Skip to content
ARITHIM ESTATE · FOR CIOS, CISOS, ARCHITECTS AND PMOS

One record of what you run, what it costs and what is changing it.

Find the shadow AI, AI agents and shadow IT nobody registered, then keep one record of what you run, what it costs and what is changing it. It fills itself from your identity, finance and contract systems, so your first report arrives before anyone has typed a thing.

Opening first to a small group of design partners. The SaaS management core comes first and the architecture and portfolio tiers follow. Every feature below carries its own status, and we have not dated any of them.

A scan line sweeps across a dotted skyline at dusk and towers appear, each window a dot: bright where an application is in use and faint where it is hardly used. Light packets travel along the road at the foot. One window in the tallest tower flickers in amber, and a small drone in the sky is ringed in amber, both standing for something nobody registered. A sample, not a real estate.
Illustrative. Not your data.
WHERE TO START

Start with the question that worries you most

Most customers begin with one of these. All of them run on the same records, so the second one is already half answered when you reach it.

CISOs and heads of cybersecurity

Shadow AI and AI agents

Which AI tools and agents are in use, who owns them and what data can they reach? Find the ones nobody registered, tier them by risk and give your board a current answer.

See shadow AI discovery →

CIOs and heads of IT operations

Shadow IT and SaaS sprawl

See every application your people use and pay for, including the ones bought on a card, and decide which to keep, merge or retire.

See the first report →

Procurement and finance

Renewals and wasted licences

Never miss a notice date again, and get unused seats back into a pool before the renewal, with finance signing each saving once.

See the renewal clock →

Enterprise architects

A portfolio that keeps itself current

A capability map and a position on every application, filled from live data so it stops going stale, with reviews sized to the risk.

See the capability map →

PMO leads

One view of change

Requests, renewals and projects on one record, with status drawn from the delivery tool and benefits measured from real spend.

See request to delivery →

Security and audit teams

Evidence on request

Access reviews, leaver checks and a log of who approved what, ready for an auditor without a scramble. See a [regulated approval workflow we built](case-compliance-vendor-approval.html).

See the smaller features →

SEE IT WORK

What does Arithim Estate do in twelve seconds?

Feeds fill the inventory, every application gets a business capability and a position, a chat request is answered with something you already own, a renewal clock runs down and a saving is signed once. Names and figures are made up to show the idea.

Arithim Estate in twelve seconds
WHERE EACH PART STANDS

How to read the status on each feature

  • Early access
    Opens first, to design partners. These parts make up the SaaS management core.
  • Soon
    Part of the first full release. The architecture and portfolio tiers sit here.
  • Planned
    Designed and agreed for the release after that.
  • Later
    On the roadmap with no release set.
WHY TEAMS CHOOSE IT

Why teams choose it

Shadow AI and shadow IT, found in days

Connect read-only to the systems you already run. Within about two weeks you see the AI tools, agents and applications nobody registered, with spend, unused seats and renewals alongside.

Records that keep themselves current

The portfolio fills from identity, finance, card and contract feeds. People are asked only for judgement: who owns this, how critical it is and what should happen next.

One answer for the CIO, the architect, the PMO and finance

Each person opens on the screen that answers their question, from the same records. The capability map, the renewal board, the project funnel and the savings ledger never disagree. The idea behind it is operational intelligence.

Decisions that reach the work

A board decision becomes a record that gates, renewals and project plans read. A retirement counts as a saving only when the spend has actually stopped.

HOW IT WORKS

One loop: discover, decide, change, prove

01

Discover

Find what the company runs and pays for, including applications and AI agents nobody registered.

02

Decide

Route every request by size and risk, set an architecture position and make the renewal decision with the facts in front of you.

03

Change

Carry the decision through a request, a renewal or a project, with owners, gates and a roadmap.

04

Prove

Show that the money moved. Finance signs each saving once, and every figure links back to its source.

THE LARGE FEATURES

The large features, and how each one works

Ten parts carry most of the product. Each animation shows one in a few seconds, and the points beside it say what it does in practice.

Early access

Shadow AI: every AI tool and agent, found and owned

Discovery, cost and owners come first. Risk tiers and re-attestation arrive with the architecture tier.

Find the AI tools, assistants and agents your people use that nobody registered, and the ones your suppliers' products bring in. Each gets an owner, a business purpose, a cost and a risk tier, in the same portfolio as everything else.

  • Found through sign-ins, OAuth grants, vendor usage reports and spend, including personal accounts used on work data
  • Each agent is tiered by autonomy, data reach and external effect
  • Ownerless and dormant AI is flagged, and people are offered an approved alternative
  • Metadata only. Prompts and outputs are never read
  • An inventory to start AI governance and EU AI Act work from

Need proof of which AI models you depend on? See the AI model inventory.

An unregistered agent found, tiered and given an ownerUnregistered AI agentAgentsupport botCustomer dataDocumentsTicketsTier 2Owner neededAsked of their managerOwner confirmedRe-attest in 90 days
Early access

Your first report in ten working days

Connect read-only to your identity provider, finance system, card feed and contract folder. Within ten working days you have a reconciled inventory of what you run and pay for, before anyone has typed a record.

  • Spend by application, vendor and cost centre
  • Unused seats, and licences paid for but not used
  • A shadow IT and shadow AI list, ranked by risk and spend
  • Renewals by notice date, suggested owners and a draft capability map
Feeds filling an application inventorySOURCESIdentityFinanceCardsContractsINVENTORYDesign suite612kChat340kCRM905kNotes app18kSurvey tool44kShadow ITANNUAL SPEND4.2MUNUSED SEATS1,420
Early access

Every number shows where it came from

Each value carries its source, its age and how sure the product is. When two systems disagree, both stay on screen and the difference becomes an exception for a person to resolve.

  • Confidence is Confirmed, Likely or Inferred, and says why
  • Out-of-date values are greyed out, with the reason
  • Conflicting figures sit side by side and are never picked silently
  • A person's judgement is recorded the same way, under their name
A figure with its source, age and a conflictSEATS · DESIGN SUITE520paid seatsVendor API4 min agoLikely400assigned in identityIdentity2 min agoConfirmed≠Disagreement kept, never picked silentlyRaised as an exception for a person to reconcile.ReconcileOut-of-date values are greyed out
Early access

One front door in chat, portal and email

People ask for a tool in Slack, Teams, a portal or by email. The product checks whether you already own something that does the job, offers a spare seat first and routes everything else by size and risk.

  • At most five questions, mapped to a business capability
  • A pooled seat or a sanctioned alternative is offered before any purchase
  • One thread and one status for the requester
  • Three requests for the same unsanctioned app in 30 days raise a demand
A chat request answered with something you already ownCHATI need a tool todraw diagrams.You already own Designsuite. 14 seats are free.Want one?Yes, use itNo, I need moreCapabilityDiagrammingROUTEAccessPurchaseReviewDemandSized by value and risk
Early access

Renewals start 120 days before the notice deadline

One clock runs per contract, anchored on the date by which you must give notice. Decisions get made in time, with the architecture position and the usage data in front of the person deciding.

  • Day 120: reclamation batch runs, and the architecture position is posted
  • Day 90: a brief with renew, right-size, renegotiate, replace or cancel
  • Day 60: decision due. Day 30: escalation, and a drafted notice for a person to send
  • Unconfirmed contract dates stay red, and the product never sends notice to a vendor
A renewal clock counting 120, 90, 60 and 30 daysDesign suiteNotice by 14 March120 daysReclaim unused seatsPosition posted90 daysRenewal briefRenew · right-size · replace60 daysDecision dueBudget owner decides30 daysEscalateNotice drafted for a personNoticeNever sent for you
Early access

Licence reclamation with a safety net

Unused seats go back into a pool, and the person decides first. Nothing is removed in the first 30 days of a new rule, and every change can be undone.

  • Candidates come with a reason and the annual cost
  • Leave, new joiners, admins, service accounts and named executives are exempt
  • The user sees Keep or Release, then their manager is asked after 5 working days
  • Batches are capped at 25 changes or 5% of a contract, with undo for 14 days
Seats released into a pool, with undoSEATSKeep or release?Not used in 90 days · 640 a yearReleaseKeepPOOLUndo for 14 daysReview only for 30 days
Soon

A capability map with a TIME position on every application

Start from a map of a starter map of business capabilities. The product proposes which capability each application serves and whether to Tolerate, Invest, Migrate or Eliminate it, showing the evidence. Architects confirm in an afternoon, and nothing waits on the map.

  • Heatmap lenses for spend, TIME mix, risk, overlap, investment and coverage
  • "Unmapped" is allowed, so the map is useful from day one
  • TIME is proposed from the evidence, and people confirm it
  • A confirmed position never changes silently
A capability heatmap switching lensLens: SpendLens: TIMESalesMarketingFinancePeopleDeliverySupportProductLegalDataSecurityPlanningOperationsEvery application is placed on a capability.Design suiteProduct designMigrateOverlap: 3 appsActive users: 61Cost per user: highInferredTolerateInvestMigrateEliminate
Soon

Architecture reviews sized to the risk

Small, low-risk requests are approved on a clock. Larger ones go to one architect. The few that carry real risk go to a design authority with a pack already assembled.

  • Tier 1 auto-approves unless the architect stops it
  • Tier 2 goes to one named architect, Tier 3 to the design authority
  • The pack shows overlap, standards, vendor risk, data classes and interfaces
  • One decision, with conditions tracked as tasks and exceptions that expire
Requests routed to three review tiersSmall toolTier 1Team platformTier 2Customer data AITier 3ClockApprovedOne architectApproved with conditionsDesign authorityOn the agendaOne decision, written to the gate and the risk log
Soon

From request to delivery on one record

A demand becomes a business case, then a funded initiative, then a set of gates, with status drawn from the delivery tool your teams already use. Nothing is copied from one record to the next.

  • Fast track, one-screen case or full case, depending on value and risk
  • Scored for strategic fit, return, cost removed, risk reduction and confidence
  • Status is derived from schedule, cost, scope and adoption, with the project manager's override beside it
  • Every initiative declares which applications it introduces, changes, replaces or retires
A demand moving through case, gates and statusDemandCaseGateBuildLiveCardCardCardCardCardSTATUS, DERIVED FROM THE DELIVERY TOOLScheduleCostScopeAdoptionProject manager override: scope amber, 2 changes addedShown beside the derived value, never in place of it
Early access

A savings ledger that finance signs once

The ledger is in the core. Benefit lines on projects arrive with the portfolio tier.

Every saving has one cause and one class, and finance signs it once. The same money can never be claimed by a reclamation, a renegotiation and a project at the same time.

  • Realised and avoided savings are totalled on separate lines
  • Potential savings are shown and never added in
  • A retirement counts only when assignments are zero and spend has stopped for two billing cycles
  • Project benefits refer to savings and never copy them
One saving signed once and landing on a benefit lineSAVINGRetire notes appRetirementRealisedFinancesigns onceBENEFITLine 3refers tothe savingLEDGERRealised412kAvoided96kPotential182knever added in
MODULES

One core, with two tiers on the same data

The SaaS management core is always included. The architecture and portfolio tiers read the same records, so adding one never means another import. Choosing between tools? Read how to choose an enterprise architecture tool.

Early access

SaaS management core

For the CIO, head of IT operations, procurement and finance. Discovery, spend and contract matching, the front door, fulfilment and seat pool, renewals, licence reclamation, offboarding, access reviews, the savings ledger and the IT pack. It also includes a read API, a read-only MCP server and continuous export to a bucket you own.

Soon

Architecture tier

For the head of architecture. Capability mapping, TIME assessment, technology risk and standards, tiered architecture reviews, the decision log, derived lifecycle, observed interfaces, the rationalisation workbench, the roadmap and the architecture pack.

Soon

Portfolio tier

For the PMO lead and head of IT portfolio. Demand triage, business cases, prioritisation, stage gates, derived status, RAID, benefit lines, weighted objectives, the portfolio pack, and import from spreadsheets and legacy project plans.

A company that buys only the core still gets a working front door. Requests that would go to the demand path wait in a named approver queue, and Tier 2 and 3 reviews go to a named IT or security reviewer. Adding a tier swaps those queues for the review board and the portfolio.

THE SMALLER FEATURES

How the smaller features work

Thirty-two smaller features sit around the ten above, grouped by the loop. Each says what it does and how it behaves when it runs.

Discover

Early access

Shadow app and risk triage

On the first sighting of an unapproved app, the product scores its risk, checks for overlap and asks the first user what it is for and what data goes in. A free, low-risk app with five users or fewer and no overlap is marked Allowed. Overlap triggers an offer of the approved app. Blocking is advisory, and revoking access is always a ticket. This is the same discovery that catches vibe-coded apps built outside IT.

Early access

Spend matching and contract capture

Card descriptors and invoices are cleaned and matched to vendors, invoices are split by line, and currency is converted at the transaction-date rate. Contract terms are extracted by AI and stay Inferred until a named person confirms them. The product never reads full card numbers and never makes payments.

Early access

Ownership and freshness chase

Each application has a business, technical and budget owner. The product suggests an owner with the reason, such as "212 active users in Marketing, renews 14 March", and asks Accept, Not me or Ask a question. Leavers' apps move to their manager the same day, and nobody gets more than one question or nudge per app a week.

Soon

Technology risk and standards

For custom, hosted and desktop software, the product joins versions to end-of-support dates and to your adopt, trial, contain and retire standards. The technical owner picks Plan upgrade, Accept risk until a date or Already handled. Accepted risk expires and re-triggers.

Soon

Observed interfaces

App-to-app OAuth grants, API connections and MCP links seen by the feeds are recorded as interfaces, so the impact of retiring an application shows what depends on it.

Decide

Early access

Seat pool, access and purchase

Access requests pick a tier, role and duration and follow the policy's approval chain. A pooled seat is used first, and the seat expires by ticket. Use is checked after 30 days. Purchases go in parallel to the budget owner, security, the architect and procurement, and the signed contract creates the vendor, application and renewal. Privileged roles never self-serve.

Soon

Demand triage

A daily digest proposes a route for each demand: fast track, one-screen case, full case, merge or reject with a reason. The demand manager accepts or changes it in one click. Findings and renewals never create initiatives on their own.

Soon

Business case

The product drafts options, cost lines and benefit lines, each bound to a measure such as "matched spend reaches zero". The project manager writes up to 300 words and a sponsor confirms. Above the board threshold, finance reviews and the portfolio board approves.

Soon

Prioritisation and scoring

Initiatives are scored for strategic fit, financial return, cost removed, risk reduction and delivery confidence, ranked, and a funding line is drawn. Manual moves need a visible reason, and the board approves the funded list. Estimated effort is totalled by team per quarter.

Soon

Rationalisation and retirement proposals

Candidates are ranked by saving multiplied by confidence, with a target app. The impact view shows users, interfaces, data to export and the last safe notice day. On approval it raises a demand carrying the evidence, fast-tracked if small.

Change

Soon

Stage gates and change control

Each gate pack marks every item evidenced, attested or missing. At the build gate the product checks that the review passed with no open conditions, that security and data protection are signed off and that a baseline exists. Variance opens a pre-filled change request, and impact changes reopen architecture review.

Soon

Derived status

Schedule, cost, scope, adoption, risk and overdue conditions each get a rating from the data. The project manager gets a card to confirm, edit or override with a reason. After Friday noon an unconfirmed status is published marked "not confirmed by PM". Adoption turns red only on Likely or Confirmed usage.

Soon

Roadmap, risks and conflicts

The roadmap is derived from each initiative's impacts and milestones. The product detects conflicts such as a retirement after a notice date, a capability left with no Invest app, or two initiatives on one app in a quarter, and posts each to the risk log as a suggestion with evidence.

Early access

Offboarding

For each leaver the product computes their footprint, checks identity deactivation, raises one ticket per non-SSO account, returns licences to the pool and proposes ownership transfer to the manager. It also flags agents, keys and card subscriptions under the leaver's identity. It deactivates and never deletes, and a legal hold stops destructive steps.

Early access

Access reviews

A campaign recommends keep or revoke with a reason: leaver, unused for 90 days, policy deviation or moved. Reviewers act from a chat or email card without logging in. Leaver revocations are accepted automatically, privileged access needs a second reviewer, and there is no approve-all on it. The campaign closes with an evidence pack for auditors.

Prove

Soon

Close and retirement confirmation

For each retired app the product checks zero assignments, no sign-ins for 30 days, unassigned licences, evidence of non-renewal and two billing cycles without spend. Nobody sets Retired by hand. The owner attests data archiving and retention, and benefits are handed over with 3, 6 and 12-month reviews.

Early access

Packs for boards and reviews

The IT pack covers spend, savings, renewals, shadow AI and coverage. The architecture pack and the portfolio pack follow with their tiers. Each opens with the decisions needed, links every figure to evidence and carries an AI-drafted narrative labelled a draft until the author releases it. They share as a page without a login, with PDF and slide exports.

Early access

Read API, MCP server and continuous export

A read API and a read-only MCP server answer questions over your data, and any proposal they make needs a person to approve it. The full dataset exports continuously to a bucket you own, with a documented schema.

Soon

Spreadsheet and legacy project import

A one-off upload with column mapping brings in project registers, project plans and spreadsheets, so the portfolio starts with history. Imported rows are marked as imported.

Soon

Browser, device and AI-vendor feeds

After the first value report you can add a browser extension, device management and AI vendor usage reports. The product shows a scope manifest for each, and an aggregate-only mode and a works council pack are available for Europe.

Next and later

Planned

Team-level capacity on the prioritisation board

Capacity by team and quarter, with no named people, so the funding line reflects what teams can deliver.

Planned

Cloud account and code-host feeds

Technology components for custom software are read from cloud accounts and code hosts, so the standards and obsolescence views stay current without manual entry.

Planned

More agent platforms and connectors

Discovery for further agent and automation platforms, more procurement connectors, a planning-tool connector for status and a ticketing connector for incidents.

Planned

Access fulfilment by group add

Where pilots ask for it and security teams accept it, access can be granted by adding a person to an allow-listed group instead of by ticket.

Later

Birthright access, joiners and movers

Access that follows a person's role when they join or move, starting with the ten applications that hold the most seats.

Later

App-level provisioning and revocation

Provisioning and deprovisioning inside applications, and approved revocation of OAuth grants and API keys.

Later

More status and planning connectors

Status from further work-management tools, and read and write-back with financial planning tools.

Later

Price benchmarks and custom dashboards

Opt-in, anonymised price benchmarks across customers, and a dashboard builder for views we have not thought of.

Planned

Sovereign deployment in your own estate

Run Arithim Estate inside your own cloud or data centre as an independent application. AI features use endpoints you control: your own GPU clusters, or models through your AWS Bedrock or Microsoft Foundry account, so the AI calls stay within the boundary you set.

Later

Your own workflows

Build your own workflows on the same records, with the same guardrails: review-only first, undo, and an exportable log. The ready-made workflows we ship stay as starting points.

Later

Your own data model

Add your own object types, fields and relationships, so the product can hold what is specific to your organisation. The standard model stays as the default.

Later

Diagrams from your records and from code

Architecture, process and value stream diagrams generated from your records and from code, the way coding assistants draw them today, and kept in step with the data. It works with the diagramming, scheduling and task tools you already use through their APIs, instead of replacing them.

CONNECTIONS

What it connects to, and what it can do there

Most connections are read-only. The product writes messages in chat and email, tickets in one queue you name and, if you switch it on, epics in one delivery project. It never writes to finance, ERP or HR systems.

Early access

Identity provider

Microsoft Entra ID, Okta, Google Workspace, JumpCloud and OneLogin. Read-only at first. Removal from allow-listed licence groups is opt-in under the second consent.

Early access

Microsoft 365 and Google Workspace

Licences, usage reports, OAuth grants and Copilot or Gemini usage, through the same read-only consent.

Early access

HR system

Workday, SuccessFactors, BambooHR, HiBob, Personio, UKG and Rippling. Named fields only. No salary, address or protected characteristics. Required before reclamation goes live.

Early access

Finance and ERP

NetSuite, Oracle Fusion, Dynamics 365, Sage Intacct and Xero, by read-only role or saved report, with file export accepted and 24 months of backfill.

Early access

Cards and expenses

Amex, Ramp, Brex, Navan, Expensify, Pleo, Spendesk and Payhawk, filtered to software merchants, with a CSV fallback. Card numbers are dropped.

Early access

Contracts

Bulk upload from day one, then a connector to one named SharePoint or Drive folder, or to Box, DocuSign or a contract platform.

Early access

Chat and email

Slack and Microsoft Teams for the front door and every card, with no channel history read. A forwarding address takes request and invoice mailboxes without mailbox access.

Early access

Ticketing

Jira Service Management, Zendesk and other service desks, through an integration user limited to one queue.

Soon

Delivery tools

Jira Cloud and Azure DevOps. Read work items and, if you switch it on, create epics in one project. Fields only.

Early access

Vendor admin APIs

Read-only admin tokens for Microsoft 365, Slack, Zoom, Atlassian, Salesforce, Adobe, GitHub, Miro, Figma and DocuSign. Your ten biggest by spend first.

Soon

Devices, browsers and network security

Intune and Jamf, a browser extension for Chrome, Edge and Firefox, and summaries from Netskope, Zscaler, Defender for Cloud Apps and Cloudflare. All optional.

Soon

AI vendor admin APIs

OpenAI, Anthropic, GitHub Copilot, Cursor and Copilot Studio usage, cost, seat and agent listings. Never prompts or outputs.

Early access

Your own bucket, and your sign-in

Write-only export to your S3, Azure Blob or Google Cloud Storage bucket, and SAML or OIDC sign-in with SCIM from your identity provider.

These are the connectors we are building first, in the order design partners ask for them. If a system you rely on is missing, tell us. A file export is the fallback for most finance, card and HR sources. If your ERP is SAP, we also build read-only reporting on SAP ERP.

THE FIRST 90 DAYS

What happens from day 1 to day 90

A company that stops at day 10 has a working SaaS management product, and the architecture and portfolio screens open already filled with its data.

Day 1

Connect

You sign the first consent. Identity, Microsoft 365 or Google, a finance export, the card feed, chat, a ticket queue and our sign-in are connected, and contracts are uploaded. You see a connector board with freshness for each source and an inventory filling within hours.

Day 5

Reconcile

Our onboarding lead works the merge queue with you and connects your ten biggest vendor accounts. You see the first reconciled inventory with confidence on each field and a list of exceptions.

Day 10

First value report

Spend by application and cost centre, unused seats, shadow IT and shadow AI, renewals by notice date, suggested owners and a draft capability map. The project funnel already shows renewal demands from the clock.

Day 15

Second consent

Offered on the strength of the report: browser extension, device management and AI vendor reports, each with a scope manifest, and a works council pack where one is needed.

Day 30

Front door live

The lead architect confirms the starter map for the top 50 applications in an afternoon. The front door goes live with the capability check and the seat pool, and reclamation rules run in review-only mode. Owner coverage passes 60%.

Day 45

Reviews and demand

Tiered architecture reviews and demand triage go live, and you see the first derived project status with the project manager's override beside it.

Day 60

TIME and first release of seats

TIME proposals appear with their evidence, the first access review runs on audit-scope apps and reclamation goes live where the HR system is connected.

Day 90

First signed saving

The first renewal is decided with an architecture position, finance signs the first saving and the first portfolio pack goes out with every figure linked to its source.

AI IN THE PRODUCT

Where AI helps, and who decides

AI drafts, suggests and summarises inside each flow. It acts alone only on a reversible change to the product's own records, above a confidence threshold. A named person decides everything else, and every AI suggestion is labelled as one.

Matching names and spend

Decides only when it is very sure, and suggests otherwise. Every merge can be split.

Reading contracts

Drafts the dates, notice period, value and terms. A named person confirms them before any clock uses them.

Mapping capabilities and TIME

Suggests links and positions, marked Inferred. Owners and the architect confirm the ones that matter.

Review packs, business cases and status

Drafts options, cases and up to 120 words of weekly status. The architect, the project manager or the sponsor decides.

Access review recommendations

Suggests keep or revoke with a reason, and accepts leaver revocations. A reviewer decides the rest.

The MCP server and packs

Answers questions over your data and drafts the narrative of a pack. A person approves any proposal and releases any pack.

AI never sends anything to a vendor, never approves a write to one of your systems, never reads prompt content and never changes a Confirmed value. If your people need help owning these decisions, see our AI readiness assessment and AI training.

SECURITY AND CONTROL

Security, data and control

We give your security team a scope manifest for each connector on the first call: every scope, the fields it yields, why we need it, how long we keep it and which feature uses it. For regulated teams we also build compliance workflows with an audit trail.

Two consents, not one

The first covers read-only access to your systems of record, plus messages, tickets and, if you choose, epics. The second follows your first report and covers the browser extension, device management, AI vendor reports and opt-in group removal after 30 days in review-only mode.

What we never ask for

Mailbox or calendar content, chat history, salary or protected characteristics, prompts or outputs from AI tools, full card numbers, write access to finance, ERP or HR systems, passwords, or network scanners on your estate.

Guardrails on every write

30 days review-only for a new rule, allow-listed apps and groups, deactivate and never delete, capped batches, exemptions, one-click undo for 14 days, a kill switch per rule and an exportable log of who approved what.

Your region, your retention

Tenant data stays in its region: the EU and the US at launch, with the UK and Australia on request. Raw usage events are kept for 90 days and daily roll-ups for 25 months.

Leaving is designed in

Continuous export of the full dataset to a bucket you own, a documented schema, a read API and a contractual wind-down of at least 90 days.

Security paperwork

A completed CAIQ and SIG Lite, a data processing agreement with standard contractual clauses, a sub-processor list and a penetration test summary. SOC 2 Type I is planned for launch, with Type II and ISO 27001 within twelve months.

WHO IT IS FOR

What each person sees first

CIO

A capability heatmap with lenses for spend, TIME mix, risk and overlap, a feed of what changed this week and the decisions due at the next board.

Enterprise architect

The capability map and a queue of reviews, low-confidence mappings, contested TIME proposals and expiring exceptions, each with one-click accept.

PMO lead

A funnel from request to realised benefit with value at every stage, an exceptions inbox, the prioritisation board and the pack builder.

Project manager

One scroll: derived status with evidence, milestones, risks, the gate checklist, budget against actuals, the applications in scope and the benefits.

IT and SaaS operations

A decisions inbox sorted by money at stake and deadline: renewals, shadow apps, reclaim batches, ownerless apps with spend and connector failures. See a connected reporting layer we built.

Procurement and finance

The renewal board by notice month, the matching queue, renewal briefs and negotiation packs, and the savings ledger with sign-off cards.

Budget owner

"My money": the apps, spend, renewals, initiatives and benefits for their cost centres, readable on a phone.

Anyone in Slack or Teams

No screen at all. The app answers "I need a tool for X", shows what they already have and carries every card for requests, owner confirmations and reviews.

GOOD TO KNOW

Good to know before you talk to us

  • This page describes a product being built. The status beside each feature is our current intention, not a delivery date, and it may change as design partners tell us what matters.
  • It does not scan mailboxes or calendars, inspect prompts, handle payments or virtual cards, or send anything to a vendor on your behalf.
  • It is delivered as a service. It can also be deployed inside your own estate as an independent, sovereign application, using your own AI endpoints: models on your own GPU clusters, or through your own AWS Bedrock or Microsoft Foundry account. Talk to us about what that needs.
FREQUENTLY ASKED

Questions about Arithim Estate

What is Arithim Estate?

A system of record for the technology a company runs, what it pays for, why it has it and what will change it. It combines SaaS management, enterprise architecture and portfolio management on one set of records, so the three views agree.

How do I find shadow AI in my company?

Connect read-only to your identity provider, finance and card feeds and the AI vendors' admin reports. Arithim Estate finds AI tools, assistants and agents that nobody registered, shows who uses them, what they cost and what data they can reach, and asks the owner to confirm. It records metadata only and never reads prompts or outputs.

Does it help with AI governance and the EU AI Act?

It gives you the inventory and the owner, risk tier and review records those programmes start from. It is not legal advice. If you also need dated proof of which AI models you depend on, pair it with the AI model inventory.

Is it SaaS management, enterprise architecture or PMO software?

All three, on the same data. The SaaS management core is always included. The architecture and portfolio tiers add capability mapping, reviews and decisions on one side and demand, cases, gates and benefits on the other.

How long until I see value?

The first report arrives within ten working days of connecting read-only to your identity provider, finance export, card feed and contracts. It shows spend, unused seats, shadow IT and shadow AI, and renewals by notice date.

Does it change anything in our systems?

In the first release it writes messages in Slack, Teams and email, tickets in one queue you name and, if you switch it on, epics in one delivery project. With your second consent it can also remove people from allow-listed licence groups after 30 days in review-only mode. It never writes to finance, ERP or HR systems.

Can we start with only the SaaS management core?

Yes. You still get the front door, the seat pool, the renewal clock, licence reclamation and the savings ledger. Requests that need a bigger decision wait with a named approver until you add a tier.

How does it find AI tools and agents we did not register?

From sign-ins, OAuth grants, vendor usage reports, browser signals and spend. It records metadata only and never reads prompts or outputs. Each agent gets an owner, a risk tier and a re-attestation date.

How does it work with Jira or Azure DevOps?

It reads work items and dates to derive project status and, if you switch it on, creates epics in one project you name. It never edits tasks or plans.

Where is our data kept?

In your region. The EU and the US are available at launch, and the UK and Australia on request. You can export everything continuously to a bucket you own.

What happens if we want to leave?

Your data is already in your bucket in a documented schema, and we commit to a wind-down of at least 90 days.

Which features can I use today?

None of them are generally available yet. The status on each feature shows what opens first to design partners (Early access), what follows in the first full release (Soon), and what is Planned or Later. Talk to us and we will tell you where your needs fall.

Can we run it inside our own environment?

That is planned. Arithim Estate can run inside your own estate as an independent application, with AI features using endpoints you control: your own GPU clusters, or your AWS Bedrock or Microsoft Foundry account. Today it is delivered as a service, and we can talk through what a sovereign deployment would need.

Related: AI model inventory

Tell us what your estate looks like today.

Tell us how many applications and AI tools you think you run, who keeps the architecture records and how projects reach the PMO. We will say where Arithim Estate would help first, and which parts are ready for you.

Talk to us about early access →

Prefer email? Write to hello@arithim.com.

Call us