Services · Governed Method Packs · Continuous governance

Your company's method becomes a rule the AI reads in every session.

Measured every month and reviewed every quarter. A pack is a versioned object, installed once in your repository and on your platform, and kept valid every month while AI tools change every few weeks.

Start with the Base packSee the six layers

A forty-minute conversation settles whether there is enough operation to measure, and you leave it knowing what your baseline would be.

About the h|ve Platform

The measurement comes from the recorded operation, not from a side spreadsheet.

Every h|ve service runs on the same platform. Here it is what makes the monthly reading honest: the reviewed rules come back as policy in execution, the authority matrix and the gates take effect in the project, and the six monthly readings come out of what the operation already recorded. Without the platform the method works, and the measurement gets weaker.

What this service touches

Map · context
Flow · orchestration
Guard · barrier
Judge · deviation
Ledger · record

Three of the pack's six layers live on the platform: the authority matrix per project, the gates at each passage and the instrumentation that reads cost, deviation and decision.

What stays on the platform after this service

What the other services installed, kept valid: the rule the tool reads, the authority matrix with the real roles, the gates with the artifacts the team produces and the monthly reading frozen against the baseline. It is the service that keeps what Scan & Go and FDE Velocity left behind from aging at the pace of the market.

The problem

After the rollout, three things happen at the same time, and none of them has an owner.

Governance written once ages. The rules describe a way of working that has changed, and the deviation only shows up once it has already become a problem. And when the question comes from outside, from a client, a risk area or an auditor, the answer depends on someone rebuilding the history by hand.

The billThe AI bill varies and nobody can explain it

A more expensive month is not a problem of absolute value, it is a problem of not knowing the cause. What has no known cause has no forecast and no ceiling.

The rulesThe policy stops describing the work

It stays in the repository, intact and false, and everyone knows it is false, which is worse than having no policy.

The outside questionSomeone asks, in writing, how you govern AI

How much of what you deliver is made by AI, who approved it, and how you prove it. Whoever has no answer ready improvises, and improvising on that question costs the contract.

If you answer for the technology, what is missing is a rule that actually holds and a way to see when it did not.
If you answer for the cost, what is missing is the cost per delivery, per model, with the variation explained.
If you answer for the risk, what is missing is the trail that says who decided, under which authority, on what basis.

None of the three has an owner. This service gives it one.

Policy and behavior

The rule that lives in the repository is read by the AI tool in every session, every time, without anyone remembering it.

That is the difference between policy and behavior, and it is the center of this service. The rule is not a PDF the team swears it has read: it is a versioned file the tool consults before writing. Switching AI tools costs rewriting one file, not redoing the governance.

What a pack is

A versioned object with six layers.

The policy the AI tool reads, the authority matrix of who approves what, the gates at each passage, the templates for specification and decision, the instrumentation that measures, and the history that records every change in the five before it. If it does not have all six, it is a presentation, not a pack.

Policy
The rules for using AI that hold in that repository, read by the AI tool itself.
Where it lives
In your repository, versioned in git
Who edits it
You approve, we propose
Authority
Who approves what, and what nobody approves.
Where it lives
On the platform, per project
Who edits it
You, with a record of who changed it
Gates
The artifact required at each passage of the flow.
Where it lives
On the platform, in the stages of the flow
Who edits it
You approve, the architect overrides with a reason
Templates
The models for specification, PR and decision record.
Where it lives
In your repository, as templates
Who edits it
Your team
Instrumentation
Which indicators are read, from where, with what cut. Layers 1 to 4 are configuration, and configuration is done once. This is the one that produces a new reading every month.
Where it lives
On the platform, with output by webhook and API
Who edits it
We define, you approve
History
The record of every change in the five layers above, with semantic versioning.
Where it lives
Next to the policy, in the repository
Who edits it
We write, you read

2.0

Major version when the authority matrix or the gates change, because who can do what changes.

2.3

Minor version when the policy or the templates change, because how things are done changes.

2.3.1

Patch when only the text changes, with no change in effect.

The useful answer to the outside question is not “we have governance”. It is “at version 2.3, in force since March, this decision required this approver, and here is the record”. The version is what turns the trail into proof.

The catalog, today

Pack 01Base

The method exists on paper and not in the work.

It is where you start. The AI usage policy stops being a document the team swears it has read and becomes the file the tool consults before writing, and each passage of the flow starts requiring an artifact to move forward.

Carries the weight: Policy and gates

What it installs

The policy the AI tool reads in every session, written in the vocabulary your team already uses and versioned in your repository.
The gates at each passage of the flow, with the artifact required at each one, built on top of the templates the team already has.

What changes

The written rule becomes behavior: the tool follows the policy without anyone remembering it, and whoever joins the team finds the method where the tool reads it, not in a PDF.

It fits when The method exists on paper and not in the work, or the policy has stopped describing what the team does.

Pack 02Cost

The AI bill varies and nobody can explain it.

The cost reading stops being an estimate. The pack reads your model vendor's invoice, breaks the cost down by delivery and by model, and compares each month against a baseline frozen at the start.

Carries the weight: Instrumentation

What it installs

The instrumentation that reads the cost per delivery, broken down by model, straight from your vendor's invoice.
The baseline, frozen and signed by both parties, against which every month is read.
The monthly reading that explains the variation, with up to three recommendations, each with a number next to it.

What changes

The AI bill stops being a surprise: the variation arrives explained against the baseline every month, and finance answers without needing technology to rebuild the month by hand.

It fits when The bill varies and nobody can explain it, or finance asks and technology does not have the answer.

Part A · the installation

Installing the pack is a project with a beginning and an end.

Fifteen working days, four to six calendar weeks, once. A service that installs governance by gate has to be delivered by gate, and each one is signed by someone on your side. The difference between four and six weeks lives on your side: the access, the calendar of whoever approves the authority matrix and the volume of recorded operation.

A0Prerequisiteshalf a day, before the clock starts
Gate · entryRead access to the repository and to the recorded operation, the authority approver identified, and the language model key confirmed as yours. If it does not pass, the clock does not start.
What comes outThe start date, which counts from the access being granted.
A1Reading the current state3 days · weeks 1 and 2
Gate · the portraitThe current state as described is correct, including where it is uncomfortable. The technical lead signs. If it does not pass, it goes back to reading.
What comes outA written portrait, with no recommendation at all, because recommending before reading is selling opinion.
A2Selection and cut1 day · week 2
Gate · the cutThe scope of version 1.0 is approved, with what stays out declared. Whoever approves the authority matrix signs.
What comes outWhich pack, which repositories and teams. A pack that tries to cover the whole company in 1.0 does not take effect anywhere.
A3Adapting the six layers6 days · weeks 3 and 4
Gate · the layersThe six layers accepted one by one. If one fails, iterate on it, never on the whole package.
What comes outPolicy in the company's vocabulary, authority with the real roles, gates with the artifacts the team already produces, templates on top of the templates that already exist.
A4Activation2 days · weeks 4 and 5
Gate · in forceThe policy pull request has been merged and the authority matrix is active on the platform. Verifiable in thirty seconds on one screen. Up to here the gates produced documents. This one produces effect.
What comes outThe policy enters the repository by pull request, never by direct commit, because the first demonstration of the method is the method itself being followed.
A5Baseline2.5 days · weeks 5 and 6
Gate · the baselineThe first reading of the six indicators, frozen and signed by both parties. Without it, no future reading means anything, and the third month becomes a conversation about perception.
What comes outA pack at version 1.0, in force, with a signed baseline. Only now does the cycle start.

Part B · the cycle

Four rhythms, and only one of them is a meeting.

Measuring is continuous and automatic, because the measurement comes out of the recorded operation. Following up is monthly and produces a one-page reading, not a presentation. Adjusting is monthly for the recommendations and quarterly for what requires a new version, because changing the authority matrix every month is the opposite of governing. Reviewing is quarterly, and ends in a new version or in an explicit decision to change nothing.

Rhythm	What happens	Our effort	Your time
Continuous	A crossed threshold fires an event to the team's channel, with nobody watching a dashboard.	Zero	Nobody
Monthly	The reading of the six indicators against the baseline, with up to three prioritized recommendations, each with a number next to it.	Half a day	Ninety minutes
Quarterly	The pack review: what changed in the work and what needs to change in the six layers. Ends in a new version or in a recorded decision not to change.	Three days	Half a day
Annual	A full rereading, compared to the original baseline, and the decision about the year ahead. The annual governance report comes out.	Three days	The decision

No recommendation enters the monthly reading without a number next to it.

A recommendation without a number is opinion, and monthly opinion is not a reading. When the AI tool changes, the catalog pack is the trunk and yours is a branch of it: every change arrives as a proposal by pull request, which you accept or refuse. A new outside requirement, a regulation, a clause, does not wait for the quarter.

The six readings, from the recorded operation

Cost per delivery, broken down by model.The cost reading is a reading of your model vendor's invoice, not an estimate.
One line per change, with author and under which rule.How much of what shipped was made by AI, and who reviewed it.
The catalog of decisions, with author, authority and basis.Who decided what, months later, without reassembling it from chat and meetings.
What the project governs and what it does not govern, declared.The written boundary is what prevents the surprise in the outside question.
Where the approved specification and what the PR delivered drifted apart.The deviation flagged in the PR itself, before it becomes an incident.
How many times the gate was overridden, by whom and for what reason.The price of the override is transparency.

Each of them is verifiable on one screen and none depends on a side spreadsheet. This is the list that replaces the abstraction “KPIs and costs”.

What stays with you

Everything, except the trunk.

Five of the eight artifacts live in your repository and on your platform from day one, and remain yours if the follow-up ends. What you lose is the propagation: you stop receiving the change proposal when the tool changes, and from there the pack starts aging at the pace of the market.

The annual governance report
A document ready to be handed to a third party: clients, auditors, committees, the risk area. With the trail that supports each statement.
Why it mattersIt is the artifact that answers the outside question for you, and it is worth more than the twelve monthly readings combined.
The pack, layers 1 and 4
The rules file and the templates, in your repository, in git.
Who uses itThe AI tool, in every session.
The pack, layers 2, 3 and 5
The configuration of authority, gates and reading, on the platform, per project.
Who uses itWhoever approves, at each passage of the flow.
The pack changelog
Text with semantic versioning, next to the policy.
Who uses itWhoever needs to prove what held when.
The baseline
One page, signed by both parties.
Who uses itEvery future reading compares against it.
The monthly reading
One page: six indicators, variation and up to three recommendations, each with a number.
Who uses itThe sponsor and whoever approves the authority matrix.
The portrait of the current state
Eight pages at most, with no recommendation, delivered during the installation.
Who uses itWhoever will decide the cut of version 1.0.
The pack user guide
One page in the repository, updated at every major version.
Who uses itAnyone who joins the team later.

It is not an audit opinion. The trail serves an audit and does not play its part.
It is not a certification. h|ve issues no seal, and no artifact above replaces SOC 2, ISO 42001 or equivalent.
It is not development. Applying a recommendation is your team's work, or FDE Velocity's, as a separate service.
It is not operating the platform. The service speeds up the cycle and creates no dependency to run it.

Who it is for

The last service in the line, on purpose.

Whoever reaches it without going through the previous ones governs an operation that does not exist yet. Three conditions need to be present at the same time: an AI operation running with volume, someone who answers for one of the three pains, and measurement that is possible.

FitsWhen it fits: the four scares

The AI bill went up and nobody can explain the variation.
Something went into production that nobody remembers approving.
A client or a regulator asked, in writing, how the company governs AI.
The team doubled and the method that worked for six does not work for fifteen.

Does not fitAnd where we point

The AI operation is not running yet, or nothing is instrumentedThere is no baseline to measure. Scan & Go first.
None of the four scares has happened yetWithout a concrete pain, the monthly reading has nothing to be read against. Wait, or a small diagnosis for the technical lead to have the portrait before the scare arrives.
What you want is someone to apply the recommendationsThis is governance during execution, not execution. FDE Velocity, as a separate service and with the conflict declared in the monthly reading.

You do not know where AI generates value and need to decide where to invest.AI Value Discovery
Leadership needs to decide where AI comes in and what remains a person's decision.AI Leadership Lab
The use exists and there is no governed flow in production.Scan & Go
The bottleneck is capacity to deliver, and you want it delivered for you.FDE Velocity
The flow exists and what is missing is control, explained cost and proof.Governed Method Packs

Who delivers

The two partners, from the installation to the monthly reading.

The installation needs someone who knows how to design authority and gates, and the monthly reading needs someone who knows how to read the six indicators and prioritize. That is the two partners, from the portrait of the current state to the annual report.

André JB RodriguesCo-founder and Business Lead

Leads the portrait of the current state, the cut of version 1.0 with whoever approves the authority matrix, the monthly reading with the sponsor and the annual report.

Gustavo HartzCo-founder and Technology Lead

Designs the authority matrix and the gates, adapts the six layers, activates the pack on the platform and maintains the trunk when the AI tool changes.

Questions

The questions that come before the start.

The ones technology, finance and risk ask, and the silent one, which is how many meetings the team will have to sit through.

How many meetings will my team have to sit through?

Ninety minutes a month and half a day a quarter. The continuous rhythm is automatic: a crossed threshold fires an event to the channel where the team already works, with nobody watching a dashboard.

Does it work without the h|ve Platform?

The method works, and the measurement gets weaker: without the recorded operation, KPIs and costs depend on what the company manages to gather on the side. If nothing is instrumented, the right service is Scan & Go first.

Does it replace an audit?

No. This is governance during execution, which is different from auditing afterwards. The trail serves an audit, and the annual report is written to be handed to one, but it does not play its part. Refusing what we do not do is what gives credit to what we do.

What happens if we do not apply the recommendations?

Three consecutive monthly readings with no priority recommendation applied trigger a review conversation. Two paths come out of it: either someone who applies shows up, or the scope shrinks to something the company can absorb.

And when the AI tool changes?

That is the reason the pack is kept valid every month. Your AI governance has a short shelf life because the tools change every few weeks, and whoever keeps it valid is either someone on your team part-time, or us. The change arrives as a proposal by pull request, which you accept or refuse.

The next step

Start with the Base pack.

A forty-minute conversation settles whether there is enough operation to measure, and you leave it knowing what your baseline would be.

We reply within one business day.

Prefer to write? hello@hybridvelocity.com