Signal state: amber · act this quarter
Wisdomia Atlas · SWOT edition · October 2026

The AI Agents 101 Atlas

The machines have stopped talking and started acting. One hundred and one players, protocols, case studies, risks and moves that decide who wins the agent economy, mapped as strengths, weaknesses, opportunities and threats, with one question running through every page: who stops the agent?

By Dinis Guarda · with Jack Widman and Phil Grady of Sheaf · Ztudium · Wisdomia
THE 101 AS A SIGNAL STRIP · HEIGHT = ATLAS WEIGHT · COLOUR = SWOT QUADRANT · TAP A BAR TO OPEN ITS ENTRY
I · The explosion in numbers

Small in dollars, vast in consequence

Three forecasters land within ten per cent of each other for 2030. Adoption is wide and shallow, failure rates are high, and machines already outnumber people inside the enterprise.

The 2030 convergence

AI agents market, US$ billions. Curves are implied by each firm's stated base and growth rate.

Grand View Research via RaftLabs · MarketsandMarkets via Toolradar · BCC Research

Wide, then shallow

Share of organisations, McKinsey State of AI 2025 (1,993 respondents, 105 countries).

109 machines for every human

Machine identities per human in the enterprise. 1 human, 79 AI agents, 29 other machine identities.

Up from 82 to 1 a year earlier. AI agents expected to grow 85% this year. 9 in 10 organisations had an identity-related breach in 12 months.

Gartner's double signal

Where agents are heading, and how many projects will not survive the trip.

II · The SWOT field

Four quadrants, one verdict: demand is not the constraint, trust is

Every one of the 101 entries sits in one quadrant. Tap a quadrant to filter the explorer below.

III · The breakthrough

OpenShell is the line. Sheaf works the signals.

NVIDIA OpenShell decides what an agent may reach, enforced at the kernel where the agent cannot argue. Sheaf proves whether what it is about to do adds up, and holds it when it does not.

Try it · the hidden contradiction

Three agents on one mortgage file. No two of them disagree directly. Flip agent C and watch the signal.

A · Income agentverified = declared
B · LTV agentdeclared = used for LTV
C · Pricing agentused for LTV ≠ verified
0.41H¹ · contradiction · held by the Gate, Warden calls a person
Pairwise check: A vs B ✓ · B vs C ✓ · A vs C ✓ — every pair looks fine.
Coherence scores 1.00 and 0.41 are Sheaf's published illustrative values. How Sheaf works
IV · Eight industries

One number, one risk, one fix

The pattern repeats in every sector: a dramatic gain in speed, a new category of risk, and a human who must still answer for the outcome.

V · Identity

Every agent needs a passport

Identity answers which software acted. Authorisation answers what it may do. Neither answers whether the decision was sound. A complete agent passport has seven pages. Tap a page to see which standard already stamps it.

VI · The platform war

Eighteen months that rewired the stack

The battlefield moved from the model to the runtime, the protocol and the payment rail.

VII · The 101

The AI Agents 101

Players, protocols, case studies, data points, risks and moves, ranked by editorial atlas weight. Search, filter by quadrant, kind or industry, and open any entry for its source.

VIII · What to do now

Five moves per audience, each inside one quarter

Tick them off. Your progress stays in this browser.

IX · The readiness check

What class does your organisation fly?

Eight questions, thirty seconds. Your boarding pass shows how ready you are to let agents act in your name.

X · Who is in this Atlas

The players shaping the agent economy

Runtime and trust

NVIDIA · Sheaf · Cisco · CrowdStrike · Palo Alto Networks · Microsoft Security · TrendAI

Agent platforms

OpenAI · Anthropic · Google · Meta · Microsoft · ServiceNow · Salesforce · SAP

Payments and identity

Mastercard · Visa · Stripe · Fiserv · Backbase · PayPal · Shopify

Industries and builders

Allianz · State Farm · Epic · IQVIA · Netflix · Polsia · OpenClaw community

The boy enchants a broom to carry water, forgets the word that stops it, and splits it with an axe, only to get two brooms. We have built the brooms. Now we need the word.After Goethe, Der Zauberlehrling, 1797 · Dinis Guarda
XI · Method and sources

How this Atlas was built

Facts come from the sources below, as of 9 October 2026. Atlas weight and SWOT placement are Dinis Guarda's editorial judgement, not a market measure. Where a figure is a vendor's own claim, the entry says so.