Nothing grabs readers’ attention like bad news. For example, technology leaders everywhere paid close attention to the recent MIT study, which found nearly all generative AI pilots fail to deliver real financial results – 95%, to be exact.
Research from RAND has been similarly bleak, showing that four out of five AI projects stall out. And S&P Global discovered that organizations are abandoning AI initiatives at double the rate of the previous year.
Field CTO for AI at MongoDB.
We’ve all read think pieces that diagnose these failures as a result of weak models, immature tools or a lack of internal expertise and capability. However, the truth is simpler and perhaps even more uncomfortable.
The reality is that most AI projects fail not because of the technology, but instead because of the strategy—or lack thereof—behind them.
The problem: Treating AI like a product you can buy
Anyone walking the floor of an enterprise software conference will be greeted by a sea of vendors hawking shelf-ready “AI solutions”. Implicitly, the message they send is that AI is a product that someone can purchase, plug in, and then watch the efficiency benefits roll in.
And it is this highly commercialized view of AI, treating it like a stock-keeping unit (SKU) that you can order from a catalogue, that leads to AI projects unravelling shortly after purchase.
This is because AI is not a pre-boxed answer in search of a use case. Instead, it is a set of techniques that only create value when they are applied to a specific, well-defined business issue. So without a clear path to ROI from the start, it is highly likely that AI projects will fail.
When organizations forget this, they fall into what many leaders sometimes refer to as the “science-experiment” trap. It appears in two common forms. In the first, we may see internal excitement about a new model or tool, and the pilot may even secure generous funding from senior leadership, followed by an impressive demo.
But within a few months, it will fade away because no one can link it to measurable outcomes.
The MIT research confirms this, finding that purchased AI tools succeed 67% of the time while internal builds succeed only one-third as often, largely because vendor solutions come with clearer use cases and success metrics tied to specific business outcomes.
The second version starts at the top. This is when leadership puts out a call for an “AI strategy,” and in the rush to demonstrate action, an organization buys hardware or software that looks impressive on paper.
Only once it arrives does the team realize they don’t have a clear use for it, or the AI fails to deliver. In both cases, we see the same story: when the business problem comes last, there’s no space for value.
Reframing AI around business reality
The good news is that there is a better way to evaluate and prioritize AI investments. At its core, it comes down to how well teams and leaders distinguish between AI ideas that are and aren’t worth pursuing.
The starting point is to identify the biggest challenges facing an organization and what matters to its core performance. For most, this will be revenue growth, efficiency, customer satisfaction, risk exposure or perhaps productivity. The conversation must start with a business need, not a technical curiosity.
The next step is to assess the data required to solve those challenges. In many organisations, data is scattered and duplicated across multiple sources or trapped in legacy systems.
Until this data is consolidated and cleaned up, there is no model on earth that will be able to work with it effectively. Once the legwork is done to get the data estate in order, the potential for AI to solve this problem becomes tangible.
Then it’s important to consider measurability, and how the organization already measures things like revenue growth, efficiency or customer satisfaction. Having this baseline in place means any improvements from AI can be demonstrated clearly.
Metrics such as cycle time, accuracy, cost per transaction or customer scores give you the ability to show impact.
Allowing AI to earn its place in the business
By applying these considerations, you’ll likely see the number of potential AI projects shrink. That’s a good thing—because the projects that remain will be directly to business value and supported by data that’s fit for purpose. Teams can then set success criteria that everyone understands and can measure over time.
In turn, this approach can shift AI from being a costly experiment to adding tangible value. It focuses investment on what is solvable now rather than what is theoretically exciting. And most importantly, it enables AI to truly deliver on its promises of transformation.
We’ve featured the best AI website builder.
This article was produced as part of TechRadarPro’s Expert Insights channel where we feature the best and brightest minds in the technology industry today. The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/news/submit-your-story-to-techradar-pro
https://cdn.mos.cms.futurecdn.net/zAD3N7HhEBrYMu6PimoCfn-970-80.jpg
Source link
