# AI Workflows, Forward-Deployed Development, and Team Coordination
- Session: 93b8daaf-77f0-41d4-bc3d-37f6a5b4f407
- Channel: Discord #🐣│cohort-voice
- Started: 2026-06-24T20:32:27.696Z
- Ended: 2026-06-24T21:01:17.222Z
- Participants: duckanbro, samkuhlmann, ECWireless, Ξ2T, takekek
- Tags: ai-workflows, forward-deployed-development, multiplayer-ai, mcp, microsoft-365, team-coordination, agentic-tools, customer-research
## TL;DR
The group discussed how AI tools are reshaping product work, consulting, customer research, and team coordination, with Travis sharing lessons from building agent-friendly Microsoft 365 workflows for a family office client and reflecting on the need for better metrics, shared context, and human collaboration around AI systems.
## Summary
The meeting centered on Travis introducing his background in crypto, DAOs, DAOhaus, and his current work with Syncrobe, a voice/AI startup exploring multiplayer AI, shared context, skills, workflows, and consulting-led revenue. He described a recent six-week client engagement with a family office group where the work began with customer research and identifying painful workflows, then evolved into building an MCP server connected to Microsoft 365, including OneDrive, Outlook, and SharePoint. That server became the core engine for creating custom workflows and skills around the organization's actual day-to-day work.
A major theme was that many companies are not yet ready for AI agents because their data, files, and processes need cleanup before automation can reliably help. Travis emphasized that successful AI implementation is not magic: it requires configuration, iteration, usage cycles, connectors, and foundational data hygiene. He noted that his team has not yet put strong success metrics in place for the pilot, although executive buy-in and a focused pilot team have helped adoption so far.
The group also discussed the idea of a forward-deployed developer or agency model. Travis framed it as being on-site or close to users, watching how people actually work, hearing their problems, and then immediately building and shipping solutions rather than writing specs and waiting through long prioritization cycles. He contrasted this concrete, embedded work with abstract product-building and highlighted the value of making systems that improve with every iteration.
Tooling and personal workflow were another focus. Travis described using Claude Desktop and Claude Code together: keeping Claude Desktop as the planning and context hub, then sending research or coding tasks into Claude Code, gathering reports back, and feeding the results into the primary context. He also described his agent work with Hermes and a personal agent named Sivart running primarily on GLM 5.2. He sees current tools as powerful but still largely individualistic, with unresolved challenges around sharing value, context, and output across teams.
The conversation closed with reflections on the pace of AI change, the risk of overwhelm, and the possibility that agents will free people from grunt work so they can focus on more meaningful collaborative work. Participants noted that AI feels like it is moving extremely fast even though ChatGPT has been public for nearly four years, and that AI may be more accessible to mainstream users than web3 despite the group being unusually deep in both spaces.
## Action Items
- Define pilot success metrics: Create concrete metrics for evaluating whether the AI/Microsoft 365 pilot is improving workflows, adoption, and business outcomes. (owner: Ξ2T)
- Continue collecting AI practitioner perspectives: Keep running and recording cohort voice interviews to gather diverse perspectives on AI workflows, collaboration, tooling, and adoption challenges. (owner: duckanbro)
- Explore team-level context sharing: Investigate how individual AI workflows, agents, and research outputs can be shared across a team so one person's work improves another person's context and productivity.
- Review transcript for reusable content angles: Use the transcript to identify repeated themes and phrasing that can be shaped into future content, especially around real people, forward-deployed AI work, and agent-friendly business systems.
## Notable Quotes
- Ξ2T: "AI multiplayer. Like now we all have these superpowers and silos."
- Travis described the opportunity as moving beyond isolated individual AI use toward shared team context, skills, and workflows.
- Ξ2T: "This shit just isn't magic at first."
- AI implementations require configuration, iteration, cleanup, and patience before they produce reliable value.
- Ξ2T: "I'm in the console next. Crafting up a solution and shipping code."
- Forward-deployed development means staying close to users, understanding the problem directly, and building immediately instead of passing specs through long approval cycles.
- Ξ2T: "Software development as a team sport."
- Travis emphasized that current AI workflows can feel too solitary and that better collaboration models are still needed.
- samkuhlmann: "it's more normie friendly than web3. we are weird"
- Sam noted that AI may be more accessible to mainstream users than web3, even if the group is unusually immersed in both.