A cinematic miniature business: service counter, inventory room, inspection gates and operating desk connected by golden data routes.

The Parity projects

A business full
of signals.
Connected by AI.

From a single sale to the bigger picture. Explore the intelligence behind more responsive business.

Step inside the systems ↘

Parity POS

Transactions become patterns.
Patterns become a better-prepared team.

Explore Parity POS

Explore. Change. Discover.

Change the situation.
Watch the thinking.

Choose a scenario. Follow the signals. See what AI gathers, reads and connects before a useful insight reaches a person.

Hospitality & point of saleIn development

Parity POS

Every sale tells a story.

See the rhythm of a venue. AI connects transactions, baskets and trading conditions to put the next useful decision in front of the right person.

◈Transactions84 orders

◷Basket mixCoffee + pastry

◇Trading window07:00–10:00

Trading floor / intelligence view

parity AI

Sales in focus84orders analysed

CoffeePastry

Sale recorded

＋32 paired baskets

Orders through the morningPattern in view

22%38%70%100%86%45%29%18%

EarlierRelative activity · peak = 100Later

32 paired baskets

Prepare the daily trading insight for the venue manager.

Trading insightInsight ready

08:30Peak trading window

The morning rush has a pattern.

In this example, 32 of 84 orders pair coffee with a pastry. The strongest demand falls between 08:00 and 09:00.

Suggested next movePrepare the coffee-and-pastry station before 08:00.

See the source signals ＋

84 example orders
32 paired baskets
Hourly sales pattern

Prepared forVenue manager

Where AI fits

AI analyses live POS data to surface business-critical information for operators. It maps relationships and trends across sales activity, turning the data into daily insights delivered to managers so they can spot changes, understand performance and act sooner.

Business case, operating model & development status

The business case

When the till, venue and back office disagree, staff inherit the problem. Parity POS is being built to give each part of the operation a clear responsibility, with a recoverable record of what happened.

Operating model

The design separates the till, the property coordinator and the central backend. The till retains work durably; the property coordinates shared state; the backend manages configuration and reporting. Connectivity determines which actions are available, and recovery must not create a second sale.

Where it stands

Current work proves bounded sale and recovery scenarios in simulation. Full operator acceptance, payment-provider integration, fiscal qualification and venue rollout remain separate steps.

Explore the operational foundation

Explore the workflow

CounterProperty connected

Order · Example 001

Morning service

Flat white1 × $5.00

Breakfast roll1 × $12.00

Total$17.00

Ready to pay

Order retained

A clear order, with its prices retained.

The operator sees the items and the bill together. Changes must retain the correct pricing basis.

CounterProperty connected

Order · Example 001

Morning service

Flat white1 × $5.00

Breakfast roll1 × $12.00

Total$17.00

Cash recorded

Payment linked to this bill

One sale. One durable record.

A simulated cash tender belongs to this bill. A repeated request must not become a second payment.

CounterChecking sale status

Order · Example 001

Morning service

Flat white1 × $5.00

Breakfast roll1 × $12.00

Total$17.00

Confirmation pending

Check existing operation before retrying

Uncertainty stays visible.

If a response is lost, the design calls for checking the existing operation before taking another payment.

CounterResult recovered

Order · Example 001

Morning service

Flat white1 × $5.00

Breakfast roll1 × $12.00

Total$17.00

Sale recorded once

Next order uses a new identity

Continue from a known result.

Recovery resolves the original operation. The next guest receives a distinct order, with earlier obligations retained.

Step 1 of 4 · Take the order

Stock & operational controlMVP in development

Inventory

See the shortage before the shelf does.

Follow stock across locations. Then add bookings, weather and events to understand what the movement means for the day ahead.

◈Main store72 units

◷Front bar18 units

◇Expected demand30 units

One stock story / every location

Main store72units

Front bar18units

Reserve stock24units

Suggested movement12 units

Expected demand30

Location × demand

Flag the shortfall and suggest a reviewed transfer.

Stock intelligenceInsight ready

12Unit gap at the front bar

The stock exists. It is in the wrong place.

The front bar holds 18 units against illustrative demand of 30. The main store holds 72, so the immediate issue is allocation.

Suggested next moveSuggest a 12-unit transfer for the stock manager to review.

See the source signals ＋

Store: 72 units
Bar: 18 units
Demand: 30 units

Prepared forStock manager

Where AI fits

AI reconciles and studies stock movements between locations, identifies trends and flags stockout risks. As part of a wider operating system, it combines inventory data with weather, bookings, events and other business activity to put changes in context. It suggests actions and alerts the right managers and staff, helping operators make faster, better-informed decisions.

Business case, operating model & development status

The business case

A stock balance is only useful when the team can explain it. The Inventory MVP is designed to link stock movements to their source and make differences visible before they become a month-end investigation.

Operating model

ERPNext is the single authority for quantity and valuation. The tailored operator experience covers receiving evidence, transfers, counts and correction workflows. POS consumption is a defined integration path; retries must not duplicate a stock movement.

Where it stands

The MVP scope, architecture and delivery slices are defined. Proofs of concept establish selected mechanisms; they do not establish a complete or live inventory product. Pilot use depends on confirmed business rules and operational acceptance.

Explore the operational foundation

Explore the workflow

Store ledgerMovement recorded

Item

House supplies

One item · Base units

Expected64

Counted—

Difference—

Opening balance40

Received+24

Transferred out—

Delivery docket linked

Start with evidence.

An illustrative delivery adds 24 units to an opening balance of 40. The intended receiving workflow retains the delivery docket.

Store ledgerMovement recorded

Item

House supplies

One item · Base units

Expected52

Counted—

Difference—

Opening balance40

Received+24

Transferred out−12

Source and destination retained

Follow the movement.

A transfer of 12 units leaves the source location with 52. The destination and movement reference stay part of the record.

Store ledgerReview needed

Item

House supplies

One item · Base units

Expected52

Counted50

Difference−2

Opening balance40

Received+24

Transferred out−12

Count difference flagged

Make the difference visible.

The physical count is 50 against an expected 52. The difference is a review item, not an invented explanation.

Store ledgerReview needed

Item

House supplies

One item · Base units

Expected52

Counted50

Difference−2

Opening balance40

Received+24

Transferred out−12

Correction awaits authorised review

A person owns the correction.

The operator reviews the source movements and count. An authorised correction must preserve the history.

Step 1 of 4 · Receive

AI engineering & assuranceInternal engineering programme

Harness

Fast ideas. Evidence that keeps up.

Let AI help build and inspect the work. Keep the contract, changed code and verification evidence connected to the exact version under review.

◈Change contractScope defined

◷CandidateVersion B

◇Check evidenceVersion B

Inspect the change / follow the evidence

Acceptance contractRecovery must preserve the sale.

Candidate B

+ recover(existingOperation)

+ retain(originalSale)

+ verify(recordedOutcome)

Evidence for B

✓ Contract present✓ Checks recorded◇ Human review

◇Review ready

Code × contract × evidence

Prepare a review brief; release authority stays with a person.

Evidence briefingInsight ready

B = BCandidate and evidence match

The evidence follows the change.

The candidate, checks and source references point to version B in this example. AI prepares the review brief and highlights the remaining human decision.

Suggested next movePresent the evidence for authorised review.

See the source signals ＋

Contract: defined
Candidate: B
Checks: B · passing

Prepared forEngineering reviewer

Where AI fits

AI can propose and implement changes. Harness develops the checks and boundaries around that work, so an agent’s claim of success is never the only acceptance signal.

Business case, operating model & development status

The business case

Generating code is only one part of delivering software. Teams also need to know what was checked, which version was reviewed and who can approve the next step. Harness makes that evidence a first-class part of the work.

Operating model

Checks are tied to the exact candidate and its policy. Review and approval have distinct responsibilities; a changed candidate needs fresh evidence. Repository assurance and production agent controls are separate workstreams with their own qualification.

Where it stands

Versioned contracts, schemas and selected verifier components are implemented and tested. Broader enforcement rollout and production agent-runtime controls remain separately gated. Harness is an internal engineering programme, not a compliance certification.

Explore the operational foundation

Explore the workflow

Acceptance recordCandidate A

A clear contract for the work

ScopeDefined

ChecksWaiting

ReviewWaiting

CandidateAEvidence forPending

Acceptance proof defined before the build

Make “done” specific.

Define the outcome, permitted changes and acceptance proof before an agent starts work.

Acceptance recordCandidate A

The evidence follows the version

ScopePass

ChecksPass

ReviewReady for review

CandidateAEvidence forA

Passing checks are evidence, not deployment

Bind evidence to the work.

Checks and review must describe the actual candidate. A green result from another version is insufficient.

Acceptance recordCandidate B

Changed work. Fresh checks.

ScopeChanged

ChecksStale

ReviewHeld

CandidateBEvidence forA

Previous review cannot authorise a changed candidate

A change invalidates old evidence.

This illustrative candidate changed after review. The gate holds because its previous evidence no longer matches.

Acceptance recordCandidate B

Evidence ready for a decision

ScopePass

ChecksPass

ReviewReview required

CandidateBEvidence forB

Release remains a separate authorised action

Fresh evidence, then a decision.

Re-check the updated candidate. The authorised reviewer decides whether the work can advance; a passing check does not deploy it.

Step 1 of 4 · Define

AI workspaces & business operationsWorking local builds

Parity-OS

The whole picture. A clearer next move.

Bring the current work, specialist AI and business context into one workspace. Turn scattered information into a useful briefing, then a reviewable draft.

◈POS insightMorning trade

◷Inventory alertBar stock gap

◇Today’s work2 open decisions

The workspace / context comes together

Trading insightSource connected

Stock exceptionSource connected

Open decisionsSource connected

parity OS

✦Morning operating briefDraft

Relevant context, connected

Sources attached to the work

Ready for your review

Sales × stock × work

Assemble a source-linked daily brief for the operations lead.

Daily operating briefInsight ready

3Signals in one daily brief

Start with what needs your attention.

This concept combines a trading insight, a stock exception and two open decisions. Each item retains its source so the operator can inspect the detail.

Suggested next moveReview stock allocation, service preparation and open decisions.

See the source signals ＋

POS trading insight
Inventory exception
Workspace decisions

Prepared forOperations lead

Where AI fits

Specialist bots work with bounded business roles and source context. The design pairs concise answers with visible evidence, uncertainty and a clear next action.

Business case, operating model & development status

The business case

The most useful AI starts with the right context. Parity-OS brings the current screen, project and source evidence closer to the assistant, so a question can lead to a useful, reviewable piece of work.

Operating model

A desktop host brings together chat, bots and dashboards. Shared design components and domain-specific packs adapt it to different businesses. Local examples include operational insights and a workflow from scope to estimate and a reviewed document. Parity’s workspace builds on the open-source Hermes foundation.

Where it stands

Working local builds demonstrate dashboards, a contextual copilot and selected business workflows. Capabilities depend on the adopted pack and configured sources. A portable pack installer, automatic updates and rollback are still being developed.

Explore the operational foundation

Explore the workflow

Parity-OSWorkspace

Operations

A clearer working day.

3Active jobs

2For review

Your work, connected

Choose a project to see its work and evidence.

Project context available

Business context, in view.

An illustrative operations dashboard brings the day’s work and open decisions into one place.

Parity-OSWorkspace

Operations

A clearer working day.

3Active jobs

2For review

Contextual copilot

“Prepare a scope summary for this project.”

Relevant source context

The question starts from this project.

The assistant uses the active work context and relevant source material, with gaps kept visible.

Parity-OSWorkspace

Operations

A clearer working day.

3Active jobs

2For review

Draft prepared

Scope, assumptions and source references collected.

Draft · Not sent

Turn context into a useful draft.

A specialist prepares a reviewable work item. A draft remains a draft until the responsible person accepts it.

Parity-OSWorkspace

Operations

A clearer working day.

3Active jobs

2For review

Ready for human review

Check the scope and assumptions before accepting.

Approval remains with you

Keep the person in control.

The operator checks the result and its evidence. The appropriate business workflow owns what happens next.

Step 1 of 4 · See the work

AI.
With a job to do.

Useful intelligence starts with thoughtful design.

We start with the people doing the work: the decisions they make, the information they need and the moments where something can go wrong.

Then we design the experience, connect the right systems and build the controls around it. AI helps us explore, create and deliver. Evidence helps us decide what is ready.

Explore our approach to solutions

Bring us the operational challenge.

Let’s make the next
working system yours.

Talk about your project

A business fullof signals.Connected by AI.

Change the situation.Watch the thinking.

Every sale tells a story.

The morning rush has a pattern.

The business case

Operating model

Where it stands

Morning service

Morning service

Morning service

Morning service

See the shortage before the shelf does.

The stock exists. It is in the wrong place.

The business case

Operating model

Where it stands

House supplies

House supplies

House supplies

House supplies

Fast ideas. Evidence that keeps up.

The evidence follows the change.

The business case

Operating model

Where it stands

A clear contract for the work

The evidence follows the version

Changed work. Fresh checks.

Evidence ready for a decision

The whole picture. A clearer next move.

Start with what needs your attention.

The business case

Operating model

Where it stands

A clearer working day.

A clearer working day.

A clearer working day.

A clearer working day.

Useful intelligence starts with thoughtful design.

Let’s make the nextworking system yours.

A business full
of signals.
Connected by AI.

Change the situation.
Watch the thinking.

Let’s make the next
working system yours.