I’ve been doing the rounds with major global service providers and IT businesses recently, discussing the many possibilities that generative AI offers. The responses have been a mixed bag. For instance, I’ve encountered IT leaders who have flat-out told me, “We’ve banned all large language models.” When I asked them, “How large?” they couldn’t provide an answer or a real reason why they’ve banned it. It seems to be a reactionary move, stemming from a lack of understanding and a desire to control the situation out of fear. They’re not alone. Recent data from Cisco reveals that 27% of businesses have blanket banned the use of AI entirely.
The root of the issue lies in the fundamental approach IT departments take towards technology. They’ve had a set way of testing for over 30 years, building disciplines around service management and software development. Testing is the very backbone of corporate IT, ensuring system stability through well-defined scripts: if X input is given, Y outcome is expected.
This meticulous approach has served traditional IT systems well, which are deterministic and follow clear logic paths. However, GenAI systems are different. They are non-deterministic, meaning they can produce varied outputs for the same input. Imagine a GenAI system that produces a “brown cat” as a result. An equally valid response could be “cat that is brown” or even “beige cat,” depending on context. Traditional IT testing frameworks, which rely on exact matches, can’t handle this variability. This mismatch leads IT departments to ban GenAI outright, as they can’t control it or guarantee its outcomes.
The challenge of technical debt and AI
Another challenge is the concept of technical debt. Technical debt (also known as tech debt or code debt) is a way of describing what happens when development teams take actions to speed up the delivery of a piece of functionality or a project that later needs to be restructured. In other words, it’s the result of prioritizing rapid delivery over perfect code.
In traditional IT systems, managing technical debt involves clear, deterministic processes. However, with GenAI, the non-deterministic nature complicates this. For instance, if a GenAI system performs an analysis and returns “brown cat” as a result, an equally correct answer might be “cat that is brown.” The exact phrasing can vary, making it challenging to apply the same testing and refactoring processes being used in traditional systems.
This discrepancy between expected and actual outputs goes against the grain of established IT testing methodologies. Unless IT departments adapt to this new way of working, they will continue to take a blanket approach of banning GenAI, and progress will be stunted. The IT department rhetoric seems to be: If AI can’t be controlled or precisely predicted, it might break things, so it’s safer to ban it altogether. This attitude needs to change.
Addressing security risks
In my view, another big reason for pushback on GenAI is that the IT department simply doesn’t understand it. The only real security risk with GenAI is the question of ‘Is my data going to be used to train a model?’ But so long as the large language model is private, then there’s virtually no issue. If the LLM is a private model, your data will not be compromised, and with off-the-shelf models like OpenAI’s GPT-4o, you can tick a box that opts you out of them using any data to train their model.
Shifting the focus from an IT goal to a business goal
To move beyond the old ways of doing things, we need to transform IT departments into business-focused entities. This means shifting leadership from IT heads to business operations leaders, with tech experts supporting business goals. GenAI is about outcomes, not just outputs. By focusing on business needs and desired results, we can fundamentally change how IT functions operate to allow room for GenAI to work.
Global systems integrators (GSIs) face the same challenges. Recently, I spoke with some AI leaders within GSIs who refused to test a corporate chatbot because they couldn’t risk the variability involved. Understanding how GenAI works and its limitations is crucial for adoption.
We need to approach this transformation by establishing a clear and collaborative environment between IT and business units. Encouraging IT departments to understand the business context and desired outcomes will help bridge the gap. Training and upskilling IT professionals on the nuances of GenAI and its testing methodologies can pave the way for smoother integration.
Involve employees too
This segways nicely into the importance of not only keeping your employees in the loop but empowering everyone in the business to contribute to AI adoption. As with every digital transformation effort, GenAI isn’t just about the latest shiny tool; it’s about fundamentally changing the way you work, and that requires everyone in the business to be on board and actively working towards the same goal. Recent research has emerged showing that 33% of employees are worried about AI taking their jobs. Businesses must be sensitive to this and create an environment where employees feel empowered and comfortable using the technology, as understanding AI is key to working with it.
For instance, at my own company Enate, we’ve embraced this approach by purchasing and deploying an AI testing platform that our testing team (not the IT department) has configured and installed. This hands-on involvement demystifies AI, allowing employees to see firsthand how it can be tailored to meet their specific needs and challenges.
Other departments use GenAI, too and using the best tools available is generally likened to having your assistant. Enate Product Managers use GPT for their first line of code, and the Marketing Team uses GPT and DALLE for a range of tasks, from proofreading to freelancer briefs. By enabling those who are directly impacted by AI to lead its operationalization, businesses can create a more inclusive and transparent setting, upon which GenAI can grow.
Moving beyond traditional IT frameworks
AI’s potential is immense, but it requires a transformative shift in how we approach technology in organizations. By moving beyond traditional IT frameworks and focusing on business outcomes, we can harness AI effectively. It’s a collective effort that involves the entire business, not just one department.
Let’s work together to embrace AI’s possibilities and drive businesses forward. If we don’t, we’ll get left behind in the AI race.
