Last Updated on August 1, 2023 by Editorial Team
Author(s): Rami Iguerwane
Originally published on Towards AI.
Navigating the AI Adoption Maze
Reflecting on five years of helping businesses innovate with AI, uncovering common hurdles, and embracing the lessons learned.
Over the past five years of working with multiple companies and helping them deploy AI solutions, I’ve witnessed some common patterns and decided to share them with you.
In this blog post, we aim to shed light on the hurdles that organizations often encounter (based on my experience) while embracing AI innovation. But fear not! This blog post isn’t just about enumerating challenges, it’s about learning from them.
Let’s dive in!
In the ever-evolving landscape of AI technologies, finding the ideal solution for a specific use case can be similar to searching for a needle in a haystack. Companies eager to embrace the power of AI often embark on an exhaustive quest, dedicating significant time and resources to scout and benchmark various providers.
When it comes to AI adoption, I believe that one size does not fit all. Each use case demands a tailored approach, sometimes necessitating a thorough exploration of multiple AI providers. On average, companies I worked with spent approximately three months engaging in this crucial activity.
As companies struggle with the complexities of scouting and benchmarking, some opt to seek external help from consulting firms. While they can offer valuable insights, their services come at a cost (very expensive in some cases).
To avoid spending too much time in the scouting and benchmarking phase, here are two things you can do:
- Pre-vetted provider lists: Curating a list of pre-vetted AI providers based on their track record and credibility can save you precious time and resources. This approach reduces the need for extensive research and minimizes the risk of choosing the wrong provider. However, keep in mind that innovation can come from unexpected sources, so don’t overlook niche players. Big tech companies (Google, OpenAI, etc.) aren’t the only players in the market.
- “Call a friend”: Collaborate with peers and industry experts to gain valuable insights into AI provider capabilities and their experience using it.
The Onboarding Odyssey of New AI Providers
As companies progress through the initial scouting and benchmarking phase, the next hurdle in the AI adoption journey is the integration process for these providers. This critical step involves setting up accounts, gaining access to their platform/API, and defining security rules, often requiring an additional three months of back-and-forth exchanges.
Moreover, each provider may have unique requirements and interfaces. Understanding the API documentation of each one can easily become overwhelming. All of this adds complexity to the adoption process, particularly for organizations new to AI.
To overcome the integration challenges, here are three things you can do:
- Pilot projects for familiarization: Just get started, don’t wait. Initiate pilot projects or Proof of Concepts to get hands-on experience and practical knowledge of the new AI solution. These smaller-scale projects will not only prepare you for full-scale integration, they will also shed light on different ways in which the technology can be useful.
- Standardized integration procedures: Decide where the data should be stored, what API protocol to use, and what are the 2–3 things you need to validate to be aligned with your company's systems.
Just before you cross the finish line: Reality vs. Expectation
After investing over 6 months and involving many company resources, it is frustrating when over 50% of the initiatives fail and fall short of expectations.
One of the most significant sources of post-implementation frustration is when providers fail to deliver on their high promises. Like many other technologies, but especially when it comes to AI, companies easily buy into marketing claims and fancy demos. They then realize that the actual performance and outcomes do not align with the expectations set during the POC. This misalignment can negatively impact the organization’s (i.e., C-levels and budget owners) perception of AI as a valuable technology.
Moreover, AI solutions pricing can be complex. Depending on the providers, sometimes it includes the model, the hosting, the support, or only a subset of these three. Many providers have moved to a pay-as-you-go model, so pricing also changes with volume.
Finally, keep in mind that AI touches various aspects of the company: from the data you use to the IT system you integrate with all the way to the business requirements. Thus, companies may underestimate the time and resources needed for seamless integration, leading to unforeseen development challenges. Such unanticipated complexities can disrupt the initial project timelines.
To reduce the post-implementation frustration, here are 3 things you can do:
- Try before you buy: You shouldn’t blindly trust what you see on the provider’s marketing page. By digging a little, you may uncover potential gaps between marketing claims and actual performance. Therefore, always run your own tests before committing.
- Transparent pricing negotiations: Ensure you discuss various cost sources (Cost of service, updates, maintenance, etc). Also, make sure you request volume pricing discounts if you can commit to them.
- Help your provider: Don’t hesitate to share feedback with your providers and point out what doesn’t work. They are continually iterating on their service and its performance, by helping them, you’re helping yourself get a better service in the long run.
Challenges abound in the dynamic world of AI adoption, from scouting the right provider to navigating onboarding complexities and post-implementation frustrations. But hopefully, armed with these actionable strategies, you can conquer these hurdles.
Remember, AI adoption is not merely about deploying cutting-edge technologies; it is a journey of innovation and transformation.
Join thousands of data leaders on the AI newsletter. Join over 80,000 subscribers and keep up to date with the latest developments in AI. From research to projects and ideas. If you are building an AI startup, an AI-related product, or a service, we invite you to consider becoming a sponsor.
Published via Towards AI