30% of Gen AI Projects will Fail by 2025. Here’s Why and How to avoid it.

Make your GenAI initiative a successful investment. Know how to avoid silly and costly mistakes.

Every business must have come across the increasing hype around Gen AI and its applications within organizations. However, we cannot deny its potential in improving business workflows, productivity, and efficiency. Thus, they are investing resources into Gen AI pilots, reaching $12 billion in 2024, following a breakout year in 2023.

But, does your business need GenAI or does it have a positive impact on your business?

Well, this is the right question to ask.

In a recent event, Air Canada paid CA$1.4k+ as compensation due to their failed Gen AI initiative, as their virtual assistant provides false information that impacts a user. A major AI blunder.

Approximately 90% of Gen AI pilots will not be moving to the production phase and 30% will be completely abandoned by 2025. Businesses are targeting Gen AI for improvements but for the sake of implementing it, without knowing its implications and loopholes.

Reasons Why Most Companies Fail to Introduce Gen-AI- Crucial Roadblocks to Consider.

94% of businesses believe that AI will help them unlock success in the future. But, how many of them would be able to achieve success as expected? I guess, a few, considering the low success rate.

In our experience of developing Gen AI projects for our clients, we have come across a few roadblocks that every business should work on.

To make the most out of your Gen AI initiative, businesses must focus on the following roadblocks to a successful Gen AI initiative.

• Gen AI talent shortage. Without the right team, no business can build a strong strategy to follow and meet goals. 62% of business executives lack the skills to execute their AI strategy.
• Data governance issues. 58% of organizations fear risks that come with the adoption of Gen AI, especially cybersecurity, data management, and governance-related issues.
• Huge upfront adoption cost. Not every business can simply just implement Gen AI, it needs a huge investment in place to make it happen. Many businesses are lagging due to the cost factor.

• Traditional Infrastructure. Gen AI needs modern infrastructure for successful implementation and scale. 35% of companies lack the infrastructure to support AI implementation. Get in touch with our digital transformation experts to update your traditional systems using Gen AI.
• The small mistake comes at a huge cost. AI has threats like hallucinations and generating dummy data or fake data that can impact the outcomes if no data strategy is in place. 27.2% of companies don’t trust generative AI outputs. A recent example has already been shared at the beginning of this blog.

These are very confusing numbers that deflect businesses from adopting the Gen AI approach within their organizations. These factors not only stop businesses from adopting Gen AI but also act as a failure if not considered at an early stage.

In addition to all these barriers, other factors can affect an ongoing Gen AI initiative. Let’s understand what these are.

Factors That Can Affect an Ongoing Gen AI Initiative

Some of the common mistakes that a business can make with their ongoing Gen AI efforts will eventually affect it.

1. Fail to adopt a change management approach

AI and Gen AI are relatively new technologies with fewer skills available. It means businesses need to upskill and train their teams and create awareness about Gen AI. It is not about just implementing AI into existing systems, teams need to understand what they can achieve using AI and what improvements they are expecting from it. However, most teams feel challenged to cope with new technology, hence failed initiatives.

It requires a complete change in culture in how they oversee Gen AI and its implications, and align their outcomes with business goals. While some have fear that Gen AI might take up their job. Generative AI might replace 85 million jobs globally by 2025, giving rise to new job roles.

Solution- this is why businesses must adopt change management from the start. They must understand the need to change and make it an organization-wide approach. Find out the reason for resistance and overcome it with the right approach and leaders in place. Hire the right leaders to upskill the workforce and prepare them to adopt new technologies.

2. Overexpecting Gen AI Outcomes

Shreya Shankar, a machine learning engineer at Viaduct, says that generative AI (genAI) has its pros and cons. The advantage is that it simplifies data preparation, which used to be very difficult. This ease can lead to unexpectedly good initial results, which gets people excited to experiment more.

However, skipping detailed data work can create unrealistic expectations. Since many people don’t evaluate their models systematically before using them, they rely on “vibes” rather than solid data, which isn’t reliable.

Many people become overconfident from the casual use of tools like ChatGPT and end up disappointed because their expectations are too high. Most of the content generated using ChatGPT is emotionless, it is not engaging, which might impact your brand value.

Solution- Success in machine learning often comes from constant fine-tuning and improvement. With genAI, this crucial work is often missing because it’s so quick to start. Additionally, interacting with generative AI can be inconsistent, like giving instructions to a teenager. It requires both skill and experience to manage these tools well.

3. Not able to maintain AI models over time

Whatever Gen AI model you choose for your business will be suitable for a certain amount of time, as the data might have grown over time. If you do not maintain your models, you might not get the accurate results that you expect.

Lack of maintenance is a major problem for getting good results from GenAI, according to a Deloitte report. With regular upkeep, you can be sure your models are working correctly and giving accurate information for business decisions.

Creating impressive GenAI applications is just the start. If you don’t have a system for tracking performance and improving based on data, you’ll face two issues:

1. You’ll spend a lot of time managing the system.
2. Your GenAI models will become outdated over time.

Neither of these is a good long-term solution. Without proper maintenance and security, your AI investments won’t stay valuable for long.

Solution- understand your long-term goal using GenAI, what areas it is impacting, and understand the data growth. Then train your models on the growing data to ensure accurate results. You can take the help of GenAI solution providers who can help you train and update your models accordingly.

4. Neglecting data issues

Carm Taglienti, a distinguished engineer at Insight, highlights that most AI project failures come from unrealistic expectations, not technology problems.

For example, a US chip company wanted AI to improve its supply chain but expected it to work perfectly immediately. They didn’t realize that AI projects need adjustments and good data. The company’s data was either hard to get or in a difficult format. This made it tough to use AI effectively.

Similarly, a PwC survey found that 84% of companies face data issues with AI. For example, a company struggled with data access due to licensing and control problems, leading to delays and frustration.

Taglienti advises companies to first evaluate AI projects based on impact, then risk, and lastly data availability. If the data isn’t ready, it’s better to pick a different project.

Solution- always check the source of your data. Make sure that you are not generating harmful data that can impact your brand’s reputation in any way.

5. Ignoring ethical and privacy implications

Gen AI is capable of drawing the best possible results only if businesses consider the ethical and privacy implications it can have due to the following reasons.

• Misinformation and deepfakes. Gen AI can blur the line between reality and fabricated information, which is alarming and can impact business reputation. Recently Facebook has started a project to identify deepfakes.
• Bias and discrimination. Gen AI will use the exact data that you will fed. If they are trained on biased data they will generate inaccurate results that can impact business decisions. Prioritize diversity in training data sets and check for harmful and biased data.
• Copyright and intellectual property. Generative AI that creates content similar to copyrighted works can lead to expensive legal issues and damage reputations. To avoid problems, make sure training content is licensed and use metadata to track content origins.
• Privacy and data security. Generative models using personal data can risk privacy by creating too-accurate synthetic profiles. This can lead to legal issues and loss of user trust. To protect privacy, anonymize data, and improve security. Follow rules like GDPR, HIPPA, and others to only use necessary data and keep personal data secure with strong encryption.

Solutions- make your GenAI initiative responsible. It is only possible if the team understands the criticality of its implications. They must be aware of the available ethical and privacy approaches to include and ensure complete GenAI security.

Other Factors that Can Impact Gen AI Projects

Some other factors can impact your Gen AI project. Consider the following.

• Allocating fewer resources will stop the Gen AI project’s growth, failing to meet business objectives.
• Treating AI as a one-time project. To keep up the momentum, businesses must conduct ongoing maintenance, updates, and adjustments to adapt to new infrastructure and environments.
• Ignoring the infrastructure needs can impact performance and limit its implementation.
• Poor integration with the existing infrastructure and systems can reduce the efficiency of the project and overall business.

Businesses must consider all the factors while starting up a Gen AI or any AI-related project. It has a vast scope only when you make the right decisions, hire the right talents, and have modern solutions to adapt to Gen AI.

Consult with Gen AI experts to understand your business goal and how Gen AI fits there.

How to Make Your Gen AI Project Work?

Doing it all right in one go can be challenging for any business or Gen AI expert. As it is a new technology everyone is still trying their hand on this new technology.

But, as per our expertise in delivering next-gen Gen AI projects, we have listed a few strategies to increase the success chances of your Gen AI projects.

• Set clear, achievable goals for your Gen AI project.
• Use high-quality, relevant data and protect privacy by anonymizing it.
• Choose the right tools and ensure they integrate well with your systems.
• Monitor the AI’s performance and be ready to make adjustments.
• Be aware of ethical issues and comply with regulations.
• Educate stakeholders about what Gen AI can and can’t do.
• Implement strong security measures to protect data.
• Regularly evaluate and improve your approach.
• Seek advice from AI experts to tackle complex challenges.

Final Thoughts

2023 was the year of Gen AI when it all started to expedite implementation and experimentation. With the growing expectations from Gen AI, businesses are blindly adopting it, without understanding its need and implications on existing systems.

Today, 92% of Fortune 500 firms have adopted generative AI to achieve 15.7% cost savings. If this is true, then businesses must be using it. However, most of the Gen AI projects do not reach production and this is highly daunting for businesses investing resources.

The fast progress of generative AI offers great opportunities but also presents challenges that can cause project failures if not handled well. Problems like overestimating AI abilities, integration difficulties, regulatory issues, and internal resistance can be significant hurdles.

At OnGraph, we’ve seen many organizations encounter these issues in their AI projects. By sharing our experiences and strategies, we aim to help others avoid these pitfalls and effectively leverage generative AI.