AI co-pilots (or digital assistants) aren't just the future, they're already built into the workings of technology.
A major step forward from memorization chatbots that spit out screenshots and FAQs, these bots are much more than just IT whisperers. They leverage the strengths of machine learning to handle strategic tasks, solve entirely new problems, handle tedious tasks, and circumvent evolving crises. In short, they democratize IT knowledge. And all in plain English.
So it's not surprising that vendors have been killing off co-pilot faster than you can say the word in the last six months. Hurray for efficiency! IT is now for everyone.
But don't be fooled by the hype: beneath the sparkling surface of efficiency and empowerment lies a web of intertwining issues that can upset the delicate balance of your IT ecosystem.
The co-pilot's paradox
If you believe the current Copilot marketing pitch, you'd think even your grandma could quickly write a complex script or troubleshoot a network problem by typing in a few simple English prompts. But we know the reality can be quite different.
The problem isn't language: these AI assistants may be fluent in English, but they don't speak the same language when it comes to technical terms, specifications, ideas, and vendor jargon.
It is not inconceivable that every vendor has a favorite co-pilot trained on their own data and optimized to favor their own product. It's a Tower of Babel situation where standardization is a moving target. It's like trying to understand the conversation of 12 people who speak different dialects of nerdiness. It's similar to what's happening with cloud technologies, but multiplied several times.
But for IT professionals, this also creates an unnecessary conundrum. Many are selling Copilot as a single AI assistant solution that will remove technology friction and increase efficiency. The problem is, in today's environment, there are many vendors, sometimes with competing products stacked next to each other.
Imagine a tired developer juggling multiple projects and platforms, having to appease the quirks of various co-pilots: this makes knowledge transfer difficult (a key promise of co-pilots) and locks developers into specific vendor ecosystems, forcing them to play favorites.
Calculating the cost of human intelligence for improving AI efficiency
You may have heard the adage, “A co-pilot complements human expertise, not replaces it.” That's true if you're already a human expert. But what if you're just getting started?
There is an undeniable risk that in five years' time, these AI assistants will simplify the realm of IT expertise: With instant answers and preemptive solutions at our fingertips, we may inadvertently rob the next generation of IT professionals of the valuable experience of learning from their mistakes.
I liken this to learning to ride a bike with training wheels: you can stand upright, but you don't gain the balance or reflexes needed to tackle real-world challenges. Similarly, a co-pilot can shield you from the tedious, frustrating, but ultimately rewarding process of troubleshooting and problem-solving. The result can be that you end up a skilled teleprompter who's a lazy learner, rather than a domain expert who quickly identifies problems.
There's also the issue of overlooking potential problems and inefficiencies. Copilot is always learning, and getting better every day, but it's not perfect because the real world, use cases, and data are not perfect.
So, if a breach occurs due to a misconfiguration that Security Copilot overlooks (because they weren't trained), who is to blame? The AI? The users who blindly trusted the AI? The vendor that created the AI? This creates a lack of accountability, leaving organizations vulnerable.
Get out of the co-pilot danger zone
So what's the solution? First, let's not waste our money. Copilot certainly has and will continue to have the potential to revolutionize IT efficiency and adoption. But we must approach Copilot with an appropriate amount of skepticism and a proactive strategy.
But I think you need to:
- Focus on interoperability: Vendors need to stop building walled gardens and start collaborating. Open APIs are a first step, but ultimately, a common language for Copilot (e.g., based on industry or domain expertise) needs to be created. This will make knowledge transfer seamless and prevent vendor lock-in. This is important because it will soon be inevitable that all users (not just IT users) will be working in the AI Copilot ecosystem.
- Balancing human and AI synergy: Copilot should be seen as a tool that complements human expertise, not a crutch to replace it. We must foster a culture of continuous learning and encourage users to challenge the AI's recommendations. As the Copilot algorithm learns faster, it will appreciate the user's challenges. Sure, this is fundamental to critical thinking, but in technology, it's essential if we don't want to suddenly stumble upon outliers or events we haven't learned from.
- Make accountability non-negotiable: Organizations need to clarify accountability for AI-assisted actions. This means creating robust governance frameworks and holding humans and AI accountable for their respective decisions. In many organizations I've spoken to, this is still in its early stages; most legal teams have yet to dig into the real-world issues of working with multiple copilots.
Conclusion
The co-pilot has the power to shape how knowledge is transferred and how work gets done, but we need to recognize that the co-pilot is not a panacea.
In the future, as the workforce becomes more AI-savvy, users may want to cycle through different copilots to accomplish more complex tasks. There is already a demand to create libraries of reusable prompts to automate multiple copilots. These are developments that companies and vendors need to drive.
Other concerns include: What happens to the various co-pilots if an employee moves to another employer? Keep in mind that these co-pilots are trained for personal two-way interactions. And could these co-pilots, which continually evolve with employee input, be subject to privacy laws and shut down because some interactions could lead to personal information?
The answers are still unclear, but it's clear that we need to start asking the tough questions and avoiding the pitfalls now if we are to maximize the co-pilot's potential without sacrificing the human element that is key to IT success.
Otherwise we would be living in a pilotless world, full of competing co-pilots.
Image credit: iStockphoto/Motor
