Build Brief

Step 1 of 4

Choose one company and one problem

Select a report saved here, import a discovery backup from another device, or start one blank case. A company is a prospect until an agreement is signed.

Do not repeat discovery, and do not guess.

This brief is for the single decision we could build and test first. It is not a company overview and it is not a request for every possible use case.

Existing discovery report available here If the report was recorded on a phone, download its discovery backup there and import the JSON file above.

Step 2 of 4

Define the decision

Describe what actually happens. Do not ask the customer to design our product. Stay with one recent episode and record the available choices.

The operating moment

Required before anything is built

What triggers the decision? coreThe first observable sign and the moment someone must decide what to do.

Who makes the decision today? coreUse a role, not necessarily a person’s name.

How quickly must they act?

Every decider in one episode, in orderInclude experts called in, set procedures and installed tools such as Squint or Copilot agents.

Which action should be compared, and against which installed approach? core

What actions are genuinely available then? coreDo not add hypothetical actions that the team could not actually take.

Which actions require approval or are prohibited?

Decision family

Annual frequency and eligible subset

Same decision per line per week core

Lines or sites with this decision

One comparable episode

Beginning, action and result

When does the episode begin?

When does it end?

What must be known before the action? core

What does the decider expect before acting, and is it written anywhere?

What result must be observed, and after how long? core

Step 3 of 4

Map the evidence

Ask where facts are recorded, not whether they “have data”. A sample export and field dictionary are enough for the first technical check.

Source systems

Names and owners, not architecture diagrams

Where is the pre-action state recorded? core

Where is the action recorded?

Where is the result recorded?

How are decisions recorded today?Typed in later means the record time differs from the decision time.

What joins the records?

How is the chosen action acknowledged in its authoritative system?

History and access

Unknown is a useful answer

How many comparable episodes exist?

Were different actions tried?

Can they share a small sample export? core

Who approves access?

What important facts exist only in people’s heads?

What data and derived learning can we use, and for which purpose?

What second company, site or system could test reuse?

Only if named worker or other sensitive data may be involved

Does the first product need named worker data?

Step 4 of 4

Define the proof

The product must improve a real decision beyond today’s workflow. Name the incumbent, the remaining gap and the economic test.

What exists already

Do not rebuild a dashboard they own

How is the decision made today?

Which tools already help?

What simple baseline should Kalyr beat?

What decision remains unsupported after those tools have done their job? core

The first product

One surface, one user, one decision

Where should the result appear?

What evidence must appear beside it?

When should Kalyr ask for missing information?

When must it abstain or escalate?

Operational and economic acceptance

This is how we earn the next decision

Operational metric core

Who owns it?

How does improvement become money? core

Annual value at risk and decision-controllable share

Finance owner and source of unit values

How will we estimate improvement against the alternative?

Minimum result that would justify continued use

What commercial decision follows a successful shadow test?

Engagement track

Where does the buyer sit?

Economic buyer and budget path

Annual minimum + qualified episode usage price reaction

Conversion condition and named next meeting

Review

One build case, without the repetition

The brief can be sent while incomplete. Missing blockers remain visible, so Minara does not have to invent answers or delay sharing useful evidence.

Decision Loop Build Brief