THOUGHT LEADERSHIP · SMALL TEAMS
What to Automate First When You’re a Two- or Three-Person Assessment Team
Small teams tend to automate whatever’s easiest, not whatever’s actually slowing everything down. A bottleneck framework for deciding where to start.
By Eduface · July 2026 · 8 min read
Small teams tend to make the same mistake when they start automating, and it’s an understandable one: they automate whatever is easiest to automate, or whatever a vendor happens to be pitching that week, rather than whatever is actually slowing everything else down. If you’ve got two or three assessors and limited time to spend evaluating tools, it’s worth borrowing an idea from a completely different field before you decide where to start.
Borrow the bottleneck idea from operations management
Eliyahu Goldratt’s 1984 book The Goal, still one of the most widely read management texts ever written despite being, unusually, structured as a novel, introduced what’s become known as the Theory of Constraints. The core idea is almost embarrassingly simple once you hear it: every process has a bottleneck, a single step that limits how fast everything downstream of it can move, and improving anything other than that bottleneck does essentially nothing for overall output (Goldratt, 1984). Speed up a non-bottleneck step and work just piles up faster in front of the real constraint. The whole system only gets faster when you fix the actual bottleneck.
Quick answer
For most small assessment teams, the real bottleneck is the first read, not grade release or notifications. Automate that, not the tidy-looking steps around it.
Marking is a process, in exactly the sense Goldratt meant. It has steps: submissions arrive, an assessor reads them, forms a judgement, writes it up, and a result gets released. In a two or three person team, one of those steps is almost always the actual constraint, and it’s rarely the one that gets automated first, because the obvious candidates, releasing results, sending notifications, logging grades, are the easy, visible parts of the process rather than the slow ones.
Where the real bottleneck usually sits
For most small assessment teams, the constraint is the first read: the initial pass where an assessor has to actually absorb a long submission and form a judgement about it from scratch. That’s the part that scales worst with volume, because it requires sustained, undivided attention, and it’s the part a tired assessor at the end of a long day genuinely cannot speed up by working harder. Everything after that first read, writing up the reasoning, checking it against a rubric, releasing the grade, moves faster precisely because the hard cognitive work is already done.
This is worth sitting with, because it cuts against where automation effort usually goes. A lot of small providers start by automating grade release notifications or building a slightly nicer spreadsheet for tracking submissions. Those changes make the non-bottleneck parts of the process feel tidier. They do almost nothing for how long a learner actually waits for a result, because the bottleneck, the first read, hasn’t moved at all.
What this means for AI-assisted marking specifically
This is exactly why AI-assisted marking earns its place at the front of the queue rather than the back. A tool that produces a structured first draft, with reasoning against each criterion, doesn’t remove the assessor’s judgement, our companion piece on human-in-the-loop marking covers why that step has to stay genuinely real, but it does change what the assessor is doing with their time. Instead of doing the full first read from a blank page, they’re reviewing a draft, which is a genuinely faster cognitive task even when they end up making substantial changes to it. That’s addressing the actual constraint, not a step downstream of it.
A short exercise worth doing before you buy anything
Time your own process for a week. Not in detail, just roughly: how long does the first read of a typical submission take compared to everything that happens after it. For most small teams, the first read dwarfs everything else combined, often by a factor of five or more. If that’s true for you, it tells you clearly where automation effort belongs. If your bottleneck turns out to be somewhere else, maybe it’s actually the final sign-off queue backing up because only one person is authorised to release results, that’s genuinely useful to know too, and it might mean a different fix entirely, like authorising a second person to approve results, rather than any marking tool at all.
Goldratt’s other point, easy to miss, is that once you fix one bottleneck, the constraint moves somewhere else, and the process of identifying and addressing it starts again. Don’t treat this as a one-off decision. Revisit it every few months as your volume and team shape change.
Frequently asked questions
What’s usually the actual bottleneck in a small assessment team’s marking process?
Most often, the first read: the initial pass where an assessor absorbs a submission and forms a judgement from scratch. It’s the step that scales worst and benefits least from simply working harder.
Why does automating notifications or grade release rarely help much?
Because those steps usually aren’t the bottleneck. Speeding up a non-constraint step just means work piles up faster in front of the real one, without actually reducing how long a learner waits.
How do I figure out where my own bottleneck actually is?
Time your process roughly for a week: compare how long the first read of a submission takes against everything that happens afterward. The step that dominates the total time is almost certainly your constraint.
Does the bottleneck stay in the same place forever once I fix it?
No. Goldratt’s framework treats this as a cycle: once you address one constraint, another step becomes the limiting one, and it’s worth revisiting the exercise periodically rather than treating it as solved permanently.
Sources
Goldratt, E. M. (1984). The Goal: A Process of Ongoing Improvement. North River Press.
Fix the first read, not the notifications
Eduface gives assessors a structured first draft with reasoning built in, so the actual bottleneck in your marking process moves, not just the parts around it.