When you pull a band-aid off, especially if there’s body hair underneath, it tends to hurt. So, a quick yank and a wee yelp and all is good again. We just stuck the Artificial Intelligence band-aid on IT systems and a whole lot of SaaS and mobile apps. Apparently, everything, everywhere, all at once has AI in it now. Why?
This isn’t necessarily a bad place to be in, but it’s good to apply some reason to cut through the hype. To also understand that the innovations we are expecting AI to deliver quickly, may take a little longer than we may think. That AI is not as all-pervasive as we feel based on the hype machine at work.
But there are some great opportunities with AI. If we focus on the right ways to leverage it and understand its weaknesses.
For all the hype of ChatGPT, it’s really just a band-aid on failing IT systems and infrastructure, Generative AI (GAI) that uses Large Language Models (LLMs), isn’t solving any systemic problems, it’s not yet connecting to very many systems in the broader business world, despite the hype. Less than 1% of businesses globally are using GAI tools and those that are, are largely experimenting only. Both from how to use it and the fact that the data they’re connecting to internally is a complete mess, complicating these experiments.
The majority of ChatGPT tools and functions are operating at the surface level of organisations large and small. A reality of the majority of enterprise level businesses is that they’re not adopting AI at a deep, systemic level because they face mounting technology debt and struggle just to maintain existing IT systems. CIOs are fighting to keep the lights on and innovation projects are being put on the back-burner as organisations are stuck in Insecurity Paralysis due to the markets doing weird things.
Back to the messiness of and lock down of data, perhaps the two most important underlying issues. AI systems need a lot of data, especially the AI tool known as Large Language Models (LLMs.) While the perception may be that most organisations, especially large ones, are very data savvy and everything’s connected, it’s quite the opposite in reality. Many companies struggle with data cleanliness and keeping their own analytics programs sorted.
Then there’s cybersecurity. Organisations are having to increase their attention and budgets, on cybersecurity. Especially where the threats from bad actors using GAI tools is mounting.
Companies also tend to keep their data very private. A lot of data protects corporate intellectual property and how they do business. AI tools that leverage this data are most often kept in-house and this means limitations to GAI tools. They may in-license large, varied data sets, but overall, they keep their data cards close to their chests. Most data systems are isolated and connect to proprietary enterprise software systems.
GAI tools like ChatGPT or Midjourney are relying on unstructured, publicly available data. The more valuable data resides in private, locked-down systems. Many companies are already ensuring GAI tools can’t access or use their data.
Within medium to large organisations, data and systems silos remain in place. Most managers from HR to marketing, sales and operations know that information is power, thus the human-centric part of organisational silos. Then there’s the reality of many IT systems that don’t play as well together as one might think.
It’s also important to keep in mind that Artificial Intelligence (AI) is not a singular tool or solution. It is a suite of tools that work alone or together in various formations to solve problems. The companies that provide these tools often compete against each other and this presents problems for organisations trying to make projects play nicely with each other. While there are some very good and excellent solutions working in several sectors, they’re very narrow in focus and application.
The hype today is mostly about GAI tools. But they’re only being applied at the surface level of society and industry. Writing reports, marketing content, gaming search engines, short videos for marketing, writing (terrible) ad copy and marketing technology companies trying to automate marketing and sales. All surface activity.
Complex and mission critical systems to organisations is where budgets go to maintain them. Since they are so critical, experimentation with this systems is extremely limited at best. They also tend to be built on legacy platforms and old code. Because of this, a lot of IT solutions are work-arounds and rest outside the main IT infrastructure.
This makes GAI and many AI tools, simply band-aids resting atop a nightmare of messy databases, poorly integrated systems and heavy technology debt laden systems. There’s a lot of hype, not a lot of reality checks as I wrote about here.
If all these enterprise systems, data lakes and data warehouses, ERP tools, accounting systems and so on, played well together and were easy to manage, there’d be no multi-billion dollar IT consulting businesses.
While that sounds like a bleak assessment of the state of AI, there are still plenty of opportunities. The best options for AI companies is being able to work at the edge of enterprise clients. Delivering SaaS based AI services or those accessible through PaaS (Platform-as-a-Service.) These are the short-term plays while enterprise solutions catch up.
Leave a Reply