How Hybrid AI Could Change the Way Organizations Learn
Leidos is investigating the combination of generative and agentic capabilities to transform training and instruction
Three Points to Remember
- Knowledge transfer and management is fundamental to the success of organizations, but many encounter difficulties training and upskilling personnel.
- Leidos is exploring the benefits of applying generative and agentic AI capabilities in tandem to training strategy, instructional design and learning-content creation.
- Hybrid AI systems in educational technology have the potential to facilitate real-time adaptive training and learning by observing user behaviors and tailoring activities.
Organizations of all sizes struggle with a common problem: how to train effectively so people learn efficiently and apply new knowledge while determining the effectiveness of training. From federal agencies upskilling their personnel to the military training its service members, success depends on whether information sticks and translates into impactful actions.
Leidos is exploring how generative and agentic AI can improve training, instructional design and learning to accelerate knowledge transfer and management inside organizations. Generative AI is good at creating content at scale; agentic AI is good at analyzing an objective and taking steps to achieve it. Bringing them together in a hybrid AI system could help deliver training that’s rapidly adaptive, timely and precisely tuned.
“Oftentimes we have the challenge of being able to measure the effectiveness of training or knowledge transfer,” said Jose Romero-Mariona, Leidos’ director of innovation for information and data sciences. “Is there anything we can do along the way, dynamically adjusting what we’re teaching, to better affect learning outcomes?”
A persistent training challenge
Traditional training programs are often built around standardized learning materials. Courses and lessons are planned in advance, delivered within a set period and evaluated at the end. It leaves little room for any adjustment let alone real-time tweaks if some learners struggle.
“We typically find ourselves with a syllabus and what we’re going to do over, let’s say, five weeks,” said Romero-Mariona, who has taught cybersecurity to U.S. Navy personnel. “Even if we get some intel on whether the instruction is being impactful, breaking from the rigorous, predetermined course schedule is difficult.”
Ineffective training carries costs and consequences. Government personnel and warfighters have limited time for training. Training that isn’t absorbed means knowledge isn’t applied. At the other side of learning desks, organizations and their leaders need to be confident that their training investments produce measurable results.
The ultimate goal is not just being able to generate information but to manage and transfer it to become actionable and tangible knowledge.
Jose Romero-Mariona
Leidos Director of Innovation for Information and Data Sciences
Two types of AI working together
Generative AI has already found its way into many workplaces, producing massive amounts of content almost instantaneously, but it often doesn’t understand context. Agentic AI, on the other hand, is designed to pursue goals, make decisions and adapt with feedback. It can mimic human logic, continually gauging progress toward intended outcomes.
In a hybrid framework, generative AI helps create learning content and materials, while agentic AI tracks how learners interact with those materials and adjusts the training experience as needed. Generative outputs are connected with agentic decision-making through feedback loops, making the system suitable to support real-time adaptive learning and what’s called cognitive knowledge management.
This can allow organizations to see if training is effective as it unfolds rather than after it ends. The AI agent can support the design of a training plan and help steer how that plan is executed.
“It’s like if you’re having a conversation with someone and trying to explain something and you can tell if it’s just not in their head,” Romero-Mariona said. “You shift the conversation or turn it back to your audience and ask them to relate to the topic.”
During a lesson, learners may be asked to interact with the training material or real-time questions might be subtly embedded for them to answer – similar to how a teacher quickly checks his or her class for understanding.
Course-correcting on the fly
Depending on how much learners are struggling, the agentic AI can task the generative AI to deploy adjustments ranging from small tweaks to reconfiguring modules to redesigning learning exercises. Over time, the system learns which approaches work best for different audiences.
“This type of framework can drive toward semi-automating different versions of a training program, when we take those lessons learned back to the AI agent,” Romero-Mariona noted. “Ultimately, it makes better use of personnel’s very limited time.”
While AI use is expanding across government, applying it to educational approaches is still in the infancy stage, according to Romero-Mariona. Beyond training, he sees the hybrid AI framework being useful in strategy planning for a range of sectors and missions – wherever combining generative capabilities to synthesize information and agentic capabilities to evaluate past and current performance of knowledge transfer can boost impact.
In health care, for example, physicians could use generative and agentic AI in tandem to fine-tune patient care plans based on machine analysis of health records and past treatment decisions.
“We still need trained physicians to make the final decisions on patient care, but when you bring in the agentic portion, now you have a smart support system that can help look at decision points and their impacts,” Romero-Mariona said. “The ultimate goal is not just being able to generate information but to manage and transfer it to become actionable and tangible knowledge.”
MORE ON LEIDOS AI CAPABILITIES
