O OrgiAi
Process discovery · Effort analysis · Organisational insight

How the work actually runs.

The only people who can describe your processes are the ones running them every day. OrgiAi interviews them, turns what they say into something you can work with, and keeps every finding tied to the sentence it came from.

Open the app See how it works Can be installed on your own servers.

The gap between the process document and the work

In most organisations, processes are written once and never read again. The work changes in the years that follow; the document does not. What is left is a procedure nobody follows and a way of working nobody wrote down.

Work drifts outside the definition

Tasks that appear in no job description become the job itself. The answer to "who does what" lives in people's heads, not in the org chart.

Waiting is invisible

How long a step takes can be measured. How long it waits for an approval is not — and that is where most of the week goes.

Knowledge sits with one person

Every organisation has a few people whose holiday stops something. Everyone knows who they are. Nowhere does it say so.

Four steps

The discovery phase of a consulting engagement, structured — and holding the same standard of evidence.

01

Interview

The employee explains in their own words, at their own pace. The assistant asks the follow-ups: how long does that take, how often, which system, who does it go to next.

02

Extraction

The conversation becomes structured work items: the step, its duration and frequency, the systems touched, the problems described, and who the work is handed to.

03

Review

Every extraction reaches a person with the employee's own sentence beside it. Nothing moves forward unapproved.

04

Analysis

Approved findings produce process maps, effort distribution, the collaboration network and bottleneck analysis.

No unapproved extraction reaches a report.

This is the product's most important constraint. What forms the report is not what the model said but what a person confirmed — because a document you put in front of a client is an opinion until its source can be shown.

What you get

All of it computed from the same interviews, so all of it agrees with itself, and every figure opens back to where it came from.

Department board

Where each unit's weekly hours go, how much of it is spent waiting on something outside their control, and which department they are held up by.

"Customs spends 54 of its 75 weekly hours waiting on somebody else."

End-to-end process map

A swimlane diagram of the company's work from start to finish: which unit does what, where work crosses a boundary, where it stops.

Downloads as one file, opens in a browser, prints.

Collaboration network

Who waits on whom — and for what. Click a person: what they receive, what they hand on, and the work behind every link.

"16 people wait on this person; they hand work to 3."

Bottlenecks

The blockages employees describe themselves, grouped by recurring theme. One person's complaint is separated from four departments' shared problem.

"Waiting for approval — 5 times, 2 roles, 4 hours a week."

Job description gap

Work people described that appears in no job description; and responsibilities the job description lists that nobody mentioned.

Direct input for hiring and role design.

Effort distribution

Weekly hours by unit, split into manual work, work inside a system, and waiting. Items with no stated duration or frequency are left out.

Totals are deliberately conservative; coverage is stated separately.

Why it can be trusted

The easiest thing an analysis tool can do is produce confident numbers with no source behind them. OrgiAi is built around not doing that.

Traceability
Every process step is tied to the interview item it came from. A manager who disputes a figure can follow it back to the sentence that produced it.
Human approval
No finding the model produced enters a report, a map or a chart until a person has approved it.
No invention
Work with no stated duration counts as zero hours; it is not estimated. A name that is not in the employee records is not added to the network. Coverage is stated on every screen — how many people the findings came from is never hidden.
Deterministic analysis
Process clustering and bottleneck grouping rest on explainable rules, not on a model. The same interviews give the same result on every run, so two reports taken two months apart can be compared.
Tenant isolation
Customer data is separated at the database level by row-level security policies. If the application connects with a privilege that could bypass those policies, it refuses to start.
Where the data lives
It can be installed on your own servers, and where the data goes — including which language model provider, if any — is your decision. Cross-border transfer obligations are planned around that choice rather than discovered afterwards.

How does the work in your company actually run?

Start with one pilot department. A handful of interviews is enough for a first process map and a first list of bottlenecks.