We announced the acquisition of Bright just a month ago. But we had been working towards a unified vision and product long before that! There are lots of exciting updates to share in coming months… starting with the latest in generative simulations. 🔥 🚀Over the last 7 years we’ve delivered more AI-powered conversation simulations than any company in the world. Lessons learned shaped the infrastructure we put in place for Zenarate’s generative + avatar-based simulations – an exciting complement to our existing NLP/NLU and software simulations. Our approach goes beyond prompt engineering to provide AI that’s ready for the unique requirements + scale of global enterprise:
- Simultaneous Conversation + Software Sims: One of our most impactful offerings through the years has been AI conversation simulations delivered at the same time as fully branching software simulations. We’ve extended this offering into our generative AI simulation offerings.
- Configurable Skills + Coaching: The top predictor of AI simulation success is the alignment of skills ratings + coaching to brand standards. Unfortunately, GPT tends to provide overly generic coaching guidance. Most providers we’ve seen entering this space have had to resort to constant prompt engineering (perhaps more suitably called prompt ‘fiddling’) to get skills rating + coaching to match company expectations.
That’s why we’ve taken extra time to build a human-in-the-loop architecture to skills rating + coaching feedback (the same method that originally made GPT such an innovation). Ongoing feedback makes Zenarate ratings + coaching increasingly accurate over time and – most importantly – enables our coaching to get very specific about how learners can improve.
- Dedicated Persona Models: It’s not hard to get an LLM to reflect a customer persona with one learner (e.g. this customer is an elderly/young; angry/calm; etc). But what about when you need 10,000 learners to interact with a customer? Will your simulated customer act consistently? Our experience indicates that due to hallucination, model drift, and natural variability in learner statements, the answer is no. And this has negative consequences for your ability to deliver consistent learning experiences at scale.
To address this we’ve built dedicated customer persona models based on proprietary data sets that provide consistency and control over how simulated customers and patients respond. You can adjust these to match the unique characteristics of your own customer-base, offering the most lifelike simulation experience possible.
- Scalable Context: What happens when a process, price, or policy changes that’s referenced across 1,000 simulations? Do you have to update them each individually? (ugh!) The architecture we’ve built makes it easy to update knowledge, policy, and other context centrally, making ongoing simulation management easy.
- Safe + Secure: The trusting relationship between a learner and their employer is essential. So we’ve taken extra precautions to ensure that simulations and coaching experiences stay on track + filter inappropriate / off-topic conversations.
- Voice or Avatar: We know that many roles require face-to-face exchanges where reading the room is key to success. We’ve built deep integrations with HeyGen + Synthesia to accelerate access to avatar-based simulations.
- NLP/NLU Based: Last but not least, generative AI isn’t always the best answer. In some cases enterprise security protocols won’t allow generative tools. In others, from a learning science perspective, enterprises want more control over conversation and coaching flows. Zenarate is the only provider offering both branching NLP/NLU or generative AI simulations.
Reach out to learn more about how Zenarate’s AI architecture + experience training a million learners with AI can open up new possibilities in learning for your organization this year.
