EdTech solutions has been part of organizational learning for decades. Since the first learning management systems of the late 1990s, most organizations have cycled through several iterations of selecting, procuring and adopting learning technologies. Along the way, we have developed a fairly structured implementation process. A typical journey runs from needs assessment through selection and deployment to measuring impact and benefits.
With the rise of AI solutions for learning and development, organizations are following a similar path. And while it is tempting to treat AI as another EdTech platform, it is worth asking whether AI’s path to delivery actually resembles traditional EdTech’s, or whether the structural differences between the two are significant enough to change what adoption requires.
England’s Department for Education’s Assessment of the Education Technology Market research report (June 2026) offers a useful starting point. It maps an eight-stage EdTech implementation journey, along with common barriers encountered at each stage. Building on that structure, let us consider: do the same eight stages hold up for AI, or does AI introduce barriers that are different in kind, not just in degree? To answer that question, here is a comparison table mapping both journeys side-by-side, highlighting where the barriers are similar and where AI’s structural differences change what leaders need to account for.
| Stage | EdTech Barriers | AI Barriers | Structural Difference |
| Stage 1: Identifying Needs | • Limited strategic planning • Lack of alignment between business needs and L&D • Lack of EdTech and digital literacy • Narrow stakeholder input | • Similar challenges with strategy, alignment and stakeholder input • Solutions looking for a problem • Lack of a shared understanding of what AI is and what it can do AI and market hype | EdTech’s literacy is relatively static while AI literacy is a moving target since what technology can do changes month to month. |
| Stage 2: Researching and Assessing | • Limited independent and unbiased evidence on impact and on EdTech research, particularly in corporate learning • Rapid technology change outpacing evidence | • AI platform proliferation outpacing selection and research capacity • Shortage of authentic technical expertise to interrogate claims • Model update outpacing research | EdTech evidence lags on one axis (products evolve, research catches up eventually). AI evidence lags on two axes at once, since both the tool and the underlying model shift, so research can be outdated before the tool it studied even finishes a full release cycle. |
| Stage 3: Making an Informed Decision | • Competing priorities • Budget and resources pressures • Decisions sometimes accelerated due to accountability pressures | • Governance frameworks largely absent • Decisions accelerated by competitive pressure • Unclear ownership of the decision | EdTech pressure plays out inside an existing, if imperfect, governance structure. AI pressure often plays out in a governance vacuum, since only a minority of organizations have any operational AI oversight in place at all. |
| Stage 4: Piloting and Trialing | • Limited staff time and capacity • Insufficient training for pilot users • Limited vendor support during pilots | • Same resources constraints Pilot-to-production chasm • Immature pilot evaluation frameworks • Data quality problems surfacing mid-pilot | EdTech's pilot barriers are mostly resourcing problems. AI's pilot barrier includes a structural translation failure: most organizations do not yet know how to convert a successful pilot into a scaled deployment. |
| Stage 5: Procuring and Accessing | • Complex and fragmented procurement processes • Cost • Data storage, privacy, security issues | • Same fragmentation AI’s unique issues that procurement was never built to assess (e.g. algorithmic bias, model IP, exit strategy) • Unpredictable cost | EdTech procurement is slow but operates against relatively settled frameworks (e.g. FERPA, PIPEDA, GDPR). AI procurement is being invented in real time, under active legal and political contest.1 |
| Stage 6: Deploying and Integrating | • Infrastructure limitation (hardware, bandwidth, incompatible systems) • Staff digital confidence and skill gaps | • Legacy system and data integration debt • A distinct information security and data governance risk around what data can be exposed to a third-party model • Persistent AI skills gaps | EdTech confidence gaps tend to close over time with training and exposure. Leaders report that workforce AI-readiness confidence is currently declining year over year even as adoption accelerates, so the gap is moving in the wrong direction rather than gradually resolving on its own. Both have cultural and skill barriers. However, AI adoption adds the trust and oversight layer that cannot be addressed by training alone. |
| Stage 7: Adopting and Training | • Limited leadership engagement and buy-ins • Lack of time and resources • Generic training by vendors | • Issue of trust and oversight gap with no precedent • Training follows adoption rather than preceding it • Training doesn’t account for how the system changes and evolves | Both have cultural and skill barriers. However, AI adoption adds the trust and oversight layer that cannot be addressed by training alone. |
| Stage 8: Measuring and Evaluating | • Lack of evaluation frameworks • Limited impact data and unbiased evidence from vendors • Measure adoption over impact • Lack of institutional support or policy | • Same under-investment in evaluation and lack of frameworks • A moving target since the platform being evaluated keeps changing shape mid-cycle • Limited vendor transparency (e.g. no access to raw usage data, undisclosed model changes, share only positive case studies) | Both are being under-evaluated and without rigorous and standardized frameworks and methodologies. Evaluating AI platforms has the added barrier of the systems’ constant changes and their abilities to adapt to their users, thus, making them harder to compare and evaluate. |
So, do the eight stages hold up for AI? The structure carries over; the practices built around it do not, and AI adds barriers of its own on top of the familiar EdTech ones. The practical takeaway for leaders? Pay attention that almost every activity once treated as a one-time milestone in an EdTech rollout becomes a continuous practice in an AI one. Literacy building cannot end at launch training. Governance and policies cannot be retrofitted after procurement. Evaluation cannot assume the thing being measured will hold still. Organizations that plan for these as ongoing capabilities, rather than boxes to tick on the way to deployment, will navigate the journey far better than those reusing their EdTech approach as-is.
1 Canada's proposed federal AI law, the Artificial Intelligence and Data Act, died on the order paper when Parliament was prorogued in January 2025, leaving no federal AI-specific regulation in place at all. Read more about it here – What’s Next After AIDA?
