Cost transparency and showback
Replace the allocation spreadsheet with an auditable model, and show every unit what it consumes before anyone is charged.
Dynamic, AI-assisted cost allocation
Cloud, data centres, licences, people and now AI, allocated by clear rules to the services and departments that use them. Live as bills and contracts change, and traceable to the bill behind every charge.
Two exports and one finance contact. Fixed fee, quoted in the first call.
Engineering · monthly charge
$39.9K42.4% of IT cost, traced to 7 resources through 5 servicesWhere the money goes · one month
$94.0K from five cost types, through 5 services, to 5 departments.AI spend · monthly, by department
$10.0KTokens and seats, allocated through 2 services to the departments that use themThe books balance · every month
Unallocated: $0.00K Nothing lost, nothing invented.
The short answer
Model it, rather than spread it. Every cost (cloud, data centre, licences, contracts, people, AI) flows by named drivers to the applications that use it, is balanced across the shared business services that depend on them, and is charged to the departments that consume those services. Every charge traces back to the bill, the rule and the driver behind it, the total charged always equals the total spent, and any change can be simulated before it is made.
For consulting firms: a defensible cost model in weeks, AI spend in your cost practice, and a live asset you leave with the client, under your name.
For finance and IT: every IT dollar traced to the team that uses it, kept live as bills and contracts change, and every change tested before you commit.
About 90 seconds, with captions. Illustrative figures; cost estimates from practitioners’ IT cost engagements.
Each problem scored on three things: how urgent it is, how widespread, and what doing nothing costs. The heavier ones come first.
The ledger shows what was bought. Which unit consumed it, through which service, is a question the accounts cannot answer.
Gravity93/100
Each month the split is rebuilt by hand, the rules drift, and the numbers do not survive the second question.
Gravity87/100
Tokens sit in cloud bills and seats in software subscriptions. Nobody can say which unit consumes it.
Gravity80/100
Units dispute their charge because nobody can trace it back to the contract, the team or the usage behind it.
Gravity73/100
A data platform or a portal used unevenly is spread by headcount, so heavy users are subsidised and light users push back.
Gravity87/100
The decision is taken first and the cost consequences found later, because there is no model to run it through.
Gravity80/100
It is bought by the unit of compute and consumed inside services that were never costed at that grain.
Gravity80/100
Enterprise cost-management platforms can take the better part of a year to roll out, and cloud-only tools miss everything that is not cloud.
Gravity67/100
Buyer and seller both need the run-rate of the separated unit and a price for each transitional service, before the deal timetable moves on.
Gravity87/100
A manual allocation can take months, contract by contract across hundreds of applications, and it lives in a spreadsheet built for one client, so the next engagement starts from nothing.
Gravity87/100
A saving resting on a split nobody can trace is easy for the client’s finance team to reopen.
Gravity80/100
No live model stays with the client, so the next conversation has to be won again.
Gravity73/100
Gravity is our read from conversations with finance and IT leaders and their advisers, on a scale of 1 to 5 per axis, combined into one gravity score out of 100. A judgement to argue with, not survey data.
The ledger is organised by what was bought. The questions a board asks are about who uses it, and what it returns.
of IT-finance teams name time-consuming spreadsheets as their top frustration.
Vendor survey, vendor-commissionedis a typical roll-out for an enterprise IT cost-management platform.
Industry estimateof leaders have only partial visibility into AI spending.
KPMG Global AI Pulse, Q2 2026more likely to report established AI return when cost visibility is strong: 15% against 3%.
KPMG Global AI Pulse, Q2 2026Boards that never asked how the network budget was spent now ask, line by line, what the AI budget buys, and who uses it.
AI cost arrives as tokens, seats and GPU hours, inside services that were never costed at that grain. Answering for it needs the same thing every other IT cost already needed: a model that traces each cost to who consumes it. At the same time, the consolidation of the large cost-management vendors has many buyers reviewing their tools. Both make this the year to replace the spreadsheet.
Structured exports and unstructured documents go in. AI reads, classifies and maps them into cost facts your team approves; the engine allocates, balances and charges; the model keeps pulling new data as it arrives.
1 · Sources
2 · AI ingestion
Vendor invoices (PDF)Illustrative
Invoice INV-2026-1043 · CRM licences · 1,200 seats × 6.50 a month
Licences → CRM → Customer portal and Corporate IT
Confidence: High
3 · Engine
Every dollar in is a dollar out
4 · Outputs
Re-runs as new data lands
Which sources connect first is agreed per engagement: a first model needs only the ledger and the invoice list. Payroll data is pseudonymised at role level.
Activity-based costing, run as a model rather than a spreadsheet: costs flow from what was bought, through what uses it, to who pays.
Each cost (a cloud bill, a contract, a licence, a team) flows to the applications and infrastructure that use it, by measured usage where it exists and by an agreed driver where it does not.
Shared costs are redistributed across the business services that depend on them, so a platform used by five services is not carried by one.
Services are charged to the departments, entities or customers that consume them. The total charged equals the total spent: nothing lost, nothing invented.
The model keeps who uses a service (a team, a named group, “about 3,000 users in one region”) apart from the entity that pays for it, so a shared platform is charged to the divisions that really consume it.
Not a spreadsheet rebuilt every month. When a bill lands, a contract renews or a workload moves, the model re-allocates and every charge updates. Play the month or drag the timeline to watch it happen.
All IT cost · Day 1 · $94.0K a month Illustrative
ComputeStorageNetworkLicencesAI services
The same model, two more ways in: open a department to see what it pays for, or read any cell to see how a resource reaches a department.
Drill-down · department to resource Illustrative
Resource × department Illustrative
| Resource | Engineering | Sales and marketing | Operations | Finance and admin | HR | Total |
|---|---|---|---|---|---|---|
| Cloud compute | $32.0K | |||||
| Data centre colocation | $18.0K | |||||
| Data centre storage | $12.0K | |||||
| Cloud storage | $8.00K | |||||
| Network and egress | $6.00K | |||||
| SaaS licences | $8.00K | |||||
| AI tokens and seats | $10.0K | |||||
| Charged | $39.9K | $27.1K | $18.4K | $5.81K | $2.90K | $94.0K |
Point at or tap a cell for its share of the resource, its share of the department’s charge, and the path it takes.
Branch the model, test a reorganisation, a new compute provider, a different AI model or a migration, and compare each department’s charge against the baseline before you commit.
BaselineScenario
The cost flow above follows your scenario. Illustrative demo model; the assumptions are shown under each choice.
The first generation of cost tools spread last month’s ledger by fixed percentages. This one runs on live operational data, re-allocates as usage moves, and uses AI to build the model and explain every result.
Cost sits on the live graph of resources, deployments, services and consumers, refreshed from the source systems, so allocations move with usage.
It proposes how accounts and vendors map to cost categories, suggests drivers from the usage data, and learns the mappings so the next run starts further ahead. Your team approves every rule.
“Why was this department charged this much?” answered in plain language, with the trail from the charge back to the bill, the rule and the driver.
Budgets, forecasts and actuals in one model: variance explained as it appears, and next quarter’s cost projected from the drivers.
AI reads documents, suggests rules and answers questions. The allocation maths itself is deterministic and traceable to source facts, so the numbers you put in front of a board come from the engine, not from a chat.
The same cost-to-service model can carry AI and token spend, CO₂ and water, and data-sovereignty exposure, each traced to the service that drives it.
Each figure carries how sure the data behind it is, every run is versioned and reproducible, and unallocated cost is shown rather than spread.
Ask the model Illustrative
Engineering is charged $39.9K, 42% of the total. The largest parts are the developer platform ($12.8K), the data platform ($11.6K) and the corporate it ($10.8K). The developer platform is dedicated to engineering, so it carries all of it; the shared services are charged by measured usage.
Trail: Developer platform, Data platform, Corporate IT → the applications behind them → the resources that feed them.
Answers are written from the demo model and follow the scenario you set in the simulator, so every number in them traces back to the flow above.
Built for IT cost first: cloud, compute, storage, network, licences, software, people and facilities. The engine itself is domain-agnostic, so the same model extends to other shared costs.
Replace the allocation spreadsheet with an auditable model, and show every unit what it consumes before anyone is charged.
Charge shared platforms out by real usage, and move from showback to chargeback when the organisation is ready.
Put two IT cost structures in one model, expose the overlap and the true run-rate of the combined estate.
Model the standalone cost of a separated unit, the stranded cost left behind, and a defensible price for each transitional service.
Map the same costs to a new organisation without rebuilding the model, and compare the options before the announcement.
AI as its own cost category, landed on the services and units that use it. Paired with our engineering benchmark, it prices cost per unit of output. The process-level view of AI against labour is built with our first pilot clients.
Where each route stands, as we see it. Legacy platforms are broader in some areas; this approach wins on speed, traceability and where it runs.
| Spreadsheets | Enterprise cost-management platforms | This approach | |
|---|---|---|---|
| How cost is allocated | By hand, in one pass | Top-down, simplified | Three phases with flexible methods per cost |
| Where the data comes from | Exports pasted in each month | Periodic batch loads | Your ledger, bills, contracts and usage, refreshed on a schedule |
| Traceability | None beyond the formulas | Limited | Every charge back to the bill, rule and driver behind it |
| What-if | A copy of the workbook | Limited, backward-looking | Branch the model, run the scenario, compare to the baseline |
| Plan against actual | Separate sheets | Often an add-on | One model, variance tracked |
| When a bill, contract or workload changes | Wait for next month’s rebuild | Wait for the next batch load | Re-allocated as the data lands |
| Unstructured data (invoices, contracts) | Retyped by hand | Mostly manual mapping | Read and mapped by AI, approved by your team |
| Time to a first model | Days, then months of upkeep | Often six to eighteen months | A first cut from a raw export in days, refined over weeks |
| Where it runs | On a laptop | The vendor’s cloud | In your own infrastructure or cloud |
The model stays with you, re-runs on a schedule, and runs in your own infrastructure or cloud.
What each department and entity consumes, by service and by cost category, with the trail from each charge back to the bill.
Budgets, forecasts and actuals side by side, with the variance explained by what moved underneath.
Branch the model, change a driver, a contract or the organisation, and compare against the baseline.
AI spend found in subscriptions and cloud bills, allocated to the units that use it, and tracked as it grows.
What each business service costs to run, built up from the resources, contracts and teams behind it.
Allocated by real usage rather than headcount, with the method chosen per service and visible to every unit.
Renewals, cloud moves, team changes and reorganisations run through the model first.
Exports to Power BI, Tableau, CSV or an API for finance and planning, and the model runs in your own infrastructure.
Stand up a first defensible model from two exports, early enough to shape the engagement rather than close it.
Every figure traces to a source, a rule and a driver, so your recommendation stands on a model, not an estimate.
Carve-outs, transitional service pricing, integrations and reorganisations run through the same model.
The model stays with the client and re-runs every month: the quarterly review becomes a standing conversation. We run it behind you.
Begin coarse for a fast result, then break out the biggest costs for more clarity. Versions and branches keep every refinement safe and reversible, and you can stop after any step.
Two exports and one finance contact: the IT ledger and the vendor and invoice list. Cost by category, with AI as its own line, mapped to the first services. A result in weeks, on its own.
Start with a first cost modelTwo exports and one finance contact. Fixed fee, quoted in the first call.
Service owners agree the drivers, one workshop each. Every department sees what it consumes, with the trail behind each figure.
Run the decisions in front of you through the model, and move to chargeback when the organisation is ready.
Cost per unit of work for the processes that matter, with labour and AI side by side. Built with our first pilot clients.
Each answer stands on its own.
Tracing IT cost (cloud, data centres, licences, contracts, people and AI) to the business services and departments that consume it, by named drivers such as measured usage or headcount. It runs in three phases: costs are allocated to the applications that use them, shared costs are balanced across business services, and services are charged to the departments that consume them. The result supports showback, chargeback, budgeting and what-if decisions, and every charge traces back to the bill behind it.
A spreadsheet allocates by hand, in one pass, with no trail beyond its formulas, and breaks as it grows. Enterprise platforms are broad but often take six to eighteen months to roll out and run in the vendor’s cloud. This approach runs a three-phase model with flexible methods per cost, traces every charge to its source, simulates changes before they are made, reaches a first model in weeks and runs in your own infrastructure.
For a first model, the general-ledger or trial-balance export for the IT cost centres (12 to 24 months) and the vendor and invoice list, with one finance contact. Showback then adds the service catalogue or application list and the drivers you can evidence: usage, headcount, ticket or transaction volumes, licence assignments. Payroll data, where it is used, is pseudonymised at role level.
Both. The allocation engine is a shipped product from our technology partner; we run the engagement, the data loading, the allocation rules with your teams and the read-out. Cost allocation, showback, scenarios, plan against actual and the audit trail run today. The process-level view of AI against labour is being built with our first pilot clients.
Every figure traces to a source fact, a named rule and a driver quantity, so your finance team can check it rather than accept it. Unallocated cost is shown rather than spread, runs are versioned and reproducible, and the total charged always equals the total spent. The method for each cost is chosen with you and visible to every department.
The engine runs in your own infrastructure or cloud. Payroll data, where it is needed, is pseudonymised at role level: the model needs roles, locations and cost, never names.
The first cost model is a fixed fee, quoted in the first call. Later steps are scoped as defined work packages: fixed where the scope and data are known, capped where they are new. If you keep the model live after the engagement, the engine is a separate subscription, and neither depends on the other.
Yes. The firm leads the client and the recommendation; we run the engine, the data work and the read-out behind it. It suits cost transparency, shared-service costing, carve-outs and transitional service pricing, integrations and restructuring, and the live model creates the follow-on conversations.
No. The model shows who uses what and what it costs, in numbers your teams can trace, and lets you test options before you commit. Decisions on what to change, and the savings that follow, stay with you and your advisers.
Start with two exports. You get a first, auditable cost model within weeks, and you decide what comes next.
Two exports and one finance contact. Fixed fee, quoted in the first call.
Who are you?
Tell us and we’ll tailor the site to you.
Who are you?
Want to join our delivery network? Send us your firm and case studies.
Not right?