The idea for my company began with a familiar frustration: my mother received a traffic ticket and did not want to spend hours sitting in a classroom to complete traffic school. At the time, I was studying law and working in test preparation, where I had seen what digital delivery could make possible. Why, I wondered, did a required course have to depend on a physical classroom?
That question eventually led me to found CyberActive in 2000. But building an online course was only one part of the work. Traffic safety education had to satisfy rules set by courts and state agencies. A student needed a practical way to learn, while the institution responsible for the requirement needed a credible way to determine whether that student had completed it.
More than two decades later, education technology is facing a new version of the same question. Artificial intelligence can generate feedback, sort records, answer routine questions, and identify patterns in student activity. The temptation is to treat every step that can be automated as a step that should be. Experience in regulated learning suggests a more useful standard: automate the repeatable work, but make responsibility for consequential decisions unmistakably human.
Putting a course online is not the same as designing online education
Early online programs could not succeed by simply posting classroom materials on a screen. Course content, student progress, assessments, identity-related checks and completion records had to function as parts of one process. The precise requirements differ by state and program, which means the system must be designed around applicable rules rather than a single universal checklist.
For students, the visible experience should be simple: they know what to do next, how much work remains and where to get help. Behind that simplicity, providers need dependable records and a clear path for resolving problems. If a student loses a connection, disputes a status or cannot access a lesson, a useful system must do more than display an error message.
This distinction applies well beyond traffic school. The U.S. Department of Education’s 2024 National Educational Technology Plan distinguishes between access to technology, educators’ capacity to design learning experiences and the way students actually use digital tools. Providing a login is not the same as creating a meaningful learning experience.
Standardize the process, not the judgment
Some educational tasks are well suited to automation. A platform can present material in a required sequence, record quiz attempts, flag missing steps, notify a learner about an incomplete assignment, and assemble records for review. Automating those tasks can reduce administrative delays and make the rules more consistent.
Other situations demand context. A student who repeatedly fails an assessment may need an explanation, a different instructional approach, or help with an accessibility barrier. An unusual activity pattern might deserve review, but it should not automatically become an accusation of misconduct. A completion dispute may involve a technical interruption rather than an attempt to evade requirements.
The decision is not between technology and people. It is about choosing the right responsibility for each. Systems can identify that something requires attention. People must retain an appropriate role in determining what it means, particularly when a decision affects a learner’s standing, certification, or opportunity.
A completion record answers one question, not every question
In compliance-based education, records matter. Courts, licensing authorities, providers and learners may all need to know whether a requirement was met. That makes timestamps, assessment outcomes, documented steps, and certificates operationally important.
But a recorded completion should not be confused with a complete picture of learning. Finishing a module shows that the system recorded the required action. It does not, by itself, establish whether a learner can apply the material in an unfamiliar situation. A quiz score adds information, but even that is only one measure.
EdTech developers should ask what evidence matches the educational goal. If the goal is remembering a rule, a knowledge check may help. If the goal is applying that rule, scenario-based questions and opportunities to correct misunderstandings may be more useful. For skills that require physical performance, digital coursework cannot automatically substitute for the separate practical instruction or assessment a program may require.
The better question is not simply “Did the learner finish?” It is “What can we reasonably conclude from the evidence we collected, and what remains unknown?”
Build accountability into the workflow
Accountability becomes difficult when it is added after a dispute. A more reliable approach is to establish, during product design, what the platform records, who can correct an error, when a learner can request review, and who has the authority to approve an exception.
That means documenting the version of the course a learner completed, preserving relevant assessment and completion information, limiting access to personal data, and making changes traceable. It also means explaining the process in language that students can understand. A system can be technically compliant and still be confusing to the person using it.
AI makes these choices more urgent. An automated recommendation may sound authoritative even when it is incomplete or wrong. UNESCO’s guidance on generative AI in education calls for a human-centered approach that protects human agency and supports meaningful educational use. The National Institute of Standards and Technology’s AI Risk Management Framework likewise emphasizes governance, transparency and the ability to understand and manage risks.
Those principles translate into practical design questions: Can a person explain why a flag appeared? Is there a way to challenge an incorrect result? Can the institution review the underlying record? Do students know when they are interacting with an AI tool? Does the system collect only the information necessary for its purpose?
The lesson from regulated education is restraint with a purpose
When the COVID-19 pandemic interrupted in-person instruction, schools and driving education providers needed ways to continue teaching. CyberActive, which had operated online courses for years, worked with traditional driving schools to offer online programs through partnerships. The episode reinforced a lesson I had learned earlier: reliable digital infrastructure matters most when the real world does not follow the plan.
That reliability does not come from removing educators or administrators. It comes from letting technology handle predictable steps while leaving room for support, explanation, and human intervention. The same principle should guide today’s AI-enabled learning products.
Before automating a new educational decision, product teams should be able to answer three questions. What evidence does the system actually have? What happens when that evidence is incomplete or incorrect? And which person remains accountable for the outcome?
If those answers are unclear, faster automation may only make mistakes harder to notice and harder to reverse. If they are clear, automation can give educators more time for the work that requires judgment and give learners a process they can understand and trust.
That, to me, is the opportunity for the next generation of EdTech: not a classroom with fewer people in it, but a better-designed learning system in which technology serves people and responsibility never disappears.
About the author
Sharon Ourian is the founder and CEO of CyberActive, an education technology company established in 2000 that provides online traffic safety, driver education, and other regulated training programs. A graduate of UCLA School of Law, she has spent more than two decades building technology for compliance-based learning and partnerships with educational institutions and public agencies.
Sources for editorial fact-checking
Sharon Ourian, CyberActive leadership profile: https://www.cyberactiveonline.com/leadership/sharon-ourian
Shama Hyder, “License To Thrive,” Forbes (Jan. 26, 2021): https://www.forbes.com/sites/shamahyder/2021/01/26/license-to-thrive-how-this-female-ceo-invented-online-drivers-ed/
MeiMei Fox, “How CyberActive Saved Dozens Of Companies,” Forbes (July 22, 2021): https://www.forbes.com/sites/meimeifox/2021/07/22/how-cyberactive-saved-dozens-of-companies-from-bankruptcy-during-the-covid-19-pandemic/
U.S. Department of Education, 2024 National Educational Technology Plan: https://eric.ed.gov/?id=ED641164
UNESCO, Guidance for Generative AI in Education and Research (2023): https://www.unesco.org/en/articles/guidance-generative-ai-education-and-research
NIST, Artificial Intelligence Risk Management Framework 1.0: https://www.nist.gov/itl/ai-risk-management-framework
