There's an irony in building software that analyzes construction schedules: the work itself follows a critical path. Miss a dependency, and everything downstream slips. So when we started using AI as a development partner on FPM, it wasn't a gimmick. It was about shipping a better product, faster.
AI as a Development Partner
FPM's core is a schedule analysis engine built in C#. It handles forward and backward passes, calendar-aware date arithmetic, dependency resolution, and forensic delay attribution across schedule update windows. The kind of code where a one-line bug can misattribute millions of dollars in delay liability.
Claude has been instrumental in building this. Not as a code-generation machine that spits out boilerplate, but as a genuine thinking partner. When we're working through the schedule math, figuring out how changes ripple through a network of thousands of activities, having an AI that can reason about graph algorithms, CPM semantics, and edge cases in the same conversation is transformative.
The work is deeply technical: concurrent delay attribution, out-of-sequence progress detection, calendar exception handling, float consumption analysis. These aren't problems you solve by asking an AI to "write a function." They require iterative design, domain knowledge, and careful validation against real P6 XER schedule data. Claude brings the ability to hold all of that context simultaneously.
MCP: Opening the Engine to Everyone
The most exciting thing we've shipped recently is our MCP (Model Context Protocol) integration. MCP is an open standard that lets AI assistants connect to external tools and data sources. For FPM, this means any MCP-compatible AI client can query your schedule data directly.
Think about what that unlocks. Instead of navigating complex schedule analysis software, a project manager can ask:
"What caused the 45-day slip on Milestone M-100 between the June and August updates?"
The AI connects to FPM via MCP, runs the delay analysis, traces driving paths, and explains the results in plain language, citing specific activities, float changes, and delay categories.
Our MCP server exposes the full analytical toolkit: schedule overviews, activity search, CPM details, delay analysis, driving path traces, relationship analysis, and more. It's the same engine that powers the FPM application, now accessible through natural conversation.
Why This Matters for Construction Disputes
Forensic schedule analysis has traditionally been the domain of a small number of specialized consultants. The tools are complex, the methodologies are arcane, and the stakes are enormous. Delay claims on large construction projects routinely involve tens of millions of dollars.
By combining rigorous CPM calculations with AI accessibility, we're working to change that equation. The analysis stays mathematically precise. Our engine is deterministic, and every result is traceable back to specific schedule data. But the interface becomes conversational. You don't need to be a scheduling expert to understand why a project is late.
What's Next
We're continuing to build FPM in the open, with AI as a partner in development and as a feature of the product itself. The MCP integration is available to early adopters. If you work with P6 schedules and want to try querying your schedule data through AI, we'd love to hear from you.
This blog will be the place where we share what we're building, what we're learning, and the occasional deep dive into CPM theory that only a scheduling nerd could love.
Stay tuned.