The middle step is the one everyone skips
Ask any NSW teacher to describe gradual release and you'll hear the same three words: I do, we do, you do. It's on staffroom posters. It's in the induction slides. It's the shorthand for explicit instruction across the sector.
But watch a run of lessons and a pattern shows up. The model is real in theory and thin in practice. The 'I do' gets rushed. The 'we do' gets cut. And students land in independent work before they can actually do the thing on their own.
That's not gradual release. That's a demonstration followed by a cliff.
What each stage is actually for
It helps to be precise about the job of each phase, because they are not interchangeable.
- •**I do** is the teacher modelling. Not describing the task. Thinking aloud while doing it, so the invisible steps become visible.
- •**We do** is guided practice. The class works the same type of problem *with* the teacher, who is checking, prompting and correcting in real time.
- •**You do** is independent practice, once the error rate is low enough that students can succeed without help.
Hollingsworth and Ybarra are blunt about the sequence. You don't move on until the class is getting it right. The whole point is to catch mistakes while they're cheap to fix, before they harden into independent practice done wrong.
How 'we do' quietly disappears
Nobody deletes guided practice on purpose. It erodes. Here's how it usually happens.
The clock does it. The modelling ran long, there are 15 minutes left, and 'have a go at questions 1 to 8' feels like the responsible choice. It isn't. Eight questions practised wrong is worse than two questions practised together.
A worked example gets mistaken for 'we do'. The teacher solves one on the board and asks 'everyone got that?' Silence reads as agreement. Then it's straight to independent work. But watching someone else solve a problem is still 'I do'. The students haven't done anything yet.
One confident student answers for the room. A question goes out, the same three hands go up, one gets it right, and the lesson rolls on. That's not the class practising. That's one student practising and 27 spectating.
What good 'we do' looks like in the room
Guided practice has a texture you can spot from the back of the room. A few markers:
- •The teacher poses a problem, everyone attempts it, and the teacher *checks a sample of responses before revealing the answer*. Mini whiteboards, a quick 'show me on your fingers', a scan of books mid-row.
- •Wrong answers get surfaced and worked through, not skipped past. The teacher wants to see the mistake so they can address it now.
- •The prompting fades across the set. The first one is heavily scaffolded. By the fourth, the teacher is mostly quiet.
A Year 9 maths example. I do: the teacher solves 2(x + 3) = 14, narrating every step, including the ones students usually skip. We do: the class does 3(x - 2) = 9 together, but the teacher checks every student's first line before anyone moves to the second. Two more like it, with less prompting each time. You do: now the worksheet, because the error rate is low and the teacher knows it.
The difference is checking. 'We do' without a check for understanding is just 'I do' with more people talking.
The other failure: skipping 'I do'
The reverse mistake is common in senior classes and in schools that lean discovery. Students get handed a rich task and told to 'have a go' before anything has been modelled. The strong kids cope by importing knowledge from home or a tutor. Everyone else guesses.
This isn't a criticism of hard tasks. It's about sequence. Model first, practise together, then release. The task can be as ambitious as you like once students have the tools to attempt it.
Why this is a planning problem, not a teaching problem
Here's the honest bit. Most teachers know all of this. The reason 'we do' gets cut isn't ignorance. It's that a proper guided practice stage takes real work to build. You need a graded set of examples that move from heavily scaffolded to nearly independent. You need checks baked in at the right points. You need enough parallel problems to actually practise, not just one.
Building that for five lessons a day, across multiple courses, is where the model dies. Not from a lack of belief. From a lack of hours.
That's the specific gap Smart Syllabus Assistant is built to close. Every lesson it generates is structured around the three phases, with the guided practice stage written in properly: worked examples that fade, checks for understanding at the transition points, and enough practice items to make 'we do' meaningful. Then it's checked against explicit instruction (Archer and Hughes; Hollingsworth and Ybarra) before it ever reaches you.
The teaching is still yours. The reading of the room, the timing, the decision to loop back when the checks come back shaky. But the scaffold arrives already built, so the middle step is there when the clock gets tight.
See a lesson with the 'we do' stage built in properly at smartsyllabusassistant.com/sample.