Leading Through AI Uncertainty: Four Guiding Principles for Education Innovators
Artificial intelligence is advancing faster than schools, colleges, and education organizations can realistically adapt. As educators, policymakers, and EdTech leaders navigate questions about academic integrity, student learning, privacy, and responsible implementation, many are looking for practical guidance rather than simple rules or blanket AI bans.
Drawing on emerging research and lessons from educational innovation, this article proposes four guiding principles to help education innovators make more thoughtful, evidence-informed decisions while AI technologies and our understanding of their educational impact continue to evolve.
For a reflection on the larger questions AI is forcing educators to confront, see our companion post: Rethinking Teaching and Learning in an AI-Enabled World: Questions Every Education Innovator Should Be Asking.
Overview
AI adoption is accelerating across education, with both student and teacher use increasing according to numerous polls and studies.
At the same time, concerns about the relationship between technology use and learning outcomes are becoming more pronounced in both K-12 and higher education institutions.
Increased usage vs. educational impact
A growing body of surveys points to rapid increases in AI adoption as educators explore ways to save time, strengthen instruction, and support student learning.
Unlike previous waves of EdTech innovation, today's AI tools are widely accessible to the general public. Powerful applications can be downloaded in minutes and are being used with varying degrees of sophistication by educators and learners alike.
Combined with AI's rapidly expanding language, search, reasoning, and content-generation capabilities, this accessibility marks a technological shift that extends well beyond previous generations of educational technology.
AI tools are also fundamentally different from most instructional technologies. They are general-purpose cognitive tools that are often as accessible to learners as they are to educators, with many students already using them in ways schools are only beginning to understand.
Evidence is also emerging that AI can dramatically improve performance on some assignments and assessments, raising difficult questions about what student work actually demonstrates and how learning should be evaluated.
With adoption occurring so rapidly, AI is reshaping teaching and learning faster than schools can develop evidence-informed policies, implementation strategies, or research-based guidance.
At the same time, AI is arriving amid growing public concern about the educational impacts of digital technologies more broadly. Questions about screen time, social media, smartphones, and digital learning have contributed to a broader skepticism that increasingly influences how AI is perceived.
This combination of accelerating innovation and growing public caution has intensified calls for evidence-informed guidance about which AI applications genuinely strengthen learning, which have limited impact, and which may undermine educational goals.
Research alone, however, cannot fully resolve these questions. Given the complexity and diversity of educational settings, findings that show promise in one context may not readily translate to others. As a result, educators, policymakers, and education innovators continue to navigate a landscape marked by uncertainty, evolving evidence, shifting policies, and ongoing debate.
Initial AI and Learning Research: Limitations and Skepticism
Current research suggests what we would expect, that AI may be driving “disruption” in learning environments, but it isn’t revolutionizing learning, at least not yet.
There are concerns about digital shortcuts (cognitive offloading), about digital reading and notetaking vs. pen, paper, book learning and the impact on retention, and about how digital distractions compete with the need for an attentive focus on learning…
As a result, we now see a negative trend line developing in terms of public perceptions of education technology broadly defined — slowing adoptions and emerging policy shifts that ban and prohibit certain kinds of student AI use and access.
Here’s an overview of some key limitations and concerns resulting in a mixed scorecard for AI adoption when it comes to improving instruction:
Concerns about privacy, safety, and predictability
For larger-scale EdTech models — with or without AI capabilities — surveys suggest that concerns about student privacy can be a key obstacle to adoption.
Lack of convincing outcome data
While AI is gaining reach and familiarity and showing promise as a learning aide, as a time saver for lesson planning (including personalization), and as a tool for tailored learning support and tutoring support, a large-scale study of AI for K-12 learning in Michigan found that factors such as teacher responsiveness, teaching experience, and lesson design deliver impacts that far outweigh those attributed to growing AI usage.
Lack of consistent outcomes across different subject matter
The same study also found that the impact of AI on learning is uneven, with less evidence for successful implementation in subject areas such as math and language arts, according to one large study.
Concerns about equal access
Education outcomes may be tainted by growing inequities among students when it comes to access to AI-assisted learning. Studies signal a growing gap between students who use AI for active, deep learning and those who use it passively for rote completion of assignments. Similar inequities may afflict the institution as well, with some schools having the resources, expertise, and infrastructure at the ready while others have much catching up to do.
Growing Questions About the Evidence Base Add to the Backlash
Growing concerns about research literature and AI mean research is resulting in more questions than answers. Studies extolling the benefits of AI for learning are drawing increased scrutiny, with some widely cited studies being retracted altogether.
Some researchers express concerns about a research landscape driven by market forces not empirical methods and independent review:
We are currently witnessing one of the largest and fastest deployments of private capital in modern economic history with US cloud and AI infrastructure providers on track to spend nearly $700 billion in 2026 (Patience, 2026). The investment has been accompanied by an intensive effort to frame AI adoption as an urgent educational priority…. The effects of this pressure also reach academic literature.
One group of researchers found that “loosely defined treatments, mismatched or opaque controls, and outcome measures with unclear links to durable learning obscure causal claims... Observed gains cannot…be confidently attributed to ChatGPT, and meta‐analytics effect sizes may over‐ or understate its benefits.”
Another researcher undertook a comprehensive audit of AI research and came to the following broad finding:
This [research literature] review employs a forensic methodological audit of 14 meta-analyses making broad claims about the impact of AI on education… The audit found that none of the examined meta-analyses provided a valid basis for the claims they advanced: none had a coherent construct, none sufficiently assessed publication bias, and all had severe heterogeneity. Statistics were misapplied, miscalculated, and misinterpreted.
A Growing Public Backlash
The scrutiny being given to existing research on AI and learning comes on the coattails of what is already a broad wave of doubt about AI deployment. According to a Data for Progress study, public opinions of AI are “closely divided, with 48% of voters viewing the technology favorably and 46% unfavorably.”
High-profile initiatives ranging from device-free classroom proposals at Yale to expanding smartphone restrictions in K–12 schools reflect growing concern about misdirected attention, waning cognitive engagement, and about how to manage technology access and turn technology capabilities into wins for learning.
Harnessing the Potential for EdTech-Driven Learning Gains in an Era of Skepticism
Even if educators were to have a strong evidence base to draw on, AI is still too new for those findings to instill strong conviction. After all, educators still have to contend with the dynamics of local implementation — predicting what will work in their setting and circumstances — published trials and research findings notwithstanding.
And, whatever claims academic researchers may make about AI in education, when overall statistical declines in nationwide assessment scores happen in tandem with increasing EdTech adoptions, it’s a scenario that inevitably feeds skepticism.
As such, product teams, grant seekers, and change leaders in schools will almost certainly be challenged to demonstrate that AI-enabled products, investments, and adoptions offer more than promises, making it imperative to:
Explain how AI tools will align key organizational goals and practices around teaching and learning
Demonstrate what guardrails will be in place to ensure privacy for example
Present concrete evidence that technology designs drive learning as opposed to simply streamlining students’ learning effort
Navigating AI in a Period of Deep Uncertainty
Key to success amid the skepticism and fast-changing and speculative AI adoption and implementation landscape is to maintain a strong focus on key instructional goals and methods.
In the short run at least… Organizations that align AI initiatives with clearly defined learning outcomes, support educators through sustained implementation efforts, and continually evaluate whether these tools are contributing to meaningful improvements in teaching and learning are the most likely to experience and be able to celebrate learning successes.
For managing AIED decisions in this climate, here are four overarching principles that can ensure greater success.
Guiding Principle #1
Stay Focused
Because emerging technologies come cloaked in marketing that may promise easy fixes for stubborn learning gaps — while making it easier to dodge the complexities of cultivating human capital and other institutional initiatives — it can be tempting to give in to a solutions-first approach and hope it works out…
The challenge here is similar to the challenges students can face when surrounded by screens and internet resources — a quick win in exchange for leveraging a sustained and clear focus. This fragmented approach weakens both teaching and learning.
The antidote for educators is to establish and maintain a strong focus on the most relevant and enduring learning objectives across the deliberation, planning, implementation, and reflection stages.
Key questions for guiding AI adoptions for instruction include:
What learner needs are we trying to address?
What instructional challenge are we trying to solve?
What outcome are we hoping to improve?
What approaches and instructional strategies and formats can best help us meet these challenges and goals?
What tools or technology capabilities can support these approaches?
How do we get everyone on board and also safeguard against innovation sprawl?
How do we support successful implementation (including learning from and improving our practices and outcomes and consistently areas for building organizational capacity)?
Answers to questions like these along with the kinds of design pillars we’ve shared in other EdPro blog posts, about digital classrooms, backwards planning and Bloom’s taxonomy, and getting better results from EdTech investments, can all help align adoptions within a coherent instructional approach.
Guiding Principle #2
Prioritize Coherent Learning Design Over Fragmented Tool Adoption
One recurring lesson from digital learning research is that educational impact rarely emerges from isolated tools operating independently of broader instructional systems.
Organizations often adopt AI to streamline discrete tasks: lesson planning, grading, tutoring, study materials…
While these uses may provide value, the implementation efforts, conversations, and perceived adoption wins may actually distract educators from deeper questions of instructional practice that make deeper connections across a spectrum of learning components, such as core educational values and beliefs, aligning learning goals with a forward-looking focus on social challenges and the workforce landscape, and consideration of current students’ unique needs, challenges, and interests.
Rather than asking, "Where can we add AI?"
Education innovators should ask: How can we strengthen our existing learning model or instructional strategy, and what gaps, challenges, or aspirational goals might we be able to address with specific kinds of AI capabilities? Which capabilities?
This shift from fragmented adoption takes the emphasis off of trying out AI tools as they surface.
Instead, educators engage in professional learning about AI capabilities as a foundation for envisioning how AI will help solve instructional challenges or amplify instructional practices in their local context.
Ideally, EdTech product teams will inform their work through collaboration with frontline educators in localized settings so educators and tech leads work together to uncover new approaches to harnessing AI for learning impact and for mutual learning in pursuit of a shared educational purpose and under lived constraints.
To reflect further on the bigger questions educators need to be asking as an AI-enabled world takes shape, see ourcompanion post: Rethinking Teaching and Learning in an AI-Enabled World: Questions Every Education Innovator Should Be Asking.
Guiding Principle #3
Cultivate Curiosity and a Thirst for Organizational Learning
Improving instruction is not going to be the result of any static implementation plan. The “rollout” needs to be rooted in shared goals and aspirations for teaching and learning This means that improving learning requires a scaffolded approach and decision making that’s collaborative but also informed by learning and data, and pursued with agility and nuanced refinements: pilot — learn — refine — scale…
Key features of strong professional learning ethos include:
experimentation and piloting that provide data and feedback
reflection and iterative improvements guided by data and established priorities
iterative improvements and targeted capacity building that set the stage for larger scaling and adoption
curiosity-driven research and exploration that nourishes aspirational goals and professional creativity within a disciplined approach to change leadership
Organizations should expect AI initiatives to evolve over time as evidence emerges, user needs become clearer, and practices mature, but as these shifts occur, it’s important to revisit core learning goals and instructional designs, ward off tool and initiative bloat, and sustain positive but disciplined accountability and evaluation frameworks.
Guiding Principle #4
Invest in Human Capacity Alongside Technology
Research consistently suggests that educator capacity remains one of the most important determinants of technology effectiveness.
Professional learning is most effective when it is:
sustained rather than episodic
collaborative rather than isolated
context-specific rather than generic
embedded within implementation efforts rather than delivered separately
Faculty learning communities, peer support models, coaching structures, and just-in-time learning opportunities frequently outperform one-time workshops because they help build accountability and capacity across teams and help educators share knowledge and problem-solve across a larger time span of practice.
AI implementation should therefore be viewed as both a proscribed technology initiative within a superseding professional learning initiative.
AI-Supported Instruction: Promising Avenues
Current implementation discussions and areas of design focus point to some of the ways AI applications are most likely to help the learning process in the short term, such as:
instructional feedback
skill and knowledge scaffolding
time management for teachers
framing learning and increasing access (rather than replacing core core cognitive work)
With this in mind, here are some of the key opportunities that AI-in-education teams and education leaders can focus on as they approach AI adoption.
Opportunity Area #1: Strengthening Educator Capacity
Potential AI contributions
lesson planning
resource development
tailored instruction and/or differentiated lesson planning
progress monitoring/formative assessment design
Examples from recent research
Teachers enhance effectiveness through AI-assisted subject matter learning.
Teachers streamline rote lesson planning tasks — to invest more time in high-quality and engaging lesson designs without added workload.
AI-assisted tutoring capabilities help teachers support all learners’ individual learning needs more consistently (opportunity 2 below).
Teachers use AI to design and generate effectively targeted formative assessment tasks or activities.
Potential pitfalls
Teachers use a AI-generated materials indiscriminately, resulting in highly passive and/or undifferentiated forms of learning.
After off-loading one-to-one learning feedback and support an AI tool, teachers no longer assess for accessibility or check for understanding between more formal formative and summative assessments.
Teachers use AI to generate rote question and answer assessments that fail to engage higher order thinking skills and build deeper subject matter competencies and mastery.
Opportunity Area #2: Providing Timely Feedback and Guided Practice
Potential AI contributions
structured writing feedback to focus and prompt (not replace) cognitive effort
generating background information and source data for applied comparison, reasoning, and or analysis activities
helping students identify knowledge, skill, or process gaps for remediation and review
Examples from recent research
improved revision behavior
stronger mathematical proof performance
increased motivation and engagement
Potential pitfalls
AI shortcuts limiting learning challenges
Lack of access or skill in using AI for relevant feedback
Lack of self-direction makes remediation tools ineffective for some learners
Opportunity Area #3: Expanding Personalized Learning Support
Potential AI contributions
tutoring
scaffolding
differentiated practice
guided support
Examples from recent research
AI-guided tutoring outcomes
using AI for step-by-step reasoning support vs. answer generation
Potential pitfalls
Teachers misgroup students or use improper or overly rigid grouping criteria
Differentiation results in ineffective or poorly constructed learning tasks
Scaffolding lacks coherence
Opportunity Area #4:Prompting Thinking Rather Than Replacing It
Potential AI contributions
Access to wide-ranging source materials for enrichment, for constructing contextualized arguments, solutions, analyses…
Open-ended feedback to guide self-correction, revision, critical thinking…
Examples from recent research
differentiated critical feedback
individualized knowledge hints and prompts
questioning strategies
Potential pitfalls
answer generation
unstructured use
AI replacing core cognitive work
Guidelines At-a-Glance
Before investing in, implementing, or scaling an AI-enabled solution, work through the following questions. Doing so will help you distinguish the building blocks of deeper approach to improving learning from approaches that rely primarily on an isolated technology, tool, or tech feature.
Final Thoughts
Systems leadership expert and seasoned education reform leader Michael Fullan encouraged innovators across the education landscape to approach school culture with an inclusive approach and a strong learning mindset.
With so many questions still unanswered and with such powerful and consequential technology innovations evolving so quickly, Fullan’s call for education leaders to build strong learning cultures is more urgent than ever.
With the pace of disruption accelerating, education technology teams, AI designers, and education leaders and instructional coaches and researchers should learn together and lead together with a purpose-driven approach.
In the shorter term the future of AI in education will not be determined by how quickly organizations adopt it, but by how effectively they connect AI-enabled innovation to meaningful learning, teaching, and educational outcomes.
For Additional Reading and Insights
Understanding AI Adoption in Schools
Stanford SCALE — AI's Growing Place in K–12 Classrooms
An overview of how teachers are incorporating AI into classroom practice, the instructional tasks where AI appears most useful, and the practical questions educators continue to wrestle with as adoption accelerates.
Education Week — More Teachers Are Using AI in Their Classrooms. Here's Why.
A practitioner-focused look at how teachers are using AI to support lesson planning, feedback, differentiation, and classroom efficiency, while highlighting ongoing implementation challenges.
The 74 — Case Study: How Two Teachers Use AI Behind the Scenes to Build Lessons and Save Time
A practical case study illustrating how AI can reduce routine planning tasks while allowing teachers to invest more time in instruction and student engagement.
This national public opinion survey examines how Americans view artificial intelligence, revealing that public attitudes toward AI remain deeply divided and vary considerably across demographic groups and levels of AI familiarity. For education leaders, EdTech organizations, and grant seekers, the findings provide valuable context for understanding why AI initiatives increasingly require transparent communication, strong evidence of educational value, and thoughtful attention to trust, privacy, and responsible implementation—not simply new technological capabilities.
Research on Learning Outcomes and Student Use
Michigan Virtual — Artificial Intelligence and Student Usage in Online Learning
One of the largest longitudinal studies examining AI usage patterns, student achievement, and learner perceptions in K–12 virtual education. Particularly valuable for understanding where AI appears promising—and where educator expertise and instructional design remain more influential than technology alone.
University of Kansas — AI and Machine Learning in K–12 Education
A concise overview of current educational applications of AI, including personalization, adaptive learning, and instructional support, with discussion of both opportunities and limitations.
Critical Perspectives on the AI Evidence Base
"ChatGPT in Education: An Effect in Search of a Cause" (2025)
This scholarly review examines the rapidly expanding body of research on ChatGPT in education and argues that many reported learning gains should be interpreted cautiously. The authors contend that inconsistent study designs, weak causal measures, and varying definitions of AI-assisted learning make it difficult to determine whether observed improvements are truly attributable to AI. For educators and EdTech developers, the paper serves as a valuable reminder that strong instructional claims require equally strong evidence.
"Presumed Effective: The Manufacturing of an Evidence Base for AI-in-Education Through Flawed Meta-Analysis" (2026)
This methodological critique reviews several widely cited meta-analyses on AI in education and raises important questions about how evidence is synthesized and communicated. Rather than challenging the potential of AI itself, the authors argue for more rigorous research methods, clearer outcome measures, and greater caution when drawing broad conclusions about educational effectiveness. The paper is particularly valuable for education leaders, grant writers, and product teams seeking to distinguish promising innovations from claims that may outpace the available evidence.
Related EdPro Articles
A reflection on the kinds of questions education innovators need to be asking today to be better positioned to design, fund, and implement meaningful education innovation in tomorrow’s increasingly AI-enabled world.
From Promise to Impact: A Practical Framework for More Effective EdTech Adoption and Integration
A practical framework for designing technology initiatives that strengthen teaching and learning through coherent instructional design and sustained implementation.
The Science of Learning: What Every Education Innovator Should Know
An overview of enduring learning science principles that can help guide instructional design, curriculum development, and educational technology decisions.
Leading the Change You Want To See
Discoverwhy The Practice of Adaptive Leadership (by Heifetz, Grashow, and Linsky) offers an authoritative and practical field guide for school leaders seeking to build confidence leading deeper change.
