Origin story
Built by the people who had to answer for the systems
Odero Otieno founded Dipp AI Technologies, Inc. in July 2026 to close the gap between an AI's advanced reasoning capabilities and an enterprise's ability to trust it with real work: verifiable, cost-disciplined, and in control of their own data.
How it started
An idea in March 2025, a company in July 2026.
Nothing here started with a model demo. It started with the operational reality of regulated enterprises that cannot deploy what they cannot account for.
March 2025
The idea, formed inside the work itself
Leading enterprise envisioning, design, architecture and engineering for Microsoft's top 100 enterprise customers — banking, payments, healthcare, insurance and other regulated sectors — the same conversation kept repeating. Every organisation could get a model to reason well. None of them could prove, after the fact, what an agent had actually done, under whose authority, at what cost, and against which data boundary. The gap was not model quality. It was the absence of a control plane.
2025
The problem confirmed at operating scale
As Managing Director of Engineering for Cloud, Data, and AI Platforms at Humana, the same gap appeared from the inside of a company that has to answer for its decisions: infrastructure supporting more than $130 billion in annual revenue and over 20 million customers, where an unattributable automated action is not a technical inconvenience but a regulatory event. Watching a human supervise an agent was never the same claim as controlling it.
Early 2026
Human-in-the-Role, written down
The architectural answer was to bind authority to a role before execution rather than review it afterwards: one directive, one bound role, gates in sequence, one committed action, one permanent record. That became the Verified Execution Cycle, and around it the seven components that make it enforceable.
July 2026
Dipp AI Technologies, Inc. incorporated
The company was founded in New York in July 2026 to build Orcher as one system — authority, cost, data boundary and evidence together, rather than four separate initiatives that never reconcile.
August 2026
The thesis published
The Enterprise Superintelligence Report, Vol. I set out the argument in public, with its figures and sources, so customers could judge the reasoning before they judge the product.
September 2026
Early access opens
Orcher opened to an invitation-only cohort of regulated-industry leaders, one workflow at a time, with every deployment briefed by the people who designed the components.
Founder
Founded and led by Odero Otieno.
He built Orcher, the agentic control plane, as the infrastructure that makes this possible, after engineering executive leadership at Fortune 100 scale at Microsoft and Humana. As Managing Director of Engineering for Cloud, Data, and AI Platforms at Humana, he architected infrastructure supporting more than $130 billion in annual revenue and 20 million-plus customers across healthcare, insurance, and finance. As Principal Director of Software Engineering for Enterprise Platforms at Microsoft, he led platform engineering for the company's top 100 enterprise customers.
Earlier, as Technical Fellow and Founder in Residence at Texas Medical Center, he built economic and data infrastructure for the $25 billion annual revenue medical complex. He pioneered Human-in-the-Role and authored The Enterprise Superintelligence Report, Vol. I.
What we believe
Four positions the architecture is built on.
Each one is a design constraint inside Orcher, not a value statement.
Authority is architecture, not policy
A control that lives in a document is not enforced anywhere a machine can check it. Every claim Orcher makes is enforced at execution or it is not made.
Cost is a governance outcome
Recommending a cheaper model is not governing spend. A ceiling that a directive cannot exceed is.
The data boundary is verified per call
“We won't train on your data” is a contract term. Redaction, region pinning and training-exclusion checks on every call are a control.
Evidence is a byproduct, not a project
If proving what happened requires a reconstruction exercise, the architecture was wrong. The record should already exist.
Orcher early access · invitation only
We are opening Orcher to a small cohort of regulated-industry leaders.
Applications are read by the founder. We take workflows where an autonomous action touches a system of record and someone has to answer for it — claims, care decisions, credit, filings, dispatch, procurement. Approved organisations get an invitation, a workspace, and a briefing with the people who built the control plane.
Open to all regulated industries · one workflow to start
01
Apply with one workflow
One class of directive, the system of record it touches, and the obligation behind it.
02
The founder reads it
A decision against the current cohort, normally inside one business day.
03
Invitation opens the account
Accounts only open for invited work addresses. Nothing self-provisions.
04
Workspace tracks the deployment
Status, briefings, invitations and the research library in one place.
Read the thesis, then talk to the founder.
Vol. I sets out the argument with its sources. A briefing puts your workflow against it.