Unlocking 9 years of project history

Bednark Studio is a leading design and technical production studio based in Brooklyn. The 100+ person company has 20+ years of experience serving global brands and creative agencies including Nike, Google and Target, with work spanning custom fabrication, architectural millwork and experiential marketing.

We came in with a simple mandate: help roll out Claude licenses fully with the team. When we dug into it, we found that the initial rollout stalled because the answers were too generic. "AI didn't know how we work at Bednark" was the predominant feeling, and the rollout had understandably created more AI skeptics than enthusiasts.

But Bednark had already done the hard part. The company maintains a detailed operational record of 520+ projects in their ERP, and every meeting was recorded and transcribed to capture the day-to-day decision making process. The problem was that their AI tools had no access to it. Our goal was to fix that.

The tasks that people were asking AI to do were concrete: finding comparable work orders to price a new deliverable, confirming what scope had been agreed on a past call, locating the latest version of a client creative.

But the pattern was the same across all of them. The information existed, but Claude couldn't reach it. That meant folks were falling back to how things have always been done and finding it by hand, spending hours scrolling through Google Chat spaces to find the dimensions for a part heading into production.

To close the gap, we built a custom Bednark MCP server, focused on connecting these systems of record:

• Innergy. The ERP containing 305,000+ structured records covering budgets, job costing and change orders across every project.
• Read.ai. External and internal meeting transcription, with 5,900+ recorded meetings.
• Project knowledge archive. Nine years of internal project correspondence and documentation.
• Google Chat. The named team and project spaces where scope calls, dimensions and shop-floor decisions get made day to day.

The sequencing mattered. Solving data access meant the workflows and skills built after had something real to stand on. And it outlasts any single platform decision, because whatever Bednark adopts next will need to connect to Bednark MCP as well. We saw this work as the highest impact first investment in a longer term AI rollout strategy, not a one-off integration.

After building this out, we hosted a Claude 101 workshop where we got everyone set up on their Claude accounts with the official Claude connectors for Gmail, Google Drive and Google Calendar, and the custom Bednark connectors for Innergy, Read.ai and the Knowledge Archive.

Within 45 minutes, everyone had a provisioned seat, a configured Claude instance, and one or two prompts of their own already run.

And the impact showed immediately. In the first 30 days, Claude usage has grown quickly in Bednark across all teams: business development, project management and production, covering both fabrication and millwork. In that period, 88% of the connector activity ran through our custom connectors.