I've watched dozens of teacher-development pilots over the last decade — some ours, many not. The ones that turn into long-term partnerships almost always look different from the ones that quietly fade after a year. The difference rarely has anything to do with the platform itself. It has to do with how the pilot was designed before the contract was signed.
If you're a superintendent, curriculum director, or operations lead about to greenlight a teacher-growth pilot this year, here are five things you should insist on — even if it's not us you end up choosing.
Define two or three measurable outcomes before you define scope
Most pilots start with the wrong question: "How many sites should we include?" The right question is: "What two or three teacher-level outcomes do we need to see by spring for this to be a yes?"
Possible answers: a measurable shift in observed instructional practice, a reduction in admin time spent on evaluation paperwork, a year-over-year change in retention at pilot sites versus comparable non-pilot sites. Pick the two or three that matter most to your board narrative.
Once you have those, scope follows naturally. Without them, you're piloting a feature set, not a result.
Insist the pilot mirrors your eventual full configuration
A common trap: vendors propose a stripped-down pilot — fewer modules, simplified workflows, lighter integration — to get to a fast yes. Then at renewal, the question becomes "what would the real version cost and look like?" and the conversation starts over.
If your final-state implementation will include observation forms, learning groups, AI-drafted feedback, and a portfolio module, your pilot should include all of them — even if at smaller scale. Otherwise you're not really evaluating the platform you'd buy.
Bring your trainer-of-record inside the pilot, not after it
If your system uses an instructional framework — Danielson, Marzano, your own — you almost certainly have someone who trains principals on it. Pull that person into the pilot. Have them help configure the rubric mappings, sit in on the first round of observations, co-author the AI feedback prompts.
Pilots where the trainer shows up after launch always feel grafted-on. Pilots where the trainer is a co-architect compound. Teachers see one coherent system, not a platform plus a separate workshop.
Bake in a renewal decision date and explicit go/no-go criteria
Decide today, in writing, what you'll need to see by month nine to renew. Decide what would make it a no. Put both in the pilot agreement.
This sounds defensive. It's actually generous to both sides. It tells the vendor exactly what to optimize for. It tells your team what to track. And it makes the renewal conversation a five-minute review of evidence instead of a six-week political negotiation.
If a vendor pushes back on this, that's information.
Demand clarity on what AI will and won't do in teacher growth
The market is flooded with platforms that quietly use AI to score teachers, label performance levels, or generate ratings — and even more that quietly use educator data to train models you'll never see. Most don't disclose any of it clearly. Many buy schools' enthusiasm for AI without explaining what they're actually doing under the hood.
Before you sign, ask three blunt questions:
- Does the AI ever assign a label, score, or rating to a teacher? The answer should be no.
- Is teacher or student personal data sent to a third-party model, retained, or used for training? The answer should be no.
- Who has the legal authorship of feedback the AI helps draft? The answer should be: the human admin, always.
For faith-based and mission-driven systems, this question is non-negotiable. The technology has to support the educator, not replace the judgment.
The strongest pilots aren't the ones with the biggest budgets
They're the ones where the school system walked in already knowing what success looked like, who would own each piece, and how the renewal would be decided. That clarity changes everything — including what's possible commercially. When a system shows up with concrete success criteria, vendors can structure pilots in ways they otherwise can't.
If any of this is useful as you evaluate options this spring, I'm always happy to compare notes.
— Ilya
