[Event Report] From 'Creating' to 'Questioning': Frontline Engineers Discuss Next-Generation Work Styles and Management in Collaboration with AI
DXHR Inc. hosted an AI utilization study session at its co-working space "NeuroHub." Frontline engineers and entrepreneurs gathered to discuss next-generation work styles and management in collaboration with AI, sharing practical insights and business implications.
📋 Article Processing Timeline
- 📰 Published: May 1, 2026 at 23:00
- 🔍 Collected: May 1, 2026 at 14:31
- 🤖 AI Analyzed: May 1, 2026 at 15:20 (48 min after Collected)
DXHR Inc. (Headquarters: Shibuya-ku, Tokyo; Representative Director: Kazunari Maeda; hereinafter "DXHR") held an event for engineers, "[LT Study Session] AI Utilization Study Session #4 for Everyone to Discuss," at its co-working space "NeuroHub" on Thursday, April 23, 2026.
The phase where generative AI was lauded as a "magic wand" has ended, and we are now entering a "practice" phase of how to integrate it into actual work and create value. This article reports on the LT (Lightning Talk) event where frontline engineers and entrepreneurs in AI development gathered. We deliver practical knowledge and business insights for working with AI, from individual task management to context design in product development and the construction of autonomous AI agents.
## Background of the Event
With the evolution of generative AI technology, efficiency in all development processes, such as coding, testing, and meeting minute creation, has advanced. However, new challenges have also emerged on the ground, such as "AI was introduced, but the expected accuracy was not achieved" and "development speed increased, but is it really leading to customer value?" This event was planned with the aim of sharing raw know-how, including the "gritty reality" of AI utilization and failure stories, and promoting knowledge exchange and networking among engineers.
## Heated Group Work Among Participants
Discussing "The Real Current State of AI"
Prior to the LT presentations, group work was conducted at each table on the theme of "What made work more efficient with AI and what didn't." As engineers and business personnel who regularly interact with AI tools were gathered, a "gritty reality from the field" that went beyond mere technical theory emerged one after another. Success stories and failure stories that could only be obtained from actual sites, such as "I introduced AI with high hopes, but it was disappointing here" and "Conversely, it was exceptionally effective for this troublesome task," were openly shared, and the venue was filled with great enthusiasm.
## LT Session
1. The Story of How Work Started Flowing When I Became a Slave to AI (Speaker: Mr. Arai)
Mr. Arai presented a case where he resolved task deadline delays by entrusting all of his 24-hour schedule management to AI. By linking with Google Calendar and following AI's instructions (notifications) like "Do this at this time," he created an environment where he could fully commit to what he truly wanted to do. Utilizing this knowledge, he is now also expanding automated services such as AI phone reception.
2. The Difficulty of Measuring AI Accuracy from a User's Perspective (Speaker: Mr. Takayama)
Mr. Takayama shared the secret story behind the development of their in-house AI meeting minute product "Secure Memo Cloud." He explained that there is a significant discrepancy between the numerical accuracy scores (WER/DER, etc.) that developers emphasize in transcription and summarization, and the actual satisfaction users experience (accuracy of proper nouns, granularity of summarization, and preference for format). He spoke about the real struggles and the importance of verification to put the service into practical operation.
3. From Finance to Management, and Then to Context Engineering (Speaker: Mr. Mokkuma)
Mr. Mokkuma raised the point that with the dramatic increase in "creation (coding)" speed due to AI, the importance of managing "what should be created" and "whether it truly has value" has relatively increased. He presented practical methods of "context engineering" to improve the quality and speed of decision-making by inputting all business "contexts" such as customer interviews and GA4 data into AI.
4. Summary of Development Environment Improvement with CI/CD Tools (Speaker: Mr. Pawn)
Mr. Pawn introduced initiatives to eliminate "double work" and "the pain of rework" in the development environment using AI. He explained the automation of branch creation using MCP and the AI-driven checking of coding standards. Furthermore, as a future outlook, he discussed the concept of having multiple AI agents with different contexts perform critical and multifaceted code reviews in parallel processing within Docker containers.
## Networking Session
After the LT session, pizza and drinks were served, and a lively networking session took place among the participants. At each table, intense information exchange unique to a closed setting occurred, ranging from advanced technical discussions such as "how to feed context to AI" to realistic failure stories like "this AI tool was disappointing." With engineers, management, and consultants mingling, the space was filled with the energy and "serendipity of knowledge" unique to a real event, where technical and business perspectives intersected.
The phase where generative AI was lauded as a "magic wand" has ended, and we are now entering a "practice" phase of how to integrate it into actual work and create value. This article reports on the LT (Lightning Talk) event where frontline engineers and entrepreneurs in AI development gathered. We deliver practical knowledge and business insights for working with AI, from individual task management to context design in product development and the construction of autonomous AI agents.
## Background of the Event
With the evolution of generative AI technology, efficiency in all development processes, such as coding, testing, and meeting minute creation, has advanced. However, new challenges have also emerged on the ground, such as "AI was introduced, but the expected accuracy was not achieved" and "development speed increased, but is it really leading to customer value?" This event was planned with the aim of sharing raw know-how, including the "gritty reality" of AI utilization and failure stories, and promoting knowledge exchange and networking among engineers.
## Heated Group Work Among Participants
Discussing "The Real Current State of AI"
Prior to the LT presentations, group work was conducted at each table on the theme of "What made work more efficient with AI and what didn't." As engineers and business personnel who regularly interact with AI tools were gathered, a "gritty reality from the field" that went beyond mere technical theory emerged one after another. Success stories and failure stories that could only be obtained from actual sites, such as "I introduced AI with high hopes, but it was disappointing here" and "Conversely, it was exceptionally effective for this troublesome task," were openly shared, and the venue was filled with great enthusiasm.
## LT Session
1. The Story of How Work Started Flowing When I Became a Slave to AI (Speaker: Mr. Arai)
Mr. Arai presented a case where he resolved task deadline delays by entrusting all of his 24-hour schedule management to AI. By linking with Google Calendar and following AI's instructions (notifications) like "Do this at this time," he created an environment where he could fully commit to what he truly wanted to do. Utilizing this knowledge, he is now also expanding automated services such as AI phone reception.
2. The Difficulty of Measuring AI Accuracy from a User's Perspective (Speaker: Mr. Takayama)
Mr. Takayama shared the secret story behind the development of their in-house AI meeting minute product "Secure Memo Cloud." He explained that there is a significant discrepancy between the numerical accuracy scores (WER/DER, etc.) that developers emphasize in transcription and summarization, and the actual satisfaction users experience (accuracy of proper nouns, granularity of summarization, and preference for format). He spoke about the real struggles and the importance of verification to put the service into practical operation.
3. From Finance to Management, and Then to Context Engineering (Speaker: Mr. Mokkuma)
Mr. Mokkuma raised the point that with the dramatic increase in "creation (coding)" speed due to AI, the importance of managing "what should be created" and "whether it truly has value" has relatively increased. He presented practical methods of "context engineering" to improve the quality and speed of decision-making by inputting all business "contexts" such as customer interviews and GA4 data into AI.
4. Summary of Development Environment Improvement with CI/CD Tools (Speaker: Mr. Pawn)
Mr. Pawn introduced initiatives to eliminate "double work" and "the pain of rework" in the development environment using AI. He explained the automation of branch creation using MCP and the AI-driven checking of coding standards. Furthermore, as a future outlook, he discussed the concept of having multiple AI agents with different contexts perform critical and multifaceted code reviews in parallel processing within Docker containers.
## Networking Session
After the LT session, pizza and drinks were served, and a lively networking session took place among the participants. At each table, intense information exchange unique to a closed setting occurred, ranging from advanced technical discussions such as "how to feed context to AI" to realistic failure stories like "this AI tool was disappointing." With engineers, management, and consultants mingling, the space was filled with the energy and "serendipity of knowledge" unique to a real event, where technical and business perspectives intersected.