The Schools Getting AI Wrong Are Starting With Their Best Teachers
Most school leaders pilot AI with their strongest, most tech-comfortable teachers. That instinct is understandable. It feels safe. It produces clean results. The problem is it also produces a distorted picture of what AI can do across a building.
The OECD Digital Education Outlook 2026 found something that cuts against this pattern directly: inexperienced tutors can raise the quality of their tutoring and improve student learning outcomes by using GenAI tools. That is not a vendor claim or a conference talking point. That is published OECD research, based on evidence about where GenAI produces measurable instructional benefit.
The implication is worth sitting with. The staff who gain the most from AI-assisted tutoring are the ones who do not yet have deep instructional experience to draw on. Paraprofessionals. Tutoring center staff. First and second-year teachers running small groups with limited feedback. These are often the people working most closely with your highest-need students, the ones for whom instructional quality has the most consequence.
Why the Pilot-With-Your-Best Strategy Backfires
Strong teachers already have internalized frameworks for what good instruction looks like. They scaffold, they adjust, they read a room. When you give them an AI tool, they can evaluate it critically, integrate it selectively, and mostly work around its limitations. Their outcomes look good in a pilot. But their outcomes were already good.
The teachers and support staff who interact with struggling students in pull-out groups, after-school programs, or one-on-one tutoring sessions often have far less instructional infrastructure behind them. Less training. Less feedback. Less experience recognizing when a student is confused versus disengaged. AI tools, according to the OECD, can help close that gap. That is where the instructional floor rises.
Separately, survey data reported by Boterview and Forbes Advisor found that 55% of teachers report better student outcomes when using AI tools. That is a majority, but it also means a large share see no difference or worse results, which is exactly what you would expect if rollout strategy, staff readiness, and use-case fit are not being considered carefully.
A Sequencing Argument, Not a Technology Argument
What the OECD finding gives school leaders is a sequencing argument. You do not need every teacher fully AI-proficient before you see returns. You need to identify where instructional support is thinnest and put the right tools there first.
That reframes the professional development question entirely. Blanket rollouts that push AI tools at experienced teachers who were not asking for them tend to produce resistance, or at best indifference. Targeted deployment, starting with the staff who have the most to gain and the students who have the most at stake, is a more honest use of limited PD time and budget.
This is a design decision as much as a policy one. It requires someone to map instructional capacity across the building, identify the roles with the least support, and match tool selection to those specific gaps. That work is not automatic. It requires deliberate planning before any tool goes live.
What the OECD Says About Teacher Expertise
There is a second finding from the OECD Digital Education Outlook 2026 that tends to get buried in the conversation about AI adoption. The report makes the case that teacher expertise should be integrated into the design process. Not the testing process. The design process.
That distinction matters. Most schools bring teachers in after decisions have already been made. Someone selected the tool, someone negotiated the contract, someone scheduled the training. Teachers arrive to be trained on something they had no hand in shaping. The OECD finding suggests that schools likely to see the best outcomes are the ones where experienced teachers have a real role in deciding how tools get used, what guardrails exist, and which use cases are appropriate for their context.
There is a practical version of this. Before rolling out any AI tool to paraprofessionals or newer staff, bring your strongest instructional voices into the room. Let them define the use cases. Let them identify the risks. Then use what they build as the framework for extending access to the staff who need the most support. That sequencing does not just produce better outcomes. It also protects the credibility of the initiative with the people who matter most to its long-term adoption.
On the Concern That AI Weakens Instruction
The worry that AI reduces the human quality of teaching is real and worth taking seriously. The OECD finding does not dismiss it. What it does is draw a sharper line between two different claims that often get conflated.
One claim is that AI replaces good teaching. The other is that AI helps staff who are not yet strong teachers become more effective. These are not the same claim, and treating them as equivalent is how the conversation stays stuck.
The evidence from the OECD supports the second claim, not the first. That is worth being precise about, especially when you are making the case to a skeptical board or a veteran department head. The argument is not that AI is good for teaching in general. The argument is that AI raises the instructional floor for staff who currently have the least support, in contexts where student need is often the highest.
That is a more defensible position, and a more honest one.
What School Leaders Can Act On Now
Start by mapping instructional support across your building. Where are the thinnest staffing situations? Which roles have the least PD, the least feedback, the least access to instructional coaching? Those are your highest- starting points for AI tool deployment.
Then bring your strongest instructional voices into decisions about tool selection and use-case design before anything launches. Not as approvers, but as architects. The guardrails they set will shape how the tools get used by everyone else.
For leaders looking to build the scaffolding for this kind of rollout, Microsoft's AI Skills Navigator for Educators is one example of infrastructure designed to support teacher-facing AI adoption. And Panorama Education's Solara, which received the 2026 EdTech Trendsetter Award, reflects the broader direction the field is moving toward in terms of data-informed instructional support.
The floor goes up. That is what the data says. What schools do with that is still a human decision, and it starts with being honest about where the floor is right now.
If you want a structured way to think through where your school or district stands before committing to a rollout, the free AI Readiness Assessment for Education is a good starting point.
We publish this kind of analysis every two weeks, so if it was useful, subscribing to the Archificials newsletter will keep it coming.
If you want to walk through the sequencing framework with your admin team, you can schedule a free demo consultation at whatever pace works for you.
FAQ
Does AI improve student outcomes, or is that just vendor marketing?
The OECD Digital Education Outlook 2026 found that inexperienced tutors can raise the quality of their tutoring and improve student learning outcomes by using GenAI tools. This is published research, not a vendor claim. Separately, survey data from Boterview and Forbes Advisor found that 55% of teachers report better student outcomes when using AI tools. The OECD finding is specific: the benefit is most measurable for staff who lack deep instructional experience, which is a more precise claim than general assertions that AI helps everyone equally. The research points to a real effect in a specific context, not a blanket improvement across all uses.
Which teachers benefit most from AI tools in the classroom?
According to the OECD Digital Education Outlook 2026, the staff who gain the most from GenAI-assisted tutoring are those who do not yet have deep instructional experience. This includes paraprofessionals, tutoring center staff, and early-career teachers working in small-group or one-on-one settings. Experienced teachers already have internalized instructional frameworks, which means AI tools produce less incremental gain for them. The OECD research suggests that piloting AI exclusively with strong, tech-comfortable teachers gives a skewed picture of what AI can do for the broader staff. The highest- deployment targets the roles with the least existing instructional support.
How should schools sequence AI rollout across teaching staff?
The OECD Digital Education Outlook 2026 supports a sequencing approach where school leaders first identify where instructional support is thinnest, then deploy AI tools to those specific roles. Before rollout, the OECD recommends integrating teacher expertise into the design process, not just the testing phase. This means bringing experienced instructional voices into early decisions about tool selection and use cases so they can shape how tools get used by less experienced staff. Blanket rollouts that push tools at all staff simultaneously tend to overwhelm experienced teachers and under-serve the staff who would benefit most. A targeted, capacity-mapped approach is a more efficient use of professional development time and budget.
Does using AI in education reduce the quality of teaching or critical thinking?
The OECD Digital Education Outlook 2026 does not argue that AI replaces good teaching. Its finding is that AI helps staff who are not yet strong teachers become more effective, which is a distinct claim. These two positions are different, and treating them as interchangeable keeps the conversation stuck. The concern that AI weakens instruction is legitimate, but the OECD evidence addresses a specific context: less experienced staff using AI in tutoring situations. The research supports AI raising the instructional floor for those staff, not replacing the judgment of experienced educators.
What role should experienced teachers play in AI adoption at schools?
The OECD Digital Education Outlook 2026 makes the case that teacher expertise should be integrated into the design process for AI tools, not introduced after key decisions have already been made. This means experienced teachers should have a real role in shaping how tools get used, which use cases are appropriate, and what guardrails are necessary for their specific context. Schools where experienced teachers help set the parameters for AI use are more likely to see strong outcomes when those tools are extended to less experienced staff. Bringing strong instructional voices in early also builds credibility for the initiative and reduces resistance during broader rollout.
