Synthesia survey finds training teams cite AI speed gains far more often than better job performance

Training teams are finding it easier to point to faster production than better workplace results from AI. In Synthesia’s 2027 learning and development survey, 75% of those answering an impact question named faster content creation as a clear gain; just 11% named improved on-the-job performance. The gap sits alongside limited measurement of what employees actually do differently after training.

The AI in Learning & Development Report 2027 draws on 831 professionals across more than 25 countries. Its findings describe practitioners’ reported experiences, rather than a direct test of AI’s effect on employee performance. Questions about impact and spending were optional, so some findings reflect smaller groups than the full sample.

This is not simply a picture of teams waiting to start. Some 37% of respondents said AI was widely adopted across their work, and another 36% used it in specific workflows. A further 24% were still exploring or piloting it. The distinction is how deeply AI has entered the work—and what teams can demonstrate afterward.

Counting attendance is easier than tracking improvement

Among 252 respondents who used AI at the evaluation stage and answered impact questions, 72% consistently measured participation and 70% measured learner satisfaction. Only 40% consistently measured on-the-job performance, while 29% tracked time to proficiency. Their evaluation was more likely to capture attendance and reactions than whether employees became more capable at work.

Confidence did not close that measurement gap. Of respondents who identified an AI impact, 60% were confident or very confident that AI caused the improvement. Yet confident respondents consistently measured on-the-job performance at roughly the same rate as everyone else: 35%, compared with 37%. Those figures cover a broader group than the evaluation-stage subset.

Asked for the biggest barrier to proving AI’s impact, 53% pointed to organizational constraints. The obstacles were not just about the technology:

• 22% cited a lack of agreed success measures.
• 17% cited access to data or measurement tools.
• 14% cited a lack of alignment among stakeholders.

Working faster is a trap if speed is the only thing you can show. When speed is the only result a team can point to, the business hears that you can do more with less.

Kevin Alster, strategic advisor at Synthesia

Personalization remains an ambition

Content generation dominated current uses, at 87% of respondents using AI, followed by video creation at 73%. Personalized guidance or tutoring reached only 29%. Still, 35% of respondents expected personalized learning experiences to deliver AI’s greatest value, compared with 42% who chose faster content production.

Synthesia offers incentives as one explanation: teams assessed on the volume of training they produce have a reason to apply AI to production. The company argues that experiments beyond volume need better performance data, visibility into AI spending and agreed measures of success. That is its interpretation of the pattern, not a cause established by the survey.

Bigger budgets, incomplete visibility

The optional spending questions reveal another blind spot. Among 557 respondents answering the current-spend question, 46% did not know what share of their learning budget went to AI. Among 542 answering the outlook question, 56% expected AI-related spending to rise over the next year. Even among those unsure of current spending, 45% expected an increase.

Cost and value influenced tool selection for 59% of respondents. But data security and privacy was the most frequently cited adoption barrier, at 28%, ahead of budget at 22%. Output quality, at 12%, and system integration, at 9%, ranked lower.

A leadership push did not necessarily mean freedom to experiment. Overall, 61% needed sign-off before piloting a new AI tool. That rose to 73% for teams whose adoption began with a leadership mandate, versus 57% for teams that started from the bottom up.