Will a curriculum-grounded AI assistant actually save teachers time?
Sanoma has launched Sanna for a free trial in seven European countries after a two-month pilot involving more than 1,500 teachers. The company describes co-design and guardrails, but publishes no measured time saving or pupil-outcome result.
The short version
- What happened
- Sanna begins a 30-day trial on 1 October in the Netherlands, Spain, Italy, Poland, Belgium, Finland and Sweden.
- Who it affects
- Teachers may gain faster first drafts for routine materials, but must spend time checking every consequential output.
- What remains unproven
- The evidence is supplied by the company launching the product; no independent evaluation of Sanna is cited.
Keep in mind
The evidence is supplied by the company launching the product; no independent evaluation of Sanna is cited.
The 1,500-teacher pilot denominator is not broken down by country, subject, completion, usage or comparison group.
No measured time saving, content-error rate, teacher retention, pupil attainment or equity outcome is reported.
Multi-source analysis · 2 linked sources. Primary analysed source date: 1 October 2026. Earliest linked supporting source: 24 September 2026.
Read the full report · 5 min The evidence, methods, implications and limitations
The launch puts a teacher-facing assistant into seven markets
Sanoma Learning launched Sanna on 1 October with a 30-day trial in the Netherlands, Spain, Italy, Poland, Belgium, Finland and Sweden. The product is designed for teachers rather than pupils. Sanoma says it can help prepare lessons, build assignments and assessments, adapt material for different learning levels, draft feedback and create presentations. Users can work through guided tools or an AI chat interface.
The claimed difference from a general chatbot is grounding in Sanoma's local learning materials and national curricula. That could make generated tasks easier to align with what a class is expected to learn, while reducing the temptation to treat a fluent but generic response as a finished lesson. Yet grounding is not the same as correctness. A curriculum source can be incomplete, a prompt can be misunderstood and a generated exercise can still contain factual or pedagogical errors.[1]
The pilot denominator is substantial, but the outcome is missing
Sanoma says initial testing in 2025 was followed by a two-month pilot in 2026 involving more than 1,500 teachers across Europe. It says feedback, studies and interviews informed development. That is useful evidence of product exposure and co-design, but the announcement does not disclose how teachers were recruited, how many came from each country or subject, how often they used the assistant, how many completed the pilot or what comparison was used.
Most importantly, the release gives no measured reduction in planning or marking time, no error rate and no pupil outcome. It therefore cannot establish the headline promise that the tool reduces workload. A before-and-after time diary, task-level accuracy audit and comparison group would allow readers to distinguish a genuinely faster workflow from work merely shifted into checking AI output. Retention after the free trial would add a practical, if still non-causal, signal of usefulness.[1]
The survey measures demand, not product performance
The launch cites Sanoma's 2026 European Teacher Survey, which received more than 20,000 responses. Depending on the market, 75% to 93% of respondents said AI tools used in education should be designed for educational purposes rather than adapted from general-purpose applications. That result supports the product strategy, but it does not show that Sanna meets the preference or improves learning. Survey wording, country samples and response patterns also matter when interpreting the range.
The OECD's recent Digital Education Outlook offers broader context: generative AI can be useful when deployed within explicit educational goals and sound teaching practice. It also warns against treating access to a tool as an outcome. Schools need to decide which tasks should be assisted, what evidence a teacher must inspect and how to preserve student thinking. The OECD report is not an evaluation of Sanna and should not be read as an endorsement.[1][2]
Human approval is necessary—and creates new work
Sanoma says teachers remain in control and can review, edit, adapt and approve all generated content. It also describes educational guardrails, user-interface choices intended to limit over-reliance and data processing within Europe. Those are sensible design commitments. Their effectiveness needs testing across languages, subjects, age groups and accessibility needs, including how the system handles unsafe requests, bias, copyrighted material and confidently wrong answers.
Review is not free. A teacher must know enough to detect a subtle misconception, unsuitable reading level or misleading assessment item. Schools should count checking time as part of the workload calculation and avoid pressuring staff to accept generated material because an assistant exists. Procurement should also specify who owns prompts and outputs, what data is retained, whether teacher interactions train future models and how incidents can be reported and corrected.[1]
What would justify wider adoption
The assessment would strengthen with independent, pre-registered evaluation across the seven markets: teacher time measured by task, blinded review of content quality, error and bias rates, accessibility checks, and pupil learning or engagement outcomes where the tool changes classroom materials. Results should be separated by language, subject, school context and teacher experience rather than averaged into one success score.
It would weaken if monitoring finds frequent curriculum errors, longer checking time, low continued use, privacy incidents or benefits concentrated in already well-resourced schools. Sanna's launch is material because it places a curriculum-linked assistant inside established European education ecosystems and exposes a large pilot denominator. The evidence currently supports a product launch and co-design process—not a proven reduction in workload or improvement in learning.[1][2]