All projects02 — Symphony

Make AI development usable by a whole team

Proposed and developed an agent-driven workflow platform, then worked on ClickUp integration, analytics, notifications, onboarding and feedback so teams could use it in their day-to-day delivery.

Employer
New Media Lab
Relationship
Employer initiative
Period
Responsibility
Platform development, integration and adoption
Delivery stage
Developed and used across project teams

Technologies & methods Agent workflows · ClickUp integration · Analytics · Telegram notifications · Team onboarding

Illustrative workflow Task integration, visible execution and team feedback form one development workflow. This is a conceptual view, not an internal system diagram.
  • My scope
  • Shared contribution
  • Other owner

From task to adoption

  1. My scope Task context ClickUp integration and project setup
  2. My scope Agent workflow Platform development and operational setup
  3. My scope Visibility Analytics, usage reporting and notifications
  4. Shared contribution · With product teams Team use Onboarding, feedback and continued iteration

Context

An AI coding agent needs more than a prompt to fit into team delivery. Work has to arrive from a task system, people need to see what happened, and users need a practical way to learn the workflow and report problems.

I proposed and began Symphony at New Media Lab, treating the implementation, integrations, operational visibility and team adoption as parts of the same engineering problem.

My contribution

I developed the workflow platform and contributed its ClickUp integration, connecting task-based work with agent-assisted execution. I worked on deployment and setup readiness, coordinated with engineering and operations stakeholders, and supported rollout into different project contexts.

I gathered feedback from PMs and engineering teams and adapted the workflow to how those teams worked. Setup, troubleshooting and knowledge transfer continued alongside feature development.

Visibility and usability

I contributed an analytics page, usage-dashboard work and a usage-report API so activity could be inspected beyond individual agent responses. Telegram notifications gave users another way to follow the workflow. I also added Simplified Chinese support for teams using it.

Onboarding and knowledge transfer

I supported setup and knowledge transfer across multiple product teams, including sessions for new users. The work included gathering feedback after setup and addressing small problems that could otherwise interrupt adoption.

Later, I prepared an eight-slide townhall deck explaining the workflow and approval responsibilities to PMs and non-technical colleagues. This was a communication artifact; the source record did not yet establish that the presentation had taken place.

Validation and delivery

The recorded work progressed from the initial proposal into integrations, deployment support, analytics, notifications and use across projects.

The platform dashboard recorded 480 resolved items over 99 days, from 19 May to 25 August 2026. That figure describes platform activity across users and projects, not tasks I personally completed or engineering hours saved. The available records do not establish an organization-wide adoption rate.

Further discussion

I’m glad to talk through this work and how a similar approach could help your team.