HABLA Intelligence^™

Signals are only useful if you know what to do next.

HABLA^™ hears what the people an organization serves keep raising, and never learns who raised it. HABLA Intelligence^™ is the layer that comes after knowing. It lets a leadership team decide what deserves attention, record what the organization chose to do, and watch the signal through the period that follows.

Request a walkthrough See what it does

More intelligence does not require more surveillance

Privacy happens before intelligence.

What a figure passes through before the layer can read it

The order is the architecture. The separation of stores comes first, then the reporting floors and the suppression that follows them, and only what survives all of that is what the analysis layer reads. Its input is defined as the output of the step above it, which is what puts every stage that protects a person ahead of the stage that analyses anything. There is no arrangement of these six in which the analysis reaches a person.

This is the anonymous activity the whole product is built on: what a person does in HABLA arrives with no name, no account and no identifier that could be joined to one. There is no individual row to be read later, so the protection is a property of the schema rather than of a policy about it.
Separate data stores, the grant model between them, and the absence of any key that would join the anonymous domain to the contact domain. This is the one stage of the six a customer could audit today, and /security/ states it field by field.
Two floors, applied before anything is a row a reader can see: an activity floor under every figure, and a population floor that decides whether a placement or an audience pack may be a line of its own at all. A cell that would identify by arithmetic is suppressed, and its neighbours with it.
What survives both floors, with the methodology panel that says what the figure is and what it is not. This is the whole of what a leadership team can open on screen, and the whole of what anything downstream may read.
The analysis layer begins here and nowhere earlier. Its input is defined as the output of the step above it, and it holds no connection to any database, so there is no earlier stage for it to reach back into.
The three-part reading, the response ladder and the record leadership writes against them. Everything generated carries a mark of its own, and the caveat naming the limits of the figure travels beside it.

Every stage that protects a person comes before the stage that analyses anything, and the analysis layer’s input is defined as the output of the step above it. The order is the architecture: there is no arrangement of these six in which the layer reaches a person, since by the time it is handed anything there is no person left in it.

Anonymous activity is what arrives; HABLA’s privacy architecture and its reporting thresholds run on it before anything is reportable at all; and the analysis layer sees only what is left after that — the same figures a leadership team’s own report renders.

AI analyzes the signal. Never the person.

That sentence is about HABLA Intelligence — the layer this page describes. It is not about HABLA AI, the assistant in the corner of this website, which is a different thing and is running now: it answers questions about HABLA out of HABLA’s own published documentation, and it works on this website rather than inside a customer’s deployment.

HABLA Intelligence reads the same suppressed aggregates a report renders, and nothing else.

The layer reads reportable aggregates and never the anonymous activity underneath them. It will not diagnose a person, name an employee or a student, score an individual’s risk, or decide anything on an organization’s behalf. Leadership reads it, leadership decides, and leadership stays responsible for the decision.

It will never:

Identify an employee or a student
Build a profile of an individual
Diagnose a mental-health condition
Calculate a risk score for a person
Decide who is disciplined, dismissed, promoted or investigated
Reconstruct one person’s path through HABLA
Predict where a signal is going — HABLA describes reporting periods that have closed

What the layer is sent, and what never reaches it

A sentence about what an analysis layer never sees is worth exactly what the boundary underneath it is worth, so the boundary is published here rather than summarized. Everything that crosses it has already been through both floors that apply to anonymous activity.

Crosses

The topic key — HABLA’s own identifier for a topic, never a customer’s wording for it
The aggregate values that survived suppression, and whether each one cleared the floors or was withheld
The labels of the reporting periods being compared
The sector, and a coarse size band
Segments as opaque labels, with the organization’s own names put back on screen by the application afterwards
The comparability state, and the caveats that print beside a figure

Never crosses

The organization’s name
Any name a customer wrote — for a site, a placement, an audience pack, or an alias on a topic
Any administrator’s name
The leadership notes on a decision record, which stay in the console
Any value beneath a floor, in any form, including the fact that one is there
Any figure that has not been through the reporting engine’s own suppression

A list like that is worth what the contract under it is worth, so here is the contract HABLA will sign: a named provider, undertaking not to train on what is sent, and stating in days how long it keeps it. HABLA has not signed one yet — which means an institution evaluating HABLA today can require this list in its own agreement first, and hold the company to it.

What you are comparing against

A signal on its own is a number with nothing beside it. Twenty of something is not a lot or a little until you know what twenty is next to. HABLA compares an organization against privacy-safe sector and national aggregates, so a leadership team can tell whether what they are seeing is theirs or everyone’s — which is the difference between “we have a problem” and “there is a problem”, and they are not the same meeting.

Every comparison row carries its denominator. A bare percentage reads as a share of people, and HABLA does not measure people — so no row is published without the count it is a share of. Two organizations can also differ for reasons that are about the deployments rather than the people — one may be older, better known, or put in more places than the other — and a comparison says so beside itself rather than in a footnote.

The floors are the reason to believe the number, and they are in force before the first figure. A cross-organization figure is published only where at least ten independent organizations contribute to it, where no single one of them exceeds forty per cent of it, and where the largest three together do not dominate it. Contributing is a separate term an organization agrees to, can decline without penalty, and can withdraw from. No cross-organization figure has been published yet, and none will be until a figure can clear all three — the combined limit is a number HABLA has not set, and the code refuses to publish rather than assume one.

What it does

Most listening products stop at the first stage. HABLA Intelligence is the name for the four after it, and each one says which side of the pricing line it falls on.

Signal → Insight → Strategy → Action → Learn

Signal

What is repeatedly reaching the surface, as aggregate activity that has cleared both reporting floors.

Core

Insight

Trends, persistence, timing and context, read without identifying anyone.

Core view, generated reading paid

Strategy

Possible organizational responses, or ways a practice that appears to be going well could be tried elsewhere.

Paid

Action

What leadership decides to do, the administrator who owns it, and when it will be reviewed.

Core

Learn

Continuing to watch the signal, and seeing what the aggregate did in the period that followed.

Core

Signal: Follow a topic across reporting periods against a criterion you choose from a fixed set of five rather than type. It is not an alert and will never be called one: it is read when a reporting period closes, it sends nothing, it notifies nobody, and it follows a topic and never a team, a shift or a department. Appreciation topics are treated on the same terms as concerns — same floors, same suppression, same aggregate-only handling — and are never scored against a unit and never used to offset a concern.
Insight: The counted view of a topic is core; the generated reading of it is not. A generated reading arrives in three parts, in this order, held there by the response format rather than asked of a model: what stands out, questions worth exploring, organizational considerations. Part two is written as questions so it cannot be read as a conclusion, and the sentence about what HABLA cannot conclude sits beside it in the same visual weight.
Strategy: Options arrive as a ladder — low-effort, operational, leadership, organization-wide — cheapest to most demanding and on one screen, so an organization that cannot afford the expensive option is still looking at something it can do this month. Each carries eight things: why to consider it, a suggested owner, the effort, the duration, implementation steps, questions to consider, possible measures, and privacy considerations. A suggested owner is a role or a department and never a person; no option offers prevalence as a measure of success; privacy considerations is a real field and not a disclaimer slot; and recording a different decision is a first-class control, so the record keeps that the choice was not HABLA’s.
Action: Every signal can carry a leadership record, and it comes with the core platform rather than the add-on. Kept over years it is an institutional history — what surfaced, what leadership knew, what leadership decided, what was done, what followed — and it is about the organization rather than about anyone in it.
Learn: HABLA compares reportable aggregate activity in the period before a recorded action with the period after it, and says plainly that it is comparing and not explaining. The record ends: three years for a decision, a watch and a strategy by default, with the organization’s own contract setting the period, and twenty-five months for a generated reading, so it cannot outlive the aggregates it describes.

Eleven fields, and these are all of them:

The signal the decision is about
The leadership question it answers
The decision
The owner — a named administrator from the organization’s own console
The action
The date the action starts
The completion state
Whether a standing watch is open on the signal
Which strategy record the action was taken from
Whether that strategy was generated rather than written by a person
Leadership notes, internal to the organization and never sent to a model

Those eleven are the record’s content. The row carries two more things and both are administrative: the reporting period it belongs to, and the dates HABLA needs in order to retire it. Every one of the eleven is written by an administrator signed in to the organization’s own console. The anonymous plane and the configuration store are separate data domains with no join key, which is what keeps this record organization-side. When that administrator’s account is deactivated the record keeps the decision and drops the person: the name is removed, and it is replaced by the role only where more than one account holds that role — otherwise the decision keeps only the organization.

HABLA reports what happened after an action. It does not claim the action caused the change.

Four things are required of every such sentence: the subject is the signal and not the people, the verb describes the movement rather than explaining it, the reporting periods compared are named, and the relation to the action is time.

Every record is written by a named administrator taking an action in a console. The layer notifies, escalates and contacts nobody, inside the organization or outside it. A person, a team and a unit are never scored — no rank, no rating, no index. Generated output is an artefact of a request rather than a new column about an organization’s people, and it carries a mark of its own wherever it appears, so a reader years later can tell which words a person wrote.

HABLA informs. Humans decide.

What is core, and what is the add-on

The framework is core. The model is the add-on.

Anything HABLA can state as a fact about an organization’s own data belongs to the core platform. Anything HABLA generates — language, interpretation, questions, a draft strategy — is the paid layer.

Core platform, at no extra charge

Reportable aggregate signals, and the methodology panel that ships with every one of them
Signal Watch, and every one of its criteria
Positive-signal reporting — an organization sees what is going well on the same terms as what is not
The leadership decision record, and action tracking against it
The before-and-after comparison for a recorded action, with the sentence that refuses to call it cause
Organizational memory as a record that can be read
Privacy, crisis resources, anonymity and accessibility — never priced, on either side of this line

The paid add-on

The add-on’s feature set is named Strategy Intelligence. It is one layer, priced once a year, and it never charges for a question, an answer or a strategy.

AI Insight — the generated reading of a signal
Strategy Builder — generated response options at every level of effort
The generated exploration behind a positive signal, and a drafted replication strategy
The generated interpretation of a before-and-after comparison
Organizational memory answered in prose rather than read as a record
Ask HABLA — a bounded set of executive questions, never an open box

Three of those sit in core against the founder’s own section layout, and the test that moved each of them is the same: a customer without the add-on must not get a less honest picture of their own signal.

Detecting a positive signal is core. Gating it would hand a non-subscriber everything going wrong and nothing going right — not a smaller product but a biased one.
Seeing what followed an action is core. The framework is signal, decision, action, and then learning; selling the last step sells the framework.
Reading the record is core. A record that can be written and not read is not a record.

Nothing about privacy sits on either side of that line. Both reporting floors, the suppression rule and the required methodology panel ship identically to a customer with the add-on and a customer without one. A threshold sold for money is not a threshold.

Tell us what it has to do.

Any vendor will tell you what their product does. Ask this one what it is not permitted to do — take the refusals above into an agreement, and hold HABLA to them before a single figure is ever generated. Tell us what this layer has to do for your organization, and we will walk you through the rest of it.

Request a walkthrough See plans and prices