Agent orchestration
The third-largest theme in the research: 194 of 268 documents, 70 organizations. Decomposed into 6 topics, each with its own issues, evidence and steps.
194 documents in the theme
12 issues named
43 sourced citations
17 organizations cited
0 sourced statistics
The topics
Each opens a dossier with the same shape: what the state of it is, the issues ranked by how many independent organizations name them - or marked as our own analysis where none does - the disagreements, the numbers, and the concrete steps under each remedy.
Multi-agent systems
Several agents on one problem, and the coordination that costs more than it looks.
Multi-agent systems work largely by spending more tokens, and the budget rarely says so 3
Multi-agent is being adopted as a maturity stage rather than chosen for a problem 1
48 documents · 27 organizations · 2 issues · 7 steps
Tools & integration
What the agent can reach, how it is granted, and what breaks at the boundary.
Tool definitions are permission grants, and are being designed as features 2
Three different kinds of tool are being governed as one ours
37 documents · 23 organizations · 2 issues · 8 steps
Autonomy limits & human control
How far the system may act alone, and what stops it.
Risk to people rises with autonomy, and the trade is almost never stated 4
Autonomy is being set once, as a binary, instead of earned incrementally 2
21 documents · 17 organizations · 2 issues · 8 steps
Context & memory
The finite budget every agent runs against, and what happens when it fills.
Context is a finite budget, and long-running agents spend it without anyone watching 2
Cost compounds invisibly because one request is not one request 2
16 documents · 13 organizations · 2 issues · 8 steps
Reliability & failure modes
Non-determinism, compounding error, and how agent failure differs from software failure.
Non-determinism breaks the testing process the organization already has 2
There is no accepted way to evaluate an agent, so every team invents one 1
15 documents · 12 organizations · 2 issues · 8 steps
Agent, or just a workflow?
The question most of the material says to ask first, and most deployments skip.
An agent is being built where a predefined workflow would do the job 3
Agent frameworks hide the one thing you most need to control 1
6 documents · 5 organizations · 2 issues · 8 steps
Where the sources disagree
No contradictions recorded yet.
From the toolkit
Agent orchestration — the implementation kit
The whole of Agent orchestration, turned into something you can run. A diagnostic that tells you which of these problems you have, and an action plan with every step owned and time-boxed.
A diagnostic you can run in a room 12 questions across 6 areas, each written so a yes or no tells you whether you have that problem. No scoring model to learn.
An action plan that names who does it 47 actions, each carrying a role and a time-box, and every fix states what exists when it is done - so you can tell a fix that landed from one that was attempted.
The same actions, sorted by person An owner map, so one column goes to one person, and a sequence that says what to do first rather than leaving you to guess.
Written for your situation Three editions - listed company, private company or scale-up, and advisory - so the owner names match the room you are actually in.
Yours to use in front of a client Every word is original work. No third-party research is reproduced in it, which is what makes it safe to hand on.
12 diagnostic questions · 47 owned actions · 26 pages · one-off, updates included
Issues were named by hand after reading the documents cited under each one. Consensus counts distinct organizations, not documents, and counts only evidence a human has verified against a located passage.
Every link opens the publishing organization's own page. Summaries and characterisations are written here; no publisher prose is reproduced.