What could your team achieve if a document that previously took two hours could be created in eight minutes? In this episode of Tech Talks Daily, I speak with Oskar Konstantyner, Chief Product Officer at Templafy, about the rapid adoption of AI agents across enterprise document workflows and what those productivity gains mean for knowledge workers. According to Templafy's proprietary usage data, AI agent adoption among its enterprise users grew from virtually zero in October 2025 to 53% by June 2026. Its analysis found that documents created without agents had a median completion time of two hours and an average of 5.6 hours across 16,000 sessions. With AI agents, the median fell to eight minutes and the average to 27 minutes across 14,000 sessions. The most common documents included pitch decks, company communications, sales materials, and product roadmaps. However, Oskar cautions against treating speed as the final measure of AI productivity. We discuss an accounting firm that could not respond to thousands of tenders because it lacked the capacity to create enough proposals. Faster document production could allow that business to participate in additional opportunities while applying its knowledge about what makes a winning submission. The benefit comes from increased commercial capacity and stronger results, rather than counting recovered hours alone. Oskar also explains what happens during those eight minutes. AI can locate relevant information, find approved content, recommend a presentation structure, apply previous lessons, and complete much of the production work. Humans remain responsible for original thinking, client judgment, factual accuracy, and final approval. In many cases, the agent may produce 60% to 90% of the document, but the beginning and end of the process remain human-led. The conversation also considers the growing volume of generic AI documents. A business already has approved slides, company descriptions, brand assets, legal statements, and sales messages. Regenerating all that material wastes tokens and risks inconsistency. Oskar argues that agents should determine when existing content should be reused, when rules should be applied, and when something genuinely new needs to be created. We also discuss why AI adoption improves when agents appear inside PowerPoint, Claude, OpenAI, and Copilot. Most employees are unlikely to abandon familiar workflows every time another AI application arrives. Is your company measuring AI success through minutes saved, or through the additional business those minutes make possible? Listen to the conversation and share your thoughts with me.

Ask this episode

“What did the council decide on item 42?” — answered with the speaker and the moment.

Transcript

Related