Agent, or just a workflow?

The question most of the material says to ask first, and most deployments skip.

6documents on this topic
5organizations represented
2issues named
8sourced citations
0sourced statistics

The state of it

One of 6 topics within Agent orchestration.

6 documents from 5 organizations address whether the thing you are about to build needs to be an agent at all. It is the shortest topic here and the one with the clearest disagreement, and the disagreement runs along a line worth noticing: the frontier labs say build the simplest thing, and the systems integrators say you will need the full architecture.

Anthropic draws the distinction precisely - a workflow orchestrates models and tools through predefined code paths, while an agent dynamically directs its own process and tool use - and then recommends finding the simplest solution possible and only increasing complexity when needed, explicitly including the option of not building an agentic system at all. OpenAI frames the same boundary around independence. Accenture argues the opposite conclusion: that single agents alone cannot drive enterprise-scale transformation and orchestration frameworks are required.

Both positions are held by parties with an interest. Read them against each other.

The issues, by agreement

How many independent organizations name each issue as a problem. An issue is only as real as the number of separate publishers that identify it, so the count is the ranking. Bars are organizations, not documents. Where the count reads ours, no publisher here states the issue and the analysis is our own.

Where they disagree

No contradictions recorded on this topic yet.

The issues in full

Each issue carries the organizations that name it, the numbers behind it, and the remedies proposed - with the concrete steps under each. Every citation points at a section of a named document, so any count here can be checked.

Issue 013 organizations name itnewest evidence Dec 2024

An agent is being built where a predefined workflow would do the job

The architectural distinction is a decision about who chooses the next step - your code, or the model. Most deployments hand that choice to the model without having asked whether they needed to.

Anthropic states the trade explicitly: agents buy task performance with latency better task performance, and you should consider when that trade makes sense. Where the steps are knowable in advance, a workflow gives you the same outcome with behaviour you can test, cost you can predict, and failures you can reproduce. The cost of getting this wrong is not a bad answer - it is a system whose behaviour you cannot characterise, attached to a problem that never required it.

How to fix it — 2 approaches, 5 steps

Decide who chooses the next step, in writing

For each candidate, state whether the sequence of steps is knowable in advance. If it is, the answer is a workflow, and the decision is made.

Done when Each proposed agent has a written answer to whether its steps are knowable in advance, anything with a knowable sequence was built as a workflow, and the latency and cost premium accepted for dynamic control is recorded.

  1. For each proposed agent, write down the steps. If you can write them, you do not need an agent.0-30 daysCTO
  2. Reserve dynamic control for cases with genuine exceptions and open-ended paths.0-30 daysCTO
  3. Record the latency and cost premium accepted in exchange, so the trade is explicit.30-90 daysCIO

Reach for the composable patterns before autonomy

Prompt chaining, routing and parallelization solve a large share of real cases with predictable behaviour and no autonomous control loop.

Done when The record for each autonomous loop in production states which deterministic pattern was tried first and how it demonstrably failed.

  1. Try the deterministic pattern first: chain, route, or run in parallel and combine.0-30 daysCTO
  2. Escalate to an autonomous loop only where the pattern demonstrably fails.30-90 daysCTO
The evidence — 4 documents
OrganizationDocumentPosition
AnthropicFrontier lab · December 2024Building effective agentsOur reading Draws the architectural distinction - workflows orchestrate models and tools through predefined code paths, agents dynamically direct their own processes and tool use - and recommends finding the simplest solution possible, increasing complexity only when needed, which may mean not building an agentic system at all.Section: workflows versus agents; when (and when not) to use agentsnames it
IBMHyperscalerAgentic AI risk and opportunityOur reading Names the choice outright rather than assuming the agentic route: it holds that the simplest answer to most business problems is a designed orchestration of conventional models and language models, that agents will not suit every problem, and that an agent is warranted only where the complexity of the problem earns it - because the harder architecture behaves less predictably.Risks and key mitigations of autonomous actionnames it
OpenAIFrontier labPractical guide to building agentsOur reading Defines a workflow as a sequence of steps that must be executed to meet a goal, and an agent as a system that independently accomplishes tasks on the user's behalf, reserving agents for workflows involving nuanced judgment, exceptions and context-sensitive decisions that have resisted conventional automation.What is an agent?; when should you build an agent?names it
OpenAIFrontier labPractical guide to building agentsOur reading Contrasts a traditional rules engine working like a checklist against an agent evaluating context and subtle patterns, positioning the agent as the answer only where clear-cut rules do not apply.When should you build an agent?names it

Issue 021 organization name itnewest evidence Apr 2025

Agent frameworks hide the one thing you most need to control

Framework abstractions make a first agent easy and make it hard to see or govern what context the model actually receives at each step - which is the variable that determines whether it works.

LangChain's response to OpenAI's guide is the useful text here because it argues against the tooling category it sells into. Its point is that most agentic frameworks are not orchestration frameworks at all but sets of agent abstractions, and that those abstractions obfuscate what the model sees at each step. For an executive the consequence is concrete: when the system misbehaves, the team cannot tell you what it was looking at when it decided, and the framework is the reason.

How to fix it — 1 approach, 3 steps

Require the context to be inspectable

Whatever the framework, the team must be able to show exactly what the model received at any step of a past run.

Done when For a past run picked at random, the team can show the exact context the model received at a given step, because the assembled context is logged alongside the action log rather than the final prompt alone.

  1. Make "can you show me the exact context at step N of a past run" an acceptance test for any framework.0-30 daysCTO
  2. Log the assembled context alongside the action log, not just the final prompt.30-90 daysCTO
  3. Reject frameworks that cannot answer it, however fast the first demo was.0-30 daysCTO
The evidence — 4 documents
OrganizationDocumentPosition
LangChainVendor · April 2025CEO responds to OpenAI on agentsOur reading Argues that most agentic frameworks are neither declarative nor imperative orchestration frameworks but simply sets of agent abstractions, and that those abstractions can obfuscate and make it hard to ensure the model has the appropriate context at each step.Response to the OpenAI guidenames it
LangChainVendor · April 2025CEO responds to OpenAI on agentsOur reading Names the cost of the convenience: the abstractions supplied to build agents make it hard to see or control what actually reaches the model at each step, which is stated as a problem with no caveat attached.What the abstraction hidesnames it
AnthropicFrontier lab · December 2024Building effective agentsOur reading Recommends starting from the simplest composable pattern rather than a framework, on the grounds that added abstraction is added complexity to justify.When (and when not) to use agentsproposes a fix
GoogleHyperscaler · September 2024Agents (Google whitepaper)Our reading Separates the agent into model, tools and an explicit orchestration layer, making the orchestration a component to be designed rather than inherited from a framework.The model, the tools, the orchestration layerproposes a fix

Who is represented

This dossier is drawn from 8 organizations working on the subject, 5 of which are cited directly in the issues above.

Consultancy — 1

Accenture 1

Academic — 1

arXiv (research) 1

Hyperscaler — 2

Google 1 IBM 1

Frontier lab — 2

Anthropic 2 OpenAI 1

Vendor — 2

LangChain 1 Cursor 1