The platform

An operational intelligence workspace, built for honest evidence.

The software designed to support an engagement: discovery, evidence, workflow maps, comparisons, baselines, approved pilots and reports, in one place. It is designed so that what is known, what is assumed and what is still unknown are never confused.

It is designed for supported engagements and is not presented as a mature self-service product. The screenshots on this page are real captures of the workspace using a fictional demonstration company.

What it does

Eight modules, one record of the work

Operational discovery
Company intake where every answer is marked known, unknown, hypothesis or verified. Interview and shadowing sessions follow structured prompts.
Evidence and assumptions
Each observation records its source — operator said, directly observed, system record or analyst inference — along with sensitivity and consent.
Workflow maps
An editable step list with actors, systems, time estimates and branches, drawn as a diagram. Only a reviewer can mark a map validated.
Intervention lab
Options compared side by side: process change, deterministic software, AI assist, AI with approval, and no change. A transparent scoring rubric, no black-box score.
KPI baselines
Measures with units and direction, recorded per period with sample size and source. Missing values stay missing; corrections are new records, not edits.
Pilot approvals
Draft, proposed, approved, running, review, scaled or stopped. A sensitive pilot cannot be approved by the person who proposed it.
Reports and playbooks
Versioned reports assembled from the records, with every line labelled by status. Playbooks only from a validated workflow with a measured pilot.
Portfolio visibility
Coverage, evidence health and pilot outcomes across the companies a person is allowed to see — never raw records from one company shown under another.

Workflow maps

The process as it really runs

Steps, actors, systems, time estimates and branches, drawn from an editable list. A map starts as a working hypothesis and is marked validated only by a reviewer.

Workflow map of a maintenance request process in five steps — tenant message received, property manager triage, supplier assignment, repair completion verification and customer update — with friction points marked on two steps. — A five-step maintenance workflow with friction recorded on two steps.Illustrative synthetic demonstration — not a real customer.

Evidence and friction

Every claim says how we know it

Friction starts as a hypothesis and becomes verified only with linked evidence. Each observation records whether an operator said it, it was observed directly, it came from a system record or it is an analyst's inference.

Two friction points, both marked as hypotheses: supplier assignment batched at end of day (high severity) and completion not verified before closing (medium severity), each with a next question and linked synthetic evidence. — Friction points with severity, frequency, confidence and the next question to answer.Illustrative synthetic demonstration — not a real customer.

Intervention lab

AI competes with simpler options

Each friction point gets several options: a process change, deterministic software, AI assistance, AI with human approval, and no change. Costs and benefits are entered with their basis; when evidence is missing the comparison says so instead of inventing a payback.

Intervention lab table comparing four options for one friction point — do nothing, rule-based software, a process change and AI with human approval — with costs, benefits and payback all marked unknown or insufficient evidence. — Four options for one friction point, with costs and payback still marked unknown.Illustrative synthetic demonstration — not a real customer.

KPI baselines

A baseline before any change

Measures have a unit, a direction and a denominator. Each observation records its period, sample size, source and reliability. Missing values stay missing, and corrections are new records rather than silent edits.

KPI baseline for median time from tenant message to supplier assignment: 7.5 hours over September from a sample of 40, labelled synthetic, low reliability, with confounders listed and no post-pilot result recorded. — A baseline of 7.5 hours from 40 jobs, marked low reliability with its confounders.Illustrative synthetic demonstration — not a real customer.

Pilot approvals

Changes are scoped and approved

A pilot moves through draft, proposed, approved, running and review before it can be scaled or stopped. It carries acceptance criteria, stop conditions and a rollback plan. For sensitive pilots the approver cannot be the person who proposed it — enforced in the database, not just the interface.

Pilot in draft stage with its scope: the option, workflow, KPI, business owner, acceptance criteria, stop conditions, control strategy and rollback plan, with unknown fields left explicitly unknown. — A draft pilot scope: unknown fields are left explicitly unknown.Illustrative synthetic demonstration — not a real customer.

Reports and playbooks

Reports built from the records

Reports are assembled from the underlying records and versioned, with every line labelled by status. A playbook can only be created from a validated workflow with a measured pilot, and it is applied elsewhere as a checklist, not an assumption.

Operational X-Ray report for the synthetic demo company, with the workflow steps, evidence base and bottlenecks each labelled hypothesis, operator said or synthetic. — An Operational X-Ray report with each statement labelled by how it is known.Illustrative synthetic demonstration — not a real customer.

By design

What the platform deliberately does

Separates each client company's records; people only see the companies they have been given access to.
Tracks consent per company and per purpose: observation, storage, AI processing and publication.
Records decisions — approvals, stops, reversals — with a note, in an audit trail users cannot edit.
Exports through the same access rules, and logs every export.

And does not

What it is not

Records how work happens in your systems. The adapter inventory does not connect to external systems today.
Not an autonomous agent. Optional AI assistance only suggests; people decide and approve.
Not sold as a self-serve product with a price list. It is designed to support an engagement.
Not a source of guaranteed savings. It reports what the evidence supports, including when that is little.

How data is protected

See your own workflow in it.

The platform supports a proposed engagement, starting with one workflow. Tell us which one you would like to understand.

Discuss a workflow Explore the platform