One area that I think is under-appreciated is how organizations understand AI and Natural Language Processing (NLP) initiatives in general. That’s what I call AI readiness or AI maturity.
Implementing AI in your business is a big step with far greater implications than your average technology investment. For most businesses, AI is ushering in transformations that affect your technical infrastructure, your data and processes, the skills of your teams, and culture in general (to name a few). An organization that is in the observation or pilot phase is very different from an organization that has completed two or three implementations. It also differentiates itself from a company that has artificial intelligence embedded in many processes.
This is why it is critical that organizations embarking on an AI / NLP journey are aware of where they are in terms of readiness at each stage. They need to understand what AI can and cannot do to ensure they bring to the table all the stakeholders who will be involved or affected by the journey. And most importantly, everyone is aligned on WHY you need NLP, the root problem to be solved, the desired results, how they will be measured, and what will happen if you don’t solve the problem NOW.
What is AI Maturity?
So what is maturity in the use of AI technologies? When we think of AI maturity, it refers to the degree to which an organization is using AI in their business. Gartner’s: An AI Maturity Model outlines five levels. From the planning stage (level 1), where organizations select use cases but have no pilots, to transformation (level 5), where AI is embedded in how the organization does business.
Many factors influence the way an organization matures. It is understood that technical capabilities and budget can inhibit an organization’s ability to go beyond an initial pilot or deployment and beyond. From there, teams face more challenges (as we’ll cover below) that can hinder progress. In the global AI software revenue forecast for 2022; Gartner quotes “Reluctance to adopt AI, lack of trust in AI, and difficulties in delivering business value from AI” as factors holding back organizations from getting to the point where they have a proactive AI strategy and robust infrastructure.
Exploring different levels of AI maturity
In our survey of NLP practitioners, we identified three levels of maturity:
- Evaluation and testing phase. Organizations that are just starting their AI journey through pilots and tests, but do not currently have NLP models in production.
- Early stage. Organizations with one or more NLP models in production for two years or less.
- Mature stage. organizations that have more than two years of production of one or more NLP models.
We asked respondents to identify the top three challenges based on their level of maturity. Here we look at how to address three key challenges.
Recommendations for increasing the maturity curve of artificial intelligence
1. Choose the right use case to prioritize
Early stage companies face many challenges. These are usually related to the visibility of the NLP project at the organizational level.
The ideal scenario is when the NLP initiative is presented throughout the organization as a strategic initiative and its economic implications. If all important departments are aware of the WHY of an NLP project, it can help facilitate cross-collaboration and teamwork.
The second challenge is the internal alignment of the expected benefits of the project. Most NLP projects start with teams who are aware of how the NLP project is built, how it is tested, and the benefits the project ultimately offers to the organization. However, it often happens that other teams that can be affected are not involved; this can be a big risk for adoption.
In this case, we recommend involving the end users involved in the implementation from the very beginning. If the end users understand what it means to them, it will align with the high-level goals and objectives of the project.
2. Build the business case
Experience with more deployments brings benefits as well as challenges. Organizations with more mature deployments will have hands-on experience to help them avoid common mistakes and pitfalls that can hinder progress.
However, the ability to justify the costs associated with building or scaling NLP modeling and tools is one area where these organizations struggle. Here we recommend leveraging past success with other NLP initiatives to demonstrate positive ROI. It will always be easier to advocate for more funding when you can demonstrate that projects are delivering results.
Another challenge is achieving a certain level of accuracy for projects. Mature organizations that have tested the technology have high expectations for the accuracy and reliability of the results. In fact, the time to value may actually be longer than for immature organizations because there is a continuous need to refine and refine results to get closer to the ideal state. In this case, we’d recommend sticking to the 80/20 rule rather than delaying launch, adjusting expectations and pushing forward NLP model in production, gathering feedback from end users and improving the solution in small steps. Don’t let “perfect” be the enemy of “good.”
3. Choose projects with high ROI
You cannot fail if you choose projects with high ROI. One of the biggest mistakes we’ve seen is when organizations decide to solve a problem that nobody cares about. Instead, we recommend focusing on problems that an NLP initiative can solve quickly and, most importantly, on problems that have an immediate meaningful economic impact on the organization.
A project that solves a known problem and gets results can also inspire enthusiasm in the rest of the organization and channel wider support.
Finally, we cannot talk about the success of an NLP project without mentioning data and security. Read our next post, Prepare for NLP program success. data and management to prepare for the data security and management aspects of your next NLP project and get the complete guide; The Roadmap to NLP Success.
Get the ultimate guide to NLP project success
Our experience at expert.ai has taught us a few things about what teams need to be successful in this important, transformative investment.
Download the guide