How Mindstake works
Mindstake follows a simple process. The platform supports the steps, but the content comes from you. You define what matters and what good looks like.
The process at a glance
Most organisations work through seven steps. Some are done once at the start; others repeat each time new evidence arrives.
Step
Your input
What you get
Agree what matters
Your purpose, your obligations, and who the reporting is for
A clear brief
Build the rubric
The areas that matter to you, and what strong and weak look like in each
A shared scoring guide
Bring in what you hold
Reports, business cases, surveys, interviews, recordings, feedback
One organised, dated evidence base
Read and score
Nothing extra - the AI reads everything against your rubric
Every relevant passage scored, with the exact words behind it
Check the picture
A walk through how the evidence has been brought together
A picture people have checked and accepted
See the results
The questions you want answered, for the period you choose
A clear view of progress, with every point traceable to its source
Keep it current
New evidence, whenever it arrives
Reporting that stays up to date and says what changed
Agree what matters
Every Mindstake engagement starts with a clear brief: what are you trying to understand, who is the reporting for, and what does good look like? This step sets the scope so that everything that follows stays focused.
The same platform adapts to very different contexts. An iwi investment fund, a health board listening to whānau, and a programme supporting founders all follow the same seven steps - but the rubric and the reporting reflect what each organisation cares about.
What you agree at this stage
- What you want to understand, and why
- Who reads the reporting, and what they need from it
- The evidence you already hold, and what is still to come
- What good looks like - not just the outcome you hope for, but what the reporting needs to withstand
Mindstake sets up a dedicated space for your organisation, so the rubric, the evidence, and the reporting all stay linked to the brief from the start.
Build the rubric
A rubric is a short scoring guide that defines what strong, moderate, and weak look like in each area that matters to you. We build it with you, in your own language. It gives Mindstake a shared standard, so every document and every voice is read the same way.
A good rubric uses plain language, is grounded in what can be seen in the evidence, and avoids broad terms like "good outcomes" without defining what that means in your context.
Scoring priorities
Mindstake crosses the areas that matter to you with a set of common questions - called priorities - to keep scoring structured and repeatable. These are configured per organisation and might include intent, design, practice, outcomes, and measurement, or whatever lens fits your work.
Area lenses
Each area gets a short set of notes - a lens - that explains what strong and weak look like there, and what kind of evidence speaks to it. This keeps the rubric flexible: you can change an area without rebuilding the whole framework.
Confidence ratings
Separate from the score itself, a confidence rating describes how clear and well-supported the evidence is. Low confidence often means a gap worth asking about - not necessarily a low score.
Bring in what you hold
Mindstake works from the evidence you already have - reports, business cases, survey responses, interviews, recordings of hui, feedback. There is no new form to fill in and nothing to rewrite.
Each item is placed against what it is about and the period it describes, so you can look at any stretch of time and see only the evidence from it. New material can be added straight from the report, whenever it arrives.
What Mindstake can read
- Documents and reports, including the charts and images inside them
- Survey exports and spreadsheets
- Interview and hui recordings, transcribed with care for te reo Māori
Introductions that identify a speaker are removed from recordings, and your evidence is held in its own space - never mixed with another organisation's.
Read and score
Mindstake's AI reads every item against your rubric. It finds the passages that matter for each area, scores them, and keeps the exact words behind every score. Nothing it produces reaches you without the passage it came from.
Because everything is read the same way, using the same definitions, a short paper and a long one - or a survey answer and an hour of kōrero - can sit side by side. This consistency is what makes the picture meaningful later.
AI that shows its working
The AI is constrained by your rubric - it does not invent areas or decide what matters. It works through your priorities and lenses, which lets it find relevant evidence described in different words across very different material.
Where the evidence says nothing about an area, the AI says so rather than guessing. A gap is shown as a gap, never as a poor score.
The AI does the reading. It does not decide the result - that comes from the steps that follow, where people check the picture and plain arithmetic brings it together.
Check the picture
Once everything is read, Mindstake brings together the passages that make the same point - across documents, across voices, and across time. You can see what is said often, what is said once, and where the evidence disagrees.
This is where people stay in charge. The picture is walked through and accepted before it reaches a report, so nothing appears that a person has not looked at.
What this step produces
- The key points in the evidence, each backed by the passages that make it
- A clear sense of how widely each point is shared
- Places where the evidence disagrees, kept in view rather than smoothed over
- Areas where there is little or nothing to go on yet
See the results
Plain arithmetic - not the AI - brings the scores together into a report. Newer evidence can carry more weight than older evidence, depending on your exact picture, so the reporting can move as things change on the ground. Every number opens back to the passages behind it.
Each investment or programme is measured against its own purpose, not ranked against the others. The question is always whether it is doing what it set out to do.
What you see
- What is being achieved, what is emerging, and what is not yet clear
- Short written summaries, where every sentence points to its source
- The voices and passages behind every finding, a click away
- Any period you choose, from the latest quarter to the full history
Keep it current
New evidence can be added whenever it arrives. It is read, checked, and brought into the picture, and the report says what changed, since when, and why. Nothing you have already seen moves quietly.
Where the evidence is thin, the report says so. Those gaps shape the next conversation - the next hui, the next survey, the next report - so understanding grows where it is needed most. The system learns with you.
What you take away
- Reporting that stays current, without starting again each round
- A clear record of what changed, and the evidence that moved it
- Questions for the next conversation, focused on what is unclear or missing
- A sound basis for reporting to boards, funders, and communities
- A better sense of what evidence to gather next time
Same engine, different decisions
Every Mindstake configuration starts with the same question: what does your organisation value, and how well are your options delivering?
Raukawa
Consistent, values-based review of documents - using the simplest Mindstake configuration.
Ngāi Tahu
Seeing whether each investment is delivering what it set out to achieve, beyond the financial.
Tapuwae Roa
Measuring alignment between organisational values and outcomes across five pou, using multi-modal data processing.
Getting started
Most teams are up and running within a few weeks. We work alongside you to set up your first decision process - criteria, rubrics, and scoring - then hand it over so you can run it yourself.
An initial conversation to understand your decision context
We help you define criteria and build your rubric
Your team runs the process on your own Mindstake instance