Anonymised, illustrative composite. Forty RFIs on an Ontario GC’s prior project averaged nine business days to answer, run over email CC chains where nobody was formally on the hook for a response. One embed-plate location RFI sat fourteen days, bouncing between three people who each assumed someone else had it.
At a glance
The GC’s RFI process ran, for years, over email: a question went out to a distribution list, and whoever felt best positioned to answer eventually did. Most of the time that worked well enough. On a sample of 40 RFIs from a prior project, the average turnaround was 9.0 business days — and one, asking for an embed-plate location on a structural detail, took 14, forwarded between the structural engineer, the architect, and the GC’s own project manager, each of whom had at some point assumed one of the other two was going to close it out.
Nobody involved was careless. The structural engineer genuinely believed the architect needed to weigh in first; the architect had flagged it back to the GC’s project manager for a site-condition check; the project manager, new to the job that week, did not realize the question had already looped back to him once. Three reasonable people, each making a reasonable assumption about who held the ball, produced fourteen days of nobody actually holding it.
Nothing about that RFI was technically difficult to answer. What made it slow was that no single person was ever formally responsible for answering it at any given moment — responsibility sat with a distribution list, which functions, in practice, like responsibility sitting with no one. The GC moved to Procore’s RFI tool, which routes each RFI through a “Ball In Court” mechanism: at any moment, exactly one assignee holds it, and “only the current ball-in-court person can forward an RFI for review to additional assignees.” Status moves through Draft, Open, and Closed, with a full history of every hand-off recorded against the RFI.
On the next comparable project, a sample of 40 RFIs averaged 3.4 business days — down from the prior project’s 9.0, a 62% reduction (9.0 minus 3.4 is 5.6; 5.6 divided by 9.0 is just over 62%). The two slowest RFIs under the new system still took 7 business days each — not fast, but for the first time, the tool showed exactly who currently held each one, rather than leaving the GC to guess which of three people on a CC line had actually picked it up.
When a design question stalls long enough to become a payment dispute rather than just a schedule risk, Ontario’s Construction Act adjudication regime moves faster than either version of this GC’s own RFI process: an adjudicator must issue a determination no later than 30 days after receiving the documents required for the adjudication (s. 13.13(1)) — see how construction adjudication differs from suing in court.
Both of those two slow RFIs turned out to be held up at the same design consultant’s office, waiting on a sub-consultant review the GC had no visibility into under the old email process. Knowing that — specifically, consistently, on both occasions — is itself a different outcome from the prior project, where an RFI sitting for two weeks produced no equivalent answer to “where, exactly, is it stuck.”
The technical difficulty of the questions being asked did not change between the two projects — the same design consultants answered both. What changed was that an RFI now has exactly one accountable holder at any given time, and moving it forward requires an explicit action from that specific person rather than a passive hope that someone on a distribution list will pick it up. A single-holder routing rule does not make anyone answer faster; it makes it impossible for a question to sit unanswered because everyone assumed it was someone else’s.
The schedule float that had been at risk on the embed-plate detail’s critical path was restored on the next comparable job, and RFI turnaround is now a tracked KPI reported at every biweekly owner-architect-contractor meeting. See what an RFI actually is and how it is used, and for a related documentation change on the same kind of project, see how a status flag caught a superseded drawing revision and how a lookahead review saved a week of crane time.
A 30-minute call is enough to tell you whether AI pays for itself here.
