ETHORITY®
Models, agents, data, tools, controls within its own platform
ETHORITY adds: Cross-provider fit, alternatives, evidence, risk, cost, ownership, review conditions
The result: A clearer decision you can review and revisit
Enterprise AI Transformation & Trust
Make an AI investment your organization can operate.
Different teams are buying copilots, agents and platforms. Leadership needs to know which systems meet business requirements, who controls them and how they fit the organization.
The situation
A business wants to evaluate an AI estate, select a platform or redesign a workflow while preserving security, accountability and implementation ownership.
What the work examines
We examine the business requirement, actual system use, dependencies, data permissions and ownership. The Decision Record compares alternatives and sets out costs, risks and the next decision. A follow-on engagement can cover architecture, integration, work redesign and operational controls.
What to measure
Owned decisions, explicit system dependencies, fewer unsupported commitments, defined operating controls and a measurable investment case.
The decision support
Registry and source intelligence, selected comparison, authority and evidence mapping, decision conditions and intended Workspace continuity under a separately agreed operating scope.
Explore the operating choices, assumptions and decision conditions. Worked examples are labelled where they appear.
Three questions every leadership team must answer
1. "What AI do we actually run, and what does it cost?" Copilot, Agentforce, AI features in every SaaS tool, internal agents, pilots in every business unit. Spend is spread across IT, business and cloud budgets. 2. "What does the AI Act require of us, and who signs off?" Each system needs a role (provider or deployer), a risk class and evidence. Legal, IT and the business units each hold part of the answer, and nobody holds the whole. 3. "Which big AI bets should we make, change or stop?" Vendors show best cases and internal teams defend their projects. Boards need an independent view of options, economics, risks and stop criteria.
What good looks like: six practices
1. One maintained AI register: owner, purpose, vendor, data, role and risk class for every system. 2. Clear decision rights: an AI committee, an owner per system and a simple approval path. 3. Evidence, not claims: vendor statements kept apart from measured and verified facts. 4. Economics per item: cost per active user, overlap and value, so the portfolio can be steered. 5. Literacy at scale: role-based training, with evidence across business units. 6. Continuous change watch: vendors, standards and guidance tracked every quarter.
None of these requires new software. All of them require clarity and ownership.
Capability is not a decision
AI providers give you capabilities. ETHORITY helps you decide which ones belong in your business.
CUSTOMER SERVICE · WORKED EXAMPLE
What changes when AI handles the routine?
Customers can ask an AI assistant directly, with a human supervisor and a clear handoff. The opportunity is to redesign the service around that behaviour.
Fictional planning example. The 30% values useful staff time at €45/hour, deducts AI and setup costs, then compares the result with today’s €600,000 annual staff effort. It is potential capacity value, not ROI or a measured customer result.
Today: a person handles every request
100,000 requests × 8 minutes × €45/hour = €600,000 in annual staff-effort equivalent.
1 Customers choose the channel
Offer direct AI chat or voice for routine questions. Keep a clear route to a person.
2 AI answers within its remit
Use approved knowledge. Uncertain answers, complaints and sensitive actions go to the service team.
3 People supervise and improve
Review samples and exceptions, correct knowledge and track repeat contacts, quality and full costs.
Best case: a redesigned service
30%
potential benefit after costs
Approved routine questions go directly to AI. People handle exceptions and supervise quality. All released capacity is put to useful work.
- Potential value in year one
- +€180,000
- Cost-equivalent per request
- €4.20
Expand when quality, demand and usable capacity support it.
Planning case: a gradual transition
10%
potential benefit after costs
Customers keep using several channels. More cases need people, and only part of the released time can be reassigned. Start with this as a hypothesis to test.
- Potential value in year one
- +€60,000
- Cost-equivalent per request
- €5.40
Pilot, measure and adapt the channel mix.
Failed pilot: the investment is lost
€30,000
pilot spend without a benefit
Adoption or answer quality misses the agreed threshold. Stop after the modelled pilot budget is spent; the existing service continues. No useful capacity benefit is assumed.
- Potential value in year one
- -€30,000
- Cost-equivalent per request
- €6.30
Pause this design; revise the workflow or choose another route.
All three cases assume the same request volume and acceptable quality. Released time needs useful work. Cash savings require actual budget or spend changes. The failed pilot uses its entire €30,000 budget without a benefit.
What ETHORITY would work on
- Choose the right routineDefine a useful request class, approved knowledge and the customer’s route to a person.
- Design human supervisionAgree decision rights, escalation, quality checks and who owns each outcome.
- Make the benefit usablePlan where capacity goes; test full costs and quality, then adapt as customer behaviour changes.
Together we would test these conditions for reaching the best case. Assessment and implementation have separately agreed scopes.
Inspect the assumptions and calculation
The baseline covers staff effort for this request class. Common existing overhead is unchanged. AI running costs include the assumed software and ongoing technical operations; supervision, knowledge upkeep and rework must be included before estimating net time released.
| Assumption | Best case: a redesigned service | Planning case: a gradual transition | Failed pilot: the investment is lost |
|---|---|---|---|
| Net time released, after review and rework | 40% | 25% | 0% |
| Released time put to useful work | 100% | 80% | 0% |
| Useful capacity-value equivalent | €240,000 | €120,000 | €0 |
| AI running costs in year one | €40,000 | €40,000 | €10,000 |
| Setup and change work | €20,000 | €20,000 | €20,000 |
| Net benefit equivalent | €180,000 | €60,000 | -€30,000 |
Net equivalent = annual staff-effort baseline × net time released × useful share − AI running costs − setup. Percentage = net equivalent ÷ baseline. Unit cost-equivalent = (baseline − net equivalent) ÷ annual requests.
The 100% useful-share assumption is deliberately optimistic. The planning case uses 80%; the failed pilot produces none. Taxes, financing and unquantified service harm are excluded. The pilot budget is not a ceiling on every possible business loss. Replace all assumptions with pilot observations before relying on the result.
Understand the implementation choices
- Improve the existing service: Clearer help content and routing, without AI.
- Assist the service team: AI drafts; a person approves every answer.
- Direct AI support: Configure a service platform for approved routine answers and human handoff. This is the design modelled above.
- Build a focused integration: Connect a specific workflow to existing systems, with its own build and operating costs.
Compare these choices against the same quality and cost requirements. They are alternatives, not grades; the figures above compare scenarios for direct AI support only.
A 90-day plan
1–3
Step: AI Act Readiness Sprint (Enterprise): register, classification, roadmap, board briefing
2–8
Step: AI Estate Audit & Mapping: costs, usage, vendor facts, a scale / fix / stop decision
6–14
Step: AI Governance Setup: operating model, approval workflow, policies, ISO/IEC 42001 map
From month 4
Step: AI Opportunity Case for the biggest bet · AI Literacy programme · AI Governance Retainer
What we do and do not do
ETHORITY delivers: - technical and organisational assessments; - governance design; - training.
ETHORITY does not: - give legal advice; - certify systems; - test them technically.
Where needed, we work with your counsel, an accredited certification body or your IT partners.
---
Define the boundary before a pilot
People and permissions
Name the process owner, data authority, reviewer and outcome acceptor. Specify the work that may happen and where a human must intervene.
Evidence and costs
Retain the baseline, case mix, quality threshold, review effort, rework and integration costs. Include a credible non-AI alternative. Released time needs an agreed destination before its value is claimed.
Expansion or stopping
Agree the conditions for a limited test, the evidence needed to expand and the reasons to redesign or stop. A market profile establishes neither workflow suitability nor a customer outcome.