most advanced systems are considered Level 2+ now, where the car can drive itself, but the driver has to keep their hands on the wheel in case there is an automated oops.
Level 3 is where things get interesting, and the driver can give up driving during certain conditions, like stop-and-go traffic or freeways. At Level 3, you can’t relax and have to be ready to assume control at any moment.
Level 4 can give the car complete control. Still, the driver may occasionally be asked to step in if the car sees a unique problem, like a flooded underpass or an unexpected accident, and needs non-critical help to solve the problem.
Level 5 is what we generally think of when we talk about autonomous driving. This level is where you don’t even need to be in the car, as it is fully autonomous.
Autonomous Levels of IT
Using the automotive autonomous levels and a guideline, I’m suggesting the following:
Level 0 again would be no automation at all. You might get reports and alerts, but someone virtually always has to act on them.
Level 1 is where you get some basic automation, but the scripts aren’t dynamic. While this level can address some of the repetitive tasks, the overall capability is pretty basic.
Level 2 adds a machine learning (ML) aspect. Some non-critical systems are automated, but mission-critical systems are still mainly under manual control. However, this is where digital assistant concepts may appear to provide timely advice and can be enabled to act on that advice in a limited way.
Level 3 uses machine learning AIs much more aggressively, and an increasing number of functions happen automatically with only some oversight. However, should a problem occur, people still need to be on hand to address it, or the system may have to shut down to prevent damage.
Level 4 starts moving from machine learning to deep learning, and increasingly the system can provide its oversight and remediation. Pressure on IT staff is vastly reduced, and only occasionally does a data scientist need to get involved. This level could be fully provided by an IT company or systems integrator and remotely monitored for alerts. You begin to have conversational AIs to handle customer requests and deal with trouble calls. IT can shift from day-to-day operations to focus on technological advancement and expansion of those operations.
Level 5 is IT in a box. This level is a full deep learning deployment where the system determines what is needed and acts on the problem independently. I doubt we can deliver Level 5 before we get to general-purpose AIs, which aren’t due until late next decade, so we have some time. By this time, much of IT will need to be retrained to do something else, because this would be similar to outsourcing IT. But in this case, you are outsourcing it to an AI, which a third-party vendor in the cloud could provide.
Wrapping Up: Trust Is Critical
Critical to both the automotive and IT AI-level progression is trust, and while automotive may essentially skip Level 3, I don’t expect that same thing to happen with IT. This difference is because IT needs to trust that each level works as intended before advancing to the next level. IT will also need time to retrain displaced employees, particularly as we advance past Level 3.
While the market has some Level 3 solutions, adoption is light, and Level 4 and 5 systems aren’t expected to be available for years yet, let alone broadly adopted.
And IT isn’t the only function being automated. Sales, particularly online and telesales, will be moving to conversational AIs. Accounting is moving aggressively to blockchain and related automated functions. And we are moving toward turning manufacturing plants into big automated 3D printers, some of which will be located at customer sites.
Granted, much of this won’t get done until mid-century or after many of us retire. Still, those just entering the workforce will likely see these automated advancements as their way of life and regular retraining as a significant part of their future careers. These efforts will lower costs, increase productivity, and make the related firms far more competitive if these advancements are done right. Done wrong, they’ll put companies under at machine speeds, so make sure you have the right partner, a partner that understands both the technology and your operations. Developing the internal skills to understand the technology and avoiding being played by a vendor become far more critical, given the ability for a poorly trained or implemented AI to do massive amounts of damage at machine speeds.
We are just at the start of digital transformation. What is coming will potentially free us from those tasks we hate, create opportunities that we don’t see, and have significant potential to change operations for good if done right and ill if not. Given how badly most early AI deployments have gone, I doubt many are even moderately ready for what’s coming and deploying AI.