5 Steps to Get your Company Ready for Generative AI in Training
The unbelievable wave of articles, announcements, and speculation triggered by the recent release of open-source generative AI models like ChatGPT and Bard is hard to process. There's no doubt we're in the middle of something huge.
Is it really that big a deal for corporate learning, training, upskilling and coaching? Yes - buckle up.
We're already seeing incredible results in developing training bots that can be tailored to deliver conversation and chat simulations, rate and coach new associate skills, and even act as a personal tutor to cover complicated products or company policies. Even before generative AI, we saw how conversation, back office, and software simulations could reduce training costs 25-50% compared to more traditional training methods. We also saw new hires gain new skills faster, impacting key metrics like NPS, error rates, sales closure, and AHT. The promise of generative AI is to get these kinds of results even faster.
To reach that destination, there's a whole range of planning and upskilling activities that L&D leaders should consider. Here are a few of the top challenges we've seen our customers face this year as they've begun to explore generative AI learning experiences with Bright.
You need a secure, ethical approach: There are meaningful topics to explore around data security, managing PII, avoiding bias, and controlling for AI model 'hallucinations.' You need to get educated more generally about how these models work, and how they do/don't protect your company and customer data. To avoid issues like the one recently experienced at Samsung, you'll need a policy for their use in training, and beyond.
You need to prep your training materials: The old adage of 'garbage in/garbage out' applies. There are some cool quick wins possible for starting to re-think ways to use existing training materials via LLM models. But you really won't want to just 'drop your training materials' into an LLM. First, because letting LLMs 'figure it out' doesn't get the results you really want (in our humble opinion) from a learning science perspective. Second, because you'll need an approach that lets you quickly update information on processes, compliance, pricing, and more to keep models current and coherent.
You need to better-define quality and performance: Beyond basic 'training', there are incredible experiences coming live at Bright through conversation, chat, back office, and other types of simulations. But it's not enough to just 'role play' with AI - you need to rate and coach as well. You'll need a thoughtful exploration of 'what good looks like' in order to fine tune your coaching models before you see the full power and results.
You need to upskill your L&D teams: Time previously spent by trainers to facilitate live training or deliver role play and certifications will need to be reinvested into AI-powered content development, simulation design, and fine-tuning your models. It's not just that your trainers don't have this skill yet - almost nobody has this skill yet! So you need a plan to get there. First, figure out where you plan to start exploring generative AI. Then document a multi-month plan with milestones for building team skill so they can help steward a forward-leaning L&D learning strategy.
You need a vision beyond ChatGPT: There are a lot of companies out there doing little more than 're-skinning' ChatGPT. They're adding little innovation beyond the baseline model. You need to understand what's really possible so that as you start to select generative AI vendor partners or tools, you're picking the one that truly fits your needs.
If you'd like to start taking steps to address the items above, and even begin using a custom large language model fine-tuned on your training materials, customer personas, and quality standards - reach out to us for a consultation and demo.