AI Assistant
A project-controls assistant that drafts, edits and reviews — where every proposed change is staged for your approval before anything is written.
Who it is for. Planners who want a senior reviewer’s second pass on a schedule, estimate or risk register without booking one, and who are comfortable supplying their own model-provider account. Enterprise buyers should talk to us about deployment before rolling it out.
What it is for
Reviewing a schedule for open ends, placeholder durations, unjustified constraints or an under-quantified risk register is expensive expert work that usually happens late, if at all. This does it against the live project model and hands back concrete, individually reviewable edits rather than prose advice.
Two things. The review is module-scoped rather than schedule-only — the same panel brings risk-register expertise in Risks, estimating expertise in Estimates and DCMA expertise in the DCMA view. And the staging model: every finding is independent, replays onto the pre-check baseline, and is previewed in the module before anything is written. This is not AI bolted onto a project tool; it is a review pass operating on the same model the scheduler does.
Awaiting capture — WF32-BOP
AI Check panel open over the Activities viewport with staged proposals, Approve and Reject controls per row, and a visible diff.
What it does
Every line below is in the product today.
Nothing is applied automatically
A proposal is staged as a non-destructive preview in the Activities grid to accept or discard. A brand-new project is built in memory and still needs an explicit save.
Review intents cannot propose changes
The assistant classifies its own intent, and only "modify" may emit a proposal. Review, verify, inspect, validate and report answer in prose and are barred from editing.
Module-scoped expertise
AI Check carries fifteen viewport mappings, each with its own domain expertise, finding categories, presets and an allow-list of operations it may emit. Modules with no applicable operations are advisory-only by design.
Findings are independent, and reversible
Accepting a finding replays only the accepted set onto the pre-check schedule as a preview, and closing the preview un-stages them.
The whole project is always in context
A review argues a risk finding against the schedule’s float and the cost base, not just against the rows on screen.
It reads the files you already have
XER and MSPDI schedules are parsed and summarised in the browser before sending, spreadsheets converted per sheet, PDFs and images sent as native parts.
A truncated reply says so
If the token limit cuts off a proposal, you are told to retry or narrow scope rather than being handed a silently incomplete change set.
It reads and writes with
Activities
The critical-path Gantt where the schedule is actually built — a virtualised table and timeline that computes your dates from logic, calendars and constraints instead of letting you type them.
Risks
A portfolio register carrying threats and opportunities, each with its own matrix and strategies, linked both ways to the schedule.
See it on your own schedule.
Send an XER before the call and we will import it and show you AI Assistant running against your data.