Once the specification is issued and procurement starts, the contractor goes out to source the products that will meet it. For each specified item, a window system, a waterproofing membrane, a paint finish, a structural fixing, they submit technical data for your approval before anything gets ordered. Your job is to check that what they are proposing matches what you wrote. Sometimes it is a like-for-like match against the named product. More often it is a proposed equal, an alternative the contractor would rather use, usually because it is cheaper, more available, or already in their supply chain.
That sounds straightforward. On a mid-sized project it is anything but. A single building can generate several hundred submittals, and each one needs to be read against the relevant clause, cross-checked against the performance requirements, and then approved, approved with comments, or rejected. Get it wrong and the wrong product ends up in the building. The review itself is rarely difficult. The difficulty is that there are so many of them, and they all want a senior person's attention.
The delay almost never comes from a single review taking too long. It comes from volume meeting fragmentation. Submittals do not arrive in a tidy batch you can sit down and clear in an afternoon. They trickle in over weeks, in whatever order the contractor's procurement happens to run, and they land on the desk of whoever is senior enough to sign them off. That is usually the same architect already running the project.
So the reviews queue behind everything else. A submittal for a balustrade fixing sits in an inbox for four days, not because it is hard, but because the person reviewing it is in three meetings and chasing a planning condition. None of that waiting shows up as work. It shows up as a gap. Multiply one four-day gap across a few hundred submittals and the cumulative drag on the programme is enormous. Studies of construction delay consistently put submittal and approval cycles among the most common causes of slippage, and the striking part is that most of that time is spent waiting, not reviewing.
Put rough numbers on it. Say a project produces three hundred submittals over the construction phase, and each one waits an average of three working days before anyone looks at it. That is nine hundred days of waiting spread across the job. Even when the contractor runs several in parallel, the critical-path items still sit in the queue with everything else, and a handful of those will be long-lead products that decide whether a trade can start on time. The cost is not the review. It is the queue in front of the review.
There is a second, quieter reason the work piles up. To review a submittal properly you need the specification open in front of you. The right clause, the performance targets, any fire or acoustic requirements, the relevant window or door schedule. Assembling all of that for every single submittal is tedious, and tedious work gets deferred. Then, when the backlog finally forces your hand, you review at speed. That is where the real risk starts.
A submittal backlog is not only a scheduling problem. It is a compliance problem waiting to surface.
When a proposed equal comes in, the real question is whether it meets every requirement in the clause, not just the obvious ones. A render system might match the specified finish and colour but fall short on fire classification. A door might hit the acoustic rating and miss the fire rating once you account for the ironmongery and the glazing. A membrane might meet the headline waterproofing performance and quietly fail the vapour requirement that mattered on this particular build-up. These mismatches are easy to miss when you are skimming under pressure, and they are exactly the sort of thing that gets caught on site, or worse, after handover.
Since the Building Safety Act, the consequences of getting this wrong are heavier. The golden thread of information means you have to be able to show that what was installed matches what was specified and approved, with the evidence to back it up. A sloppy product submittal review process leaves holes in that record. Catching a non-compliant substitution at submittal stage costs you an email. Catching the same substitution on site costs a variation, a delay, and an uncomfortable conversation with the client. Catching it after practical completion costs a great deal more than that.
This is where the work starts to look different. Most of what makes submittal review slow is comparison. You hold the product data sheet in one hand and the specification in the other and check it line by line. That is pattern matching against a known standard, and it is precisely the kind of task that AI handles well.
Tools like Avoice can pre-screen an incoming submittal against the relevant specification clause and flag where the proposed product meets the requirement and where it does not. Because Avoice ingests a firm's own specifications, schedules, and project documents, it checks the submittal against what you actually wrote, not a generic clause library bolted on from somewhere else. If a proposed membrane misses a stated performance target, or a substituted door conflicts with the fire strategy set out in your own schedule, that conflict gets surfaced before the submittal ever reaches your desk.
The aim is not to take you out of the decision. It is to change what lands in front of you. Instead of a raw PDF and a blank approval stamp, you get a structured comparison. Here is the clause, here is what the contractor has proposed, here is where it lines up, and here is the one line that does not. The reading you used to do yourself has already been done. You arrive at the part that needs you, rather than spending an hour getting there.
It changes the failure mode too. When reviews are done by hand under time pressure, the things that slip through are the boring details, the third performance value down the data sheet, the clause cross-reference nobody had time to follow. A consistent first pass does not get tired on the two hundredth submittal of the job. It applies the same scrutiny to the last one as it did to the first, which is exactly the opposite of how a person works at four o'clock on a Friday with a deadline looming. That consistency is worth as much as the speed.
None of this turns submittal review into something automatic. A machine can tell you that a product misses a stated requirement. It cannot tell you whether that miss matters on this project.
Plenty of submittal decisions are judgment calls. A proposed equal might deviate from the letter of the spec in a way that is completely acceptable given the context. Another might satisfy every clause and still be wrong, for reasons that live in the design intent rather than in any written requirement. Knowing the difference is the architect's job, and it stays the architect's job. What changes is where your attention goes. You spend it on the decisions that genuinely need a human, instead of burning it on the line-by-line comparison that does not.
That is the right division of labour. The tool does the checking. You do the judging. Anyone who has spent a wet Friday afternoon cross-referencing data sheets against a spec knows exactly which half of that they would rather keep.
Speeding up product submittal review claws back programme time, and that alone would be worth it. But it does something more interesting than that. It changes the texture of the project. When submittals clear in a day instead of a week, procurement stops stalling, the contractor stops chasing, and the build keeps moving at the pace it should. The architect stops being the choke point in a process that was never really about design in the first place.
It also tightens the record. A review process that checks every submittal against the specification, consistently and in the same way each time, produces the audit trail the golden thread now demands almost as a by-product. Avoice structures that checking around recognised standards and keeps it grounded in your firm's own documentation, so the link between what was specified, what was proposed, and what was finally approved stays intact and traceable.
The same shift has already happened to every other part of architectural documentation. Hand-drafting gave way to CAD. 2D drawing gave way to BIM. Each time, the profession kept the judgment and handed off the repetition, and nobody now wants the repetition back. Product submittal review is next in line, and it is overdue. If you want to see how this works against a live specification rather than in the abstract, Avoice runs demos built around your own projects.
