Jul 28, 2026
Dipp AI Technologies is founded to make agentic execution accountable
A new artificial intelligence research and product company, founded to close the gap between what enterprises can deploy and what they can prove.
By Odero Otieno
Dipp AI Research
We publish what we can defend.
Every figure we use in public is sourced, dated, and attributed. Where a number comes from our own analysis, we say so plainly.
How Dipp AI Research is sourced
Fig. RES-07
Inputs
D1
Primary filings
Regulatory text, standards drafts, disclosures.
D2
Vendor telemetry
Published benchmarks and incident reporting.
D3
Dipp AI analysis
Labeled as ours wherever it appears.
Publication rule
R
Sourced, dated, attributed
Every figure carries a page reference back to the volume it came from.
Vol. I · August 202637 pages31 figures
Volume I · August 2026
The Enterprise Superintelligence Report
Thirty-seven pages on the trust gap in enterprise AI: documented incidents, the failure modes of Human-in-the-Loop, the seven components of an agentic control plane, and the economics of verified execution. 191 citations.
Figure · Orcher™
Executive summary — headline findings
FIG. P03-01
78%
of enterprises run AI agents in production
Vol. I, p.3
28%
can trace an agent action end to end
Vol. I, p.6
54%
reported an agent-related incident
Vol. I, p.2
38%
can name an accountable owner
up from 7% — Vol. I, p.2
Orcher™ · Dipp AI Technologiesdippai.com
Library
Reports and research notes.
Published work, newest first. Search by topic or year; registered readers can download the full PDFs.
Filter by topic and date
12 of 12 shown
Standards
We track the regimes that will decide this.
Standards bodies, regulators, and benchmark authors are all moving. Vol. I maps where each of them currently stands.
Figure · Orcher™
Standards and regulatory landscape, 2026
FIG. P29-01
| Body or regime | Instrument | What it obliges you to show |
|---|---|---|
| OWASP | Agentic Top 10 2026, AISVS, MCP Top 10 | That identity and privilege abuse paths are closed (30+ CVEs catalogued) |
| MITRE | ATLAS — 16 tactics, 84 techniques | That adversarial agent behaviour is modelled and detectable |
| NIST | AI Agent Standards Initiative, Feb 2026 | Risk management with named accountable owners |
| EU | AI Act Article 14 + Omnibus | Effective human oversight, evidenced rather than asserted |
| CISA / Five Eyes | Joint guidance, May and Jul 2026 | Operational controls over autonomous execution |
| China | AI Agent Law, 15 Jul 2026 | Three-tier decision authorization by statute |
| Illinois | SB 315, 6 Jul 2026 | Third-party audit of automated decisions |
| Singapore | MAGF | Governance framework for agentic deployment |
| US states | ~100 measures across 38 states | ADMT disclosure and appeal rights |
6 days to map a new requirement with hashed evidence11 weeks without it
Orcher™ · Dipp AI Technologiesdippai.com
Figure · Orcher™
Three benchmark gaps nobody measures yet
FIG. P28-01
G1
Governance-Evidence Sufficiency
Would the record produced by this run satisfy the regime the enterprise answers to? Proposed as DEMM-Bench.
G2
Cost per Verified Outcome
Not tokens or latency — the fully loaded cost of one directive carried to a provable result.
G3
Owner-Harm Resistance
How often the system takes an action that harms the party it acts for. Measured at 3.7% and 14.8% in published work.
59.4% contamination observed in agent benchmarksSWE-bench Verified retired 23 Feb 202680.9% → 45.9% moving to SWE-bench Pro
Orcher™ · Dipp AI Technologiesdippai.com
Further reading · September 2026
The complete September 2026 series, all nine pieces.
Published Mondays and Wednesdays through September 2026. Free to read, no gate — from the trust gap through the frontier-era commentary.
Aug 10, 2026
Dipp AI publishes the Enterprise Superintelligence Report, Vol. I
A 37-page examination of why enterprise agentic adoption has outrun the ability to account for it — and the seven-component architecture that closes the gap.
Aug 12, 2026
Dipp AI publishes its enterprise data boundary commitment
A contractual promise not to train on customer data is not a control. Dipp AI publishes the architectural commitment it holds itself to instead.
By Odero Otieno
Aug 14, 2026
Orcher ships cost governance with hard spend ceilings
Recommending a cheaper model is advice. Refusing to dispatch above a ceiling is governance. Orcher now enforces the second.
Aug 15, 2026
Enterprise data boundaries: exclusion by architecture, not by promise
The enterprise question is not where data sits. It is what can reach it, what may leave, and whether anyone can prove which of the two happened.
Aug 17, 2026
Introducing Orcher, the agentic control plane
One layer between the enterprise and every model it uses, responsible for authority, routing, cost, compute, data boundary and evidence.
By Odero Otieno
Aug 18, 2026
Dipp AI opens the Orcher design partner programme
A small cohort of regulated enterprises will run Orcher against real workloads, real audit exposure and real spend — and shape the control plane while they do it.
By Odero Otieno
Aug 19, 2026
Dynamic model routing under a cost ceiling
Route by stakes, not by habit. A technical account of the decision that determines cost, compliance and latency in a single step.
Aug 20, 2026
The Orcher immutable audit ledger enters external review
Evidence that only its author can verify is not evidence. The ledger is being tested by people who did not build it.
Aug 24, 2026
Human-in-the-Role: a technical note on binding authority
How a standing grant is expressed, bounded, enforced, revoked and evidenced — the mechanics behind the phrase.
By Odero Otieno
Aug 26, 2026
Dipp Intelligence: how governed use compounds into institutional advantage
Models are rented and replaced. The record of what your enterprise actually did, under whose authority, and at what cost, is owned and improves with use.
Aug 28, 2026
Measuring accountability: the benchmark gap in agentic evaluation
Agent benchmarks measure whether a task was completed. Enterprises need to know whether it should have been attempted, by whom, and at what cost.
Sep 2, 2026
The Trust Gap: Why AI Adoption Has Outrun Accountability
Enterprises can trust what a model says. Very few can prove what an agent actually did, under whose authority, once it has already happened.
By Odero Otieno
Sep 7, 2026
Human-in-the-Loop Watched. Human-in-the-Role Would Have Stopped It.
A person was actively supervising every action an AI coding agent took, in real time, and a production database still got deleted. The failure was not attention. It was architecture.
By Odero Otieno
Sep 9, 2026
Introducing Orcher: The Control Plane for Enterprise Superintelligence
Seven components, two layers, one verified execution cycle — built so autonomy can run at machine speed without the enterprise losing the ability to prove, on demand, what happened and under whose authority.
By Odero Otieno
Sep 14, 2026
Introducing the Verified Execution Cycle: A New Way to Meter, Bill, and Audit AI
The industry has spent 2026 trying to bill AI agents for outcomes instead of tokens. Almost none of it has solved the harder problem underneath: proving an outcome actually happened before anyone gets billed for it.
By Odero Otieno
Sep 16, 2026
Introducing the Data Control Gateway: Why “We Won't Train on Your Data” Isn't a Control
A promise in a contract is not enforced anywhere a machine can check it. The Data Control Gateway redacts, routes, and verifies an enterprise's own data boundary at the point of every call.
By Odero Otieno
Sep 21, 2026
Cost Governance: Why Recommending a Cheaper Model Isn't Governing Cost
Model routing is not new. Almost every gateway vendor offers it. What is still missing is the difference between a router suggesting the cheaper option and a system that will not let the expensive one run unless the stakes justify it.
By Odero Otieno
Sep 23, 2026
Elastic Compute Governance: Utilisation Is a Governance Outcome, Not a Procurement Problem
The enterprises winning the next five years will not be the ones who bought the most GPU capacity. They will be the ones whose capacity is actually doing verified work — and one 23,000-cluster study suggests almost none of it currently is.
By Odero Otieno
Sep 28, 2026
Dipp Intelligence: The Only Moat That Compounds
Four of the five classic competitive moats have lost their predictive power. The one thing that cannot be bought, borrowed, or replicated is the verified record of how an organisation's own people actually work.
By Odero Otieno
Sep 6, 2026
Frontier Autonomy Needs a Control Plane: Reading the September 2026 Releases
GPT-6 Astra, Claude Fable 5.1, Claude Mythos 5.1, and Muse Spark 1.3 do not make control optional. They make it the whole game.
By Odero Otieno
Orcher early access · invitation only
We are opening Orcher to a small cohort of enterprises that must close the AI trust gap.
Applications are read by the founding team. 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.
For enterprises closing the AI trust gap · 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 founding team 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.
Corrections welcome, in writing.
If a figure we published is wrong, we want the citation. Research correspondence goes straight to the founding team.