How Medico Digital Is Putting AI to Work Across the Agency

One of the primary goals we wanted to achieve was a move from block bookings to a much more granular, task-based approach to resourcing. The main challenge was that previously  setting up all those tasks, maintaining timelines, and keeping everything up to date required a lot of manual work for our Client Partners. With the Productive MCP, we’ve been able to build a sequence of skills that they can run at specific points throughout the month. 

This helps us create tasks from our templates, identify gaps in planned allocations, and see which team members have the best capacity when new clients come on board. It’s also made our resource planning much more fluid. Instead of reviewing capacity once a month, we can now run weekly stand-ups and reassign tasks based on each department’s workload and client activity.

We’re using the MCP to take a lot of the manual admin out of our day-to-day operations. Because most of our clients are on retainers, there’s a significant amount of end-of-month work involved in closing budgets, creating invoices, and checking for discrepancies. I can now manage much of this at a high level and ensure the process is consistent across the agency. 

We’ve also combined Productive’s existing automations with the MCP to mass-create and finalize invoices for our smaller clients. Previously, someone would spend around half a day doing this, whereas we can now create and spot-check more than 50 invoices in around 30 minutes. 

We’re also using the MCP to improve time-tracking governance by cross-referencing Productive with Google Calendar, checking that team members are providing sufficient notes, and identifying patterns in how long certain tasks take to help quote future projects.

One of the biggest advantages is that Productive no longer has to exist in isolation. We can layer the Productive MCP with our third-party call recorder, Google Drive, Slack, Gmail, and our custom Medico Digital MCP, which includes client analytics and advertising data. We can then bring all of that information into Claude to generate much more context around a client. 

This is particularly useful when a client is being handed over or when someone needs a second pair of eyes on an account, because we can quickly understand what’s already been done, what’s planned, and what’s happening across the different channels.

In the future, we also plan to use this setup to create a new version of client reporting that is semi-automated and hosted in our own intranet for clients to view on demand.

We’ve started building workflows that go well beyond basic automation. For example, I’ve set up a skill that uses three-month rolling averages to project team capacity for the next quarter. Previously, we focused more on the total number of hours clients had paid for than on the actual hours required to service them. We can now bulk-create predicted capacity based on historical averages and continuously update it at the end of each month. 

We’re also using MCP for deal management, where our business development team can trigger workflows when deals reach certain probability stages, prompting them to add services and allocations. We can even add proposals into Claude and have the Productive MCP populate the relevant line items. 

Another workflow we’ve built is for web support, where client emails are routed to a central Productive project, and the MCP can automatically triage them, suggest an estimated due date, and identify the developer with the best capacity and skills to handle the issue. 

We’ve also used it to make large-scale changes across our Productive setup. For example, when we needed to correct budget line items that had been set up using pieces instead of hours, I was able to review and update them across the entire agency in one afternoon — something that would probably have taken a full week to do manually.

We have multiple initiatives that are currently planned to further enhance our usage of the MCP and provide even further efficiencies and more joined-up ways of working. One of those is exploring leveraging the Productive Notetaker rather than our own current third-party solution to integrate client context even further and increase the ability to add automated tasks following client calls.

From a business development perspective, we are also beginning to scope the idea of creating proposals within Productive via the MCP, and then using Productive to sign client documents (which would then trigger an automation notifying the delivery teams that a client has signed and work needs to begin).