In Microsoft Foundry, a workflow is a declarative definition of a sequence of steps involving AI agents, conditions, and human interactions. Instead of writing imperative code to manage execution flow, you define what should happen and when using a visual, node-based approach.
<\/p>\n\n\n\n
The platform handles:<\/p>\n\n\n\n
- \n
- Execution order and state management<\/li>\n\n\n\n
- Passing outputs between steps<\/li>\n\n\n\n
- Conditional branching<\/li>\n\n\n\n
- Human-in-the-loop pauses<\/li>\n\n\n\n
- Error handling and retry<\/li>\n<\/ul>\n\n\n\n
You focus on the logic. Foundry manages the execution.<\/p>\n\n\n\n
The Core Concepts: Nodes and Edges<\/strong><\/p>\n\n\n\n
Every Foundry workflow is a graph. The graph has nodes (the workers) and edges (the connections between them).<\/p>\n\n\n\n
![\"\"](\"\/wp-content\/uploads\/2026\/04\/image-22-1024x683.png\")
<\/figure>\n\n\n\nNodes (called Executors in Foundry) are the workers. They can be:<\/p>\n\n\n\n
- \n
- AI agents – the reasoning, language-capable workers<\/li>\n\n\n\n
- Custom logic components – deterministic code for calculations, data transforms<\/li>\n\n\n\n
- Human approval steps – pauses waiting for a person<\/li>\n<\/ul>\n\n\n\n
Edges define how data and control flow between nodes. There are five types:<\/p>\n\n\n\n
Edge Type<\/strong><\/td> Behaviour<\/strong><\/td><\/tr> Direct<\/td> Node A always feeds into Node B<\/td><\/tr> Conditional<\/td> Node B only triggers if a condition is met<\/td><\/tr> Switch-case<\/td> Routes to different nodes based on predefined conditions<\/td><\/tr> Fan-out<\/td> Sends one message to multiple nodes simultaneously<\/td><\/tr> Fan-in<\/td> Combines outputs from multiple nodes into one final step<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n The Three Main Workflow Patterns<\/strong><\/p>\n\n\n\n
Pattern 1: Sequential Workflow<\/strong>
Steps execute in a fixed, ordered sequence. Each step’s output becomes the next step’s input.<\/p>\n\n\n\n![\"\"](\"\/wp-content\/uploads\/2026\/04\/image-23-1024x455.png\")
<\/figure>\n\n\n\nBest fo<\/strong>r: Predictable multi-stage pipelines where each step depends on the previous one.
Real example: <\/strong>Customer support triage. Extract the customer issue -> Classify as billing\/technical\/general -> Enrich with customer account history -> Route to the appropriate support queue.
Pros:<\/em> Easy to reason about, predictable execution, straightforward debugging
Cons:<\/em> One slow step blocks everything; not suited for parallel workloads<\/p>\n\n\n\nPattern 2: Human-in-the-Loop Workflow<\/strong><\/p>\n\n\n\n
The workflow pauses at a defined point and waits for a human decision before continuing.<\/p>\n\n\n\n
![\"\"](\"\/wp-content\/uploads\/2026\/04\/image-24-1024x563.png\")
<\/figure>\n\n\n\nBest for:<\/strong> High-stakes or irreversible actions. Anything where you need a human to verify before committing.
Real example:<\/strong> Financial reconciliation agent. Agent identifies discrepancies, prepares a summary report, then pauses. A finance team member reviews and either approves the corrective journal entries or sends back for investigation.
Pros<\/em>: Maintains human oversight for sensitive operations; builds trust in AI-assisted processes
Cons:<\/em> Workflow speed is limited by human response time; requires clear notification and escalation design<\/p>\n\n\n\nPattern 3: Group Chat (Collaborative) Workflow<\/strong>
Multiple specialised agents share a common conversation. Control shifts dynamically based on context.<\/p>\n\n\n\n![\"\"](\"\/wp-content\/uploads\/2026\/04\/image-25-1024x690.png\")
<\/figure>\n\n\n\nBest for:<\/strong> Tasks requiring different specialisations – where no single agent has all the expertise needed.
Real example: <\/strong>Report generation. Research agent gathers and summarises sources. Writing agent drafts the narrative. Review agent checks for factual consistency and flags gaps. Output is a collaborative product of all three.
Pros: <\/em>Specialisation leads to better quality; agents can build on each other’s work
Cons: <\/em>More complex to configure and debug; requires a well-designed orchestrator to avoid loops<\/p>\n\n\n\nWorkflow Events: Understanding Execution State<\/strong><\/p>\n\n\n\n
One of the more technical aspects of Foundry workflows is the event system. Understanding what events fire and when is essential for debugging.<\/p>\n\n\n\n
![\"\"](\"\/wp-content\/uploads\/2026\/04\/image-26-1024x563.png\")
<\/figure>\n\n\n\nWhen a workflow behaves unexpectedly, start by examining the event log in sequence. The gap between events – or a missing event – tells you exactly where things broke down.<\/p>\n\n\n\n
Designing for Failure<\/strong><\/p>\n\n\n\n
Every node in a workflow can fail. The questions to ask when designing:<\/p>\n\n\n\n
- \n
- What happens if this node times out? Define timeouts per node and at the workflow level.<\/li>\n\n\n\n
- What happens if an agent call returns an error? Retry? Skip? Notify a human? Abort?<\/li>\n\n\n\n
- What happens if a tool call fails mid-workflow? Is the workflow idempotent? Can it safely retry from the beginning?<\/li>\n\n\n\n
- What is the compensation strategy? If step 5 fails after steps 1-4 completed, do you need to undo those steps?<\/li>\n<\/ol>\n\n\n\n
A workflow without failure design is a workflow that will behave unpredictably in production. Design the unhappy path first.<\/p>\n\n\n\n
A Note on Monitoring Workflows<\/strong><\/p>\n\n\n\n
Foundry workflows emit telemetry you can capture in Azure Monitor and Application Insights. Set up:<\/p>\n\n\n\n
- \n
- Workflow duration – alerting on unusually long runs<\/li>\n\n\n\n
- Error rate – alerting on node failures<\/li>\n\n\n\n
- Token consumption – tracking LLM costs per workflow execution<\/li>\n\n\n\n
- Human approval lag – how long are human-in-the-loop steps actually taking?<\/li>\n<\/ul>\n\n\n\n
That last one is often ignored. If human approval steps are taking 4 hours average when the SLA requires 1 hour, the bottleneck is human, not technical – and you need to know that.<\/p>\n\n\n\n
Pros and Cons of Agent Workflows<\/strong><\/p>\n\n\n\n
Pros:<\/em><\/p>\n\n\n\n
- \n
- Automate complex multi-step processes that previously required human coordination<\/li>\n\n\n\n
- Visual, declarative approach is accessible to solution architects and business analysts<\/li>\n\n\n\n
- Human-in-the-loop patterns provide oversight without sacrificing automation<\/li>\n\n\n\n
- Nodes are modular – individual agents can be updated without rebuilding the workflow<\/li>\n<\/ul>\n\n\n\n
Cons:<\/em><\/p>\n\n\n\n
- \n
- Debugging distributed workflow execution is harder than debugging sequential code<\/li>\n\n\n\n
- State management across nodes requires careful design<\/li>\n\n\n\n
- Fan-out\/fan-in patterns introduce complexity around output merging<\/li>\n\n\n\n
- Human-in-the-loop workflows are only as fast as the humans in them<\/li>\n\n\n\n
- Costs accumulate across every node – multi-agent workflows can be expensive<\/li>\n<\/ul>\n\n\n\n
My Rule for When to Use a Workflow vs. a Single Agent<\/strong><\/p>\n\n\n\n
A single agent with good tools can handle a surprising amount of complexity. Before building a workflow, ask:
Can one well-configured agent with the right tools complete this end-to-end?<\/em>
If yes – use a single agent. Simpler to configure, simpler to debug, lower cost.
If the process genuinely requires different specialisations, parallel execution, conditional branching based on business logic, or human approval gates – that’s when you reach for a workflow.
Don’t over-engineer. Start with the simplest thing that works.<\/p>\n\n\n\nNext: I’ll cover multi-agent orchestration patterns using the Microsoft Agent Framework – where the orchestration moves into code and you get much more fine-grained control.<\/em><\/p>\n","protected":false},"excerpt":{"rendered":"
A single well-configured AI agent is impressive. But most real business processes aren’t single tasks – they’re sequences of tasks, with decisions, branches, approvals, and handoffs in between. That’s where agent-driven workflows come in.Think of workflows as the difference between hiring one very capable person and building a team with a clear process. The person […]<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[1],"tags":[],"_links":{"self":[{"href":"https:\/\/bogdanburuiana.com\/index.php\/wp-json\/wp\/v2\/posts\/1507"}],"collection":[{"href":"https:\/\/bogdanburuiana.com\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/bogdanburuiana.com\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/bogdanburuiana.com\/index.php\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/bogdanburuiana.com\/index.php\/wp-json\/wp\/v2\/comments?post=1507"}],"version-history":[{"count":1,"href":"https:\/\/bogdanburuiana.com\/index.php\/wp-json\/wp\/v2\/posts\/1507\/revisions"}],"predecessor-version":[{"id":1514,"href":"https:\/\/bogdanburuiana.com\/index.php\/wp-json\/wp\/v2\/posts\/1507\/revisions\/1514"}],"wp:attachment":[{"href":"https:\/\/bogdanburuiana.com\/index.php\/wp-json\/wp\/v2\/media?parent=1507"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/bogdanburuiana.com\/index.php\/wp-json\/wp\/v2\/categories?post=1507"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/bogdanburuiana.com\/index.php\/wp-json\/wp\/v2\/tags?post=1507"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}