Enable every decision of consequence.

Step 01

Some decisions cannot be wrong.

A decision of consequence is one where being wrong costs money, liberty, safety, rights or lives — and where someone must be able to show afterwards what was read, what was concluded and who decided. Judges, financial-intelligence analysts, registrars, procurement officers, care coordinators, group executives and programme directors make them every day, mostly by hand, from documents no model has ever read.

Step 02

Intelligence is solved. Accountability is not.

Frontier models are commoditised. Cloud is funded. Citizen-facing AI is mature and crowded. What is not solved is the layer between the model and the institution: the case files, suspicious-transaction reports, registry filings, tender documents, contracts, leases, board papers, field reports and audit trails. As AI Acts arrive, officials will have to show what the AI read and who decided. Generic assistants cannot produce that record.

• ETDA draftETDA draft · Electronic Transactions Development AgencyDraft AI legislation from Thailand's Electronic Transactions Development Agency
• EU high-risk rulesEU high-risk rules · European Union high-risk AI rulesThe EU AI Act's duties for high-risk AI systems
• CMS-0057-FCMS-0057-F · CMS-0057-FA United States health-regulator final rule on interoperability and prior authorisation
• SR 11-7SR 11-7 · SR 11-7United States banking supervisors' guidance on model risk management

The position →

Step 03

Six kinds of institution own, fund, legitimise and deliver these decisions.

Governments and regulated enterprises own the decisions. Development banks and global funds finance them. The UNUN · United NationsThe organisation of member states, founded in 1945, whose agencies, funds and programmes set international norms and run much of the world's humanitarian caseload. system and regional bodies set the norms. NGOsNGO · Non-governmental organisationA not-for-profit body independent of any government. In this landscape, NGOs are often the implementing partners who deliver programmes on the ground. and implementing partners deliver on the ground. If you have never worked in this world, start here.

See the forest →Five-minute primer →

Step 04

Money flows down. Evidence flows up. Legitimacy flows across.

Shareholder governments capitalise banks and replenish funds. Banks and funds finance programmes. Governments and implementing partners deliver them. Results travel back up as reports, often compiled by hand once a year. Every hand-off is a decision of consequence built on documents.

How the system moves →

Step 05

Partners open the door. A rapid proof earns the seat. Replication does the rest.

NGOsNGO · Non-governmental organisationA not-for-profit body independent of any government. In this landscape, NGOs are often the implementing partners who deliver programmes on the ground. and implementing partners hold the relationships that get us into the room. UNUN · United NationsThe organisation of member states, founded in 1945, whose agencies, funds and programmes set international norms and run much of the world's humanitarian caseload. agencies and regional bodies confer legitimacy. MDBsMDB · Multilateral development bankMultilateral development bank: a bank owned by governments that lends and grants for development and global funds set the scorecard and finance what a rapid proof of concept has shown to work. Technology vendors are the platform we run on — never the audience we serve. One proven decision becomes a catalogue service; one agency becomes the pattern for the ministry; one country the template for the hemisphere.

Through the fabric →

Step 06

OrgBrain: a governed layer of intelligence under the decision.

Four layers, drawn the same way in every conversation: the client's infrastructure, a governed knowledge layer, domain agents, and the people who keep the decision. Three principles are non-negotiable: LLMLLM · Large language modelA general-purpose text model. Sphere's first principle is LLM-agnostic: the client chooses the model; OrgBrain sits above it.-agnostic, every answer cited, data never leaves.

• LLM-agnosticLLM · Large language modelA general-purpose text model. Sphere's first principle is LLM-agnostic: the client chooses the model; OrgBrain sits above it.
• every answer cited
• data never leaves

Open the stack →

Step 07

Trillions ride on these decisions. One hundred countries at US$100m a year is US$10bn.

• US$2tn to US$4tn a yearvalue at stakeS-06McKinsey Global Institute, The economic potential of generative AIMcKinsey Global Institute · as of 2023-06as of 2023-06
• About US$200bn a yearserviceable marketS-01Sphere Corporate Strategy, September 2026Sphere · as of 2026-09as of 2026-09Estimate
• US$10bn a yearobtainable revenueS-01Sphere Corporate Strategy, September 2026Sphere · as of 2026-09as of 2026-09Estimate

The opportunity →

Step 08

We earn our seat every day.

Four written guardrails — data sovereignty, a person keeps the decision, auditability, consent and exit — each end in a question the client can put to any vendor, including us. Adjudication, enforcement, sanctions, targeting, threat-to-life triage and benefit denials are red lines. They are never automated.

• Adjudication
• enforcement
• sanctions
• targeting
• threat-to-life triage
• benefit denials

The trusted partner →

Step 09

Now Latin America and the Caribbean. Next Thailand and ASEAN. Then partner-led everywhere.

National programmes are live or starting in Ecuador, Paraguay, the Dominican Republic and Colombia. The October 2026 Annual Meetings in Bangkok open three doors at once: the Royal Thai Government, Central Group and Oracle Thailand.

Now, next, then →