It comes as no surprise to learn that the number-one shadow IT application last year was ChatGPT, and it seems a safe bet that the generative IT chatbot will hang onto that position through 2024. But when shadow IT encompasses Gen AI — particularly the chatbots and applications that aim to make people more productive — an IT organization could end up in a real bind.
The battle against shadow IT has always been Sisyphean, whether the challenge came from Wi-Fi access points brought into the office, personal smartphones, or any of a host of niche SaaS applications that users or whole departments expense without bothering to tell IT. And IT’s response invariably seems to be a two step: First, ban the app or device and explain the risk it presents; then, when no one heeds the warning and policy, either bring the shadow IT element under IT’s umbrella or offer an approved app or device that promises the functionality that attracted users in the first place.
That second option may seem appealing to organizations that see the potential in generative AI. Vendors like Microsoft, Cisco, Zoom, and Google have all released gen AI-driven assistants that could be used in place of Chat GPT. And, for that matter, OpenAI seems to have launched a more aggressive push for its own Enterprise version of Chat GPT.
The traditional “fine, you all get this option” approach tended to offer ubiquitous deployment of the previously-shadow IT capability, or at least deployment targeted at the specific departments that were most insistent about needing this tool. In any case, the strategy eventually became one in which anyone who wanted the shadow IT capability could get an approved or IT-managed version.
That’s not necessarily going to be the case with the gen AI assistants, at least not right away. I’m not aware of many large enterprises that are planning to roll out Microsoft Copilot or Google Gemini to anyone who wants the genAI tools, given the $30/user/month cost. So enterprises now are struggling with the decision of who gets these tools and who waits, possibly indefinitely. And users who are forced to wait may decide to just keep on using ChatGPT – potentially doing things such as uploading company data and information like a company’s strategic goals into the broader tool as they make work-related requests.
The two forces are fundamentally at odds for enterprise management: If you’re not going to roll out gen AI assistants to everyone, then by definition you’re imposing top-down control over who gets those tools. And IT’s exercise of top-down control is what gives us shadow IT in the first place.
Zoom execs would probably be quick to point out that this is not a problem with their assistant, AI Companion, which is available at no additional cost. This at least allows IT to be a bit more prescriptive about what AI tool is allowed.
This issue isn’t likely to be settled any time soon. But if you’re looking for a deeper dive on the topic of AI assistants, their risks and potential, I’d encourage you to consider joining us in Santa Clara, CA Oct. 1 – 2 for Enterprise Connect AI, a debut event focused exclusively on the tough decisions and transformative opportunities that confront enterprises with the explosive growth of AI for the enterprise. You can get some basic information here, and registration will open in late spring/early summer. Stay tuned – it’s a topic you can’t afford to miss out on.