Skip to content
ETHORITY®
Search ETHORITY
DE

ETHORITY · AI calculators

Work out what AI will cost and return before you decide.

A board will ask which AI use comes first, what it costs over three years, what the saved time is worth and which EU AI Act duties may apply. Seven calculators help you answer these questions with your own numbers, and every assumption stays visible.

Why saved time only pays when you plan its use →
100 h → 40 h → 400 €Released → converted → monthly netScenario: the starting values are assumptions. Replace them with your own numbers.

IMPROVE

Capacity Conversion

AI saves time, but saved time only creates value when you decide how to use it. This calculator shows how many of the hours AI frees up turn into money, and in which month the investment pays back.

Pilot · scenario · modeled
How to use · four steps
  1. Enter how many work items you complete per month and how long each takes.
  2. Add how often AI gets it right first time, how fast your team adopts it and the time needed to check results.
  3. Decide what the freed-up hours will be used for.
  4. Test what happens if checking takes longer, and name who is responsible.

Your numbers stay in this browser. Only what you copy or put into an enquiry is shared.

Method and assumptions

Effective AI effort includes review and baseline rework for rejected units. Released hours = adopted units × (baseline minutes − effective AI minutes) / 60. Converted hours = positive released hours × conversion share. Net = converted hours × value per hour − monthly AI cost. The 12-month view ramps adoption while costs apply in full.

Read how the calculation works →

Name who is responsible for the next step

To turn this scenario into measured results, record the starting point, the accepted results, the real effort and spend, the sources and the owner's review. Keep rejected results and rework in the record too.

Written comparison of up to three offerings: AI Decision Brief · €195 net →See the cost and value models we use →

The button prepares an email draft. You check it and send it yourself, and you agree any order or appointment directly with ETHORITY.

ETHORITY · LONG READ

Plan AI in numbers: outcomes, full cost and a governed stack

54% of German AI users cannot say what AI costs them. Seven calculators turn saved hours, models, stacks, full cost and AI Act dates into a plan with owners and evidence.

Most AI plans stall on two questions. What does a useful result cost? And what does it run on? In Germany, 54% of firms that use AI cannot say what it costs them (Bitkom, Sep 2026). Worldwide, about nine in ten organisations report regular AI use, yet only 37% attribute any EBIT effect to it (McKinsey, Aug 2026). The gap is not enthusiasm. It is arithmetic, done on the wrong boundary.

Start from accepted outcomes, not hours

A supplier's claim of hours saved is activity, not value. ETHORITY's method separates four steps that are often merged: time released, capacity used, accepted outcome and attributable value. The Capacity Conversion calculator makes the difference visible. In the method's worked example, 1,000 accepted cases need 300 hours today and 200 hours with AI. That releases 100 hours. Only 40 of them go to a documented destination, avoided paid overtime, so the Scenario supports €1,600 of avoided cost, not the €4,000 a salary proxy would claim. Net of €1,200 programme cost, €400 a month remain.

The same calculator shows how fragile that is. Add 36 seconds of extra review per case and the monthly result falls to €240. That test comes from the customer-service Scenario of the AI Opportunity Case, where 36 seconds turn a best case negative.

Count all six cost blocks

Licences are the smallest cost driver German firms name; cloud and compute (51%) and data preparation (50%) come up more often. The Total Cost of AI Ownership calculator therefore asks for six blocks on one boundary: one-time integration, licences, usage, human review, governance and change. Its defaults are the cost boundary of the study's customer-service case: €11,000 of full monthly cost over 4,000 accepted outcomes, or €2.75 per accepted outcome. The decision rule is simple: compare that number with today's cost per outcome on the same boundary, before any pilot.

Choose the stack with the constraints first

Technology choices are business choices. The Stack Builder starts from the topic area, the data type and three constraints: EU data location, no training on your data, published price. It returns the six technology layers of the Good AI Prism, from Trust & Governance to Chips & Compute, and names the registry offerings that meet the constraints, each fact with its evidence mark and the date it was read. The Model Cost Estimator prices the same workload across 17 models: in its default the span runs from $70 to $6,980 a month, and two Google list prices double between December and January.

Govern by date

The EU AI Act sets dates, not moods. Article 50 transparency duties apply from 2 Aug 2026; high-risk duties for Annex III uses such as hiring and credit scoring apply from 2 Dec 2027 under Regulation (EU) 2026/1744. The EU AI Act Navigator turns seven answers into an indicative class, the duties of your role and the dates that bind you. It is orientation, not legal advice. The AI Journey Planner then places these dates on a DISCOVER to GROW roadmap, with an owner and an evidence gate at the end of every phase.

A Scenario is a starting point

Every result in the suite is labelled Scenario. It becomes an observed result only when your own records show accepted units, effort and cost for the same month. Each calculator ends with a governance panel: owner, review date, human oversight and the evidence needed. Copy the summary, bring it to your sponsor and decide on the evidence.

Sources and further reading