Implementing and integrating AI into your organization, especially your first foray, can be a bit daunting. Through this series of posts, I will seek to allay those fears by illustrating key strategies and best practices regarding development and deployment, through the eyes of the key researchers and practitioners.
First, we want to start with the problem in mind. The specific use case may not be explicitly defined until further information has been gathered, but we want to make sure our approach centers around the intent to assists or augment the end-users. Doing so helps ensure their initial buy-in, continued cooperation, and increases the likelihood of solution adoption at the conclusion of the project.
If you’re reading this article, you’re likely familiar with OpenAI’s ChatGPT, a chatbot based on a large language model and similar paid or public offerings such as Google Gemini, Microsoft CoPilot, and Claude from Anthropic. These interactive large language models are the most commonly illustrated examples of AI, but they are far from the only options. As of this writing over 550,000 models were listed on HuggingFace.co’s repository, specializing in categories such as computer vision, natural language processing, audio, reinforcement learning, tabulation and graph machine learning. While selecting a tool at this juncture would be premature, it’s important to know that AI models are typically trained to learn a defined set of skills and therefore differ in the capabilities in which they can perform various tasks.
This assessment can occur through direct or indirect engagement methods. Direct methods include one-on-one interviews, focus groups, and workshops. One-on-one conversations allow for in-depth exploration of the end-user's challenges, desired outcomes, and how AI might provide solutions. Prepare structured yet open-ended questions tailored to their specific roles or pain points. Focus group discussions with a representative sample of end-users can provide valuable insights into broader trends and patterns of needs within the target audience. Facilitated workshops with multiple stakeholders foster collaborative discussions. These can help identify common needs, potential conflicts, and diverse priorities across different groups that will be affected by the AI system.
Other methods of direct discovery include job shadowing and process analysis. Observing end-users in their work environments gives a direct understanding of the tasks they perform, the bottlenecks they face, and potential areas where AI could offer automation or assistance. Breaking down existing workflows step-by-step helps pinpoint inefficiencies, repetitive tasks, or decision points where AI integration could offer value.
Indirect feedback can be gathered through well-designed surveys that allow for collecting quantitative and qualitative data from a larger number of potential end-users, identifying common needs and preferences. Taking process analysis a step further, leveraging data analysis against usage patterns, employee and departmental metrics can sometimes reveal hidden patterns in user behavior that might be bolstered through AI.
Needs gathering shouldn't be a one-time activity. It's a continuous process as the AI system is developed and feedback is received. Include stakeholders representing various departments, levels of expertise, and potential end-users of the AI system to ensure well-rounded understanding, and clearly articulate the potential benefits and limitations of AI to stakeholders from the beginning to manage expectations and foster acceptance.
As AI becomes more sophisticated, it may become capable of taking on more complex tasks that were previously thought to be dependent on human intelligence alone. The long-term impact of AI on the job market is still uncertain, and this uncertainty creates real anxiety for workers who fear their skills may become obsolete. When communicating AI initiatives to your user base, be cognizant of those concerns. Provide your users with ample training to use the tools or interfaces provided and create a safe space for them to hone their abilities in leveraging AI in the workplace and share with others. As an aside, it’s also recommended to choose just one tool to provide to end users starting out, as to not provide them with whiplash alternating from model to model as each leapfrogs the other.
Once this process is repeated with multiple user groups, it will develop into a strategic vision for the company – a vision that may include future capabilities may be involved one, three or even five years down the road. Once that roadmap is established, measurable objectives can be defined, and a team can start to develop its initial foray into AI. The continuation of this journey will be the subject matter of future posts. Stay tuned!
Contact us at info@nvisionkc.com right away to start a journey towards AI integrations that are based on facts. Let us show you the manifold possibilities on how to optimize your operations and strategy for your company.
.png)