Nataliya Andreychuk is the CEO of Viseven. She is a leading digital transformation evangelist for the pharma and life sciences industries.
Have you ever watched someone step off a boat, and it immediately started leaning to one side or even capsizing because their weight was keeping it balanced? The same thing can happen in companies. When key knowledge holders leave, especially in organizations with data silos, the whole system might collapse.
Replacing a productive worker can cost more than 150% of their annual salary. That figure goes beyond the time spent covering for the employee and recruiting and training someone new. It also reflects the lost productivity when no one else can perform the critical tasks the employee handled.
Transferring knowledge while the expert is still on board isn’t always an option. Their workload is often so intense that they don’t have time to properly train others or review their work. In other cases, employees nearing departure may lose motivation, making them less engaged in mentoring activities.
Top experts may leave for many reasons—career development, family issues or work-life balance. Your focus should be on aspects you can control to retain that institutional knowledge.
AI Model Fine-Tuning: Keeping Expertise Alive
Fine-tuning an AI model empowers companies to turn the model into an expert that never quits. Simply put, it means taking a pretrained large language model and training it on specific data so it can perform a particular task exceptionally well. For instance, a life sciences marketer can use such a model to ensure materials comply with local regulations. At my company, we trained a bot on our internal knowledge to help new employees quickly understand our evolving offerings, while letting senior staff focus on other tasks rather than mentoring. It also allows everyone to refresh their own knowledge as needed.
With fine-tuning, AI becomes a living knowledge base for your organization. Your guidelines, tone, values, terminology, frameworks and approaches can be safely stored in a dedicated memory server so the information can flow easily from one employee to the next. Nothing gets buried in email threads, siloed in departments or lost before you find the perfect hire, who, by the way, will still need time to learn how your company functions.
Fine-tuning also means you’re no longer stuck with general best practices or abstract ideas. Instead, you can deliver content, processes and reports that are on-brand, consistent and compliant, without bottlenecking knowledge in just a few veteran employees.
The Right Way To Do It
When choosing an AI solution vendor, it is essential to clearly define your goals to avoid being swayed by trendy, flashy tools. You should also consider picking a solution designed for your domain. These tools are typically trained on the language and nuances of your industry, which means you will spend less time and effort fine-tuning them.
Your team should prepare data samples to show the AI how things work in your organization. Aim for thousands of examples to make your tool as precise as possible.
Equally important is ensuring your team members track the performance of the AI model and report regularly and in a timely manner, especially when they find errors. The best time to set up the monitoring is during a trial phase so you don’t end up retraining your AI model later.
Many experts say that AI is only as good as the data behind it. That’s certainly true: You need clear, structured and well-organized data (a topic that deserves its own article). But I’d add that AI is also only as good as the prompts it receives.
Recognizing this, the United States Golf Association (USGA) and Deloitte worked together to refine prompt engineering to deliver highly accurate chatbot answers about golf rules that have been cherished since 1894. To achieve similar results, train your employees in crafting prompts effectively to ensure the AI results are always consistent and on point.
The Takeaway
If your business depends on processes that require special skills and deep knowledge, you need to think about how to preserve them. Fine-tuning an AI model can help. That doesn’t mean you’ll no longer need the most experienced people on the team, but you’ll be able to capture some aspects of their work and the frameworks they use. What’s more, fine-tuning can also help lighten your team’s workload by offloading some of the mundane tasks, making employees less likely to burn out and quit.
Forbes Agency Council is an invitation-only community for executives in successful public relations, media strategy, creative and advertising agencies. Do I qualify?