CritPath AI

CritPath AI is a schedule-risk management platform built specifically for R&D programs — biotech, deep tech and federally funded research — where a deterministic Gantt chart is a promise nobody in the room can defend. Most project tools show a single finish date and say nothing about how likely it is. The established tools that do this math are Windows desktop software priced for capital projects and defense programs, and they usually sit with a specialist rather than with the program team. CritPath AI puts the full engine in a browser at $10 per user per month. THE ENGINE Critical Path Method across all four dependency types, with float, near-critical detection and a WBS hierarchy that rolls up automatically. Monte Carlo simulation over your task network — 10,000 iterations by default, built on three-point PERT-Beta estimates, with optional duration correlation and the task-linked risks from your register sampled into every run. You get the full completion distribution, P50/P75/P90/P95 finish points, a criticality index showing how often each task actually lands on the critical path, and a ranked sensitivity analysis of what is driving your finish. Protect the date where it is genuinely at risk: Critical Chain and Drum-Buffer-Rope find your constraint, size project and feeding buffers — statistically, or from your simulation at a confidence level you choose — and track consumption on a live fever chart that shows whether protection is burning faster than the work is finishing. Integrated cost-schedule risk: labor cost driven by sampled durations, plus three-point direct costs and risk events from the same iterations, producing P50 and P80 cost and an explicit contingency figure. DECISIONS THAT LIVE IN THE SCHEDULE Go / No-Go / Pivot / Defer gates are first-class schedule objects with a named owner, an audit trail and an optional two-person co-sign mode. A No-Go or a Pivot triggers a fresh run of the schedule math. So does marking a risk occurred, or accepting an AI proposal: retroactive rescheduling re-cascades the critical path, the buffers and the simulation across the whole program, so a re-plan that used to be a manual PMO exercise happens before the meeting ends. RESOURCES AND ECONOMICS Staff the plan and see what it costs. Durations driven by effort, resource leveling validated against the published PSPLIB academic benchmarks, and a staffing optimizer that ranks candidate reassignments by cost of delay, criticality and float, then shows what each one buys in days and dollars. WSJF and Cost of Delay are computed from each task's real position in the live schedule rather than typed into a spreadsheet. Earned-value reporting adds schedule and cost performance, forecast cost at completion and variance, and a portfolio dashboard rolls every program up. AI THAT PROPOSES, PEOPLE WHO APPROVE A schedule-aware copilot reasons over your actual dependency graph, not a generic chatbot pinned on a task list. Beyond it you can build AI employees — an Estimator, a Risk Analyst, a sub-PM — each with skills you define and grounded in your organization's own knowledge; autonomous runs switch on at general availability. They read the live schedule and return typed proposals. Nothing reaches the plan until a named human accepts, and every acceptance is audited. Turn on the adversarial reviewer and a separate, isolated pass grades each run against your own quality rubrics; in blocking mode, a failure makes acceptance impossible. EVIDENCE YOU CAN HAND TO AN AUDITOR A per-organization, append-only, tamper-evident audit hash chain with in-app integrity verification. A per-project maturity assessment scores your actual configuration and evidence against the AACE RP 132R-23 Level 4 ladder — three-point estimates, a risk register wired into the schedule, confidence targets, buffer management, decision gates, duration correlation, integrated cost-schedule risk and quantified cost of delay — so the compliance claim is auditable per project instead of asserted in a brochure.