Search for a command to run...
No comments yet. Be the first to comment.
The bigger picture. Why RevOps is broken, what AI agents are doing to HubSpot portals, and how to think about governance before things get out of hand.
Howly, the read-only HubSpot workflow mapping and audit tool, has been named a New & Notable app in the HubSpot Marketplace for Q2 2026. HubSpot recognized Howly in its quarterly marketplace roundup c
Ask Claude what's wrong with your HubSpot workflows. Get a prioritized list back, with links to fix each one yourself.
A property change, a list membership, five workflows firing in sequence. Here's how to see the whole chain instead of trusting that it works.
Quick Answer: A HubSpot workflow audit is the process of inventorying every workflow in a portal, mapping how they connect through direct enrollment, list membership, and property changes, and scoring
Quick Answer: A HubSpot workflow audit for a new client starts with taxonomy, not automation. Before touching a single workflow, an experienced consultant maps the naming conventions, separates workfl
July's Howly release removes the 50-workflow cap on free accounts, so the entire portal is visible on the free plan. Data agent and agent hub actions now surface directly on the workflow canvas. The audit report has been rebuilt with best practice recommendations, one-click links back into HubSpot, and full brand customization. Navigation also got faster, with a new search box, a Recent Changes panel, and trigger-type filtering.
The portal has 220 workflows, a mix of active and inactive ones, and roughly a third were built by an account manager who left the agency two years ago. There is no documentation, no naming convention, and the client just asked you to "make sure nothing breaks" before their migration next month.
This is the work, and it isn't the building. It's the understanding, and on a large HubSpot portal, understanding is the hard part.
To audit a large HubSpot workflow setup, work in five stages. First, inventory every workflow, active and inactive, with its object type, trigger, status, and last-modified date. Second, map dependencies: how each workflow connects to others through direct enrollment, list membership, and property changes. Third, score portal health by finding orphaned workflows, stale workflows, broken enrollments, and naming collapse. Fourth, document the result in a format the client can read and act on. Fifth, plan new workflows against the existing map so you can see what a change will connect to before you build it. For portals past roughly 50 workflows, the dependency-mapping stage is where manual auditing breaks down. A dedicated mapping tool like Howly builds the connection map automatically in 10 to 25 seconds, read-only, instead of the days it takes to trace it by hand.
A portal with 15 workflows can be held in one person's head. A portal with 220 cannot.
The native HubSpot workflow list shows you names, statuses, and folders, but it does not show you connections. It will not tell you that turning off "Set as MQL" silently breaks the enrollment trigger on three downstream workflows, or that a single contact property feeds enrollment criteria across nine separate automations. That information exists in the portal, but it is scattered across 220 individual workflows, and the only way to assemble it natively is to open each one and read its triggers and actions.
For an agency, this problem multiplies, because you are not auditing one portal. You are auditing every client portal you manage, repeatedly: before a migration, before a handoff, before a quarterly review, and every time a client asks "can you just change this one field?"
Three patterns make large portals dangerous to change:
Hidden dependency chains. Workflow A enrolls contacts into a list, Workflow B triggers off that list, and Workflow C triggers off a property that Workflow B sets. None of this is visible from the workflow list, so when you change A, C behaves differently with no warning.
Property blast radius. A single property like Lifecycle Stage can be referenced in dozens of enrollment triggers, branch conditions, and set-property actions. Renaming it, or changing its values, ripples across the portal in ways that are nearly impossible to predict by reading workflows one at a time.
Naming collapse. Naming conventions have collapsed. A portal where workflows are named "New Workflow — Copy (2)" or "Mike's test — FINAL" has lost the organizational structure that makes scale manageable. You cannot audit what you cannot identify.
The blast radius of any change is the real risk. And the blast radius is exactly what the native interface hides.
The portal that took years to accumulate 220 workflows can now grow far faster, because building a workflow no longer requires opening the workflow tool and assembling it by hand. HubSpot's Breeze AI lets users build workflows by describing the goal in natural language, which means anyone on a team can spin up automation in a sentence. That is a genuine productivity gain, and it is also the fastest sprawl engine the platform has ever had.
The problem is that AI accelerates creation without touching visibility. A workflow built from a prompt connects to the same lists and properties as one built by hand, but it arrives with no naming discipline, no documentation, and no awareness of what already exists in the portal. Ten team members each describing a few goals a month produces dozens of new workflows that nobody mapped, named consistently, or checked against the existing dependency graph. The accumulated, undocumented result is workflow debt, and it compounds the same way technical debt does: quietly, until a change you thought was safe breaks something nobody knew was connected.
AI is also adding a new connection type to track. HubSpot's Run Agent workflow action, in private beta as of early 2026, lets a workflow trigger an AI agent as one of its steps, and Breeze data actions let workflows call AI to read and write record properties inline. These are dependencies in their own right: a workflow that depends on an agent's output, or an agent that writes a property a dozen other workflows read, is exactly the kind of connection a dependency map has to capture. As agents become composable steps inside workflows rather than standalone features, the graph gets denser, not simpler.
https://youtu.be/14_EyOX-SyA?si=pBNvMGVe-92SW8Eo
I Used HubSpot Breeze AI to Build Workflows. Here's What Howly Found.
The conclusion is the one this guide keeps returning to. AI does not reduce the need to audit; it raises it. If your properties and naming are inconsistent, letting AI build workflows on top of them just automates the chaos faster. The discipline that follows, mapping what exists, scoring its health, documenting it, and planning new workflows against the map, is what keeps AI-accelerated creation from turning into AI-accelerated debt.
This is the process that works on portals of any size. The first three stages are the audit, the fourth turns it into something the client can use, and the fifth keeps the portal from sliding back into the mess you just cleaned up.
Start with a complete list of every workflow, not just the active ones. Inactive workflows are not harmless: an inactive workflow can still hold the only documentation of a process the client intends to resume, and a "temporarily" paused workflow is one of the most common sources of silent breakage when someone reactivates it without understanding what changed around it.
For each workflow, capture:
Name and folder
Object type: Contacts, Companies, Deals, Tickets, and any custom objects
Status: active or inactive
Enrollment trigger: what causes a record to enter
Number of actions and what kinds (set property, send email, create task, enroll in another workflow)
Last modified date, your first signal of staleness
The output of this stage is a flat inventory. It tells you what exists, but it does not yet tell you how any of it connects, which is the next stage and the one that matters most.
This is where the time goes, because dependency mapping is the difference between a list of workflows and an understanding of a system.
There are three connection types you must trace, and all three are invisible in the native workflow list:
Direct enrollment. One workflow enrolls a record directly into another. These form enrollment chains: sequences of workflows linked end to end. Break one link and the whole chain downstream stops receiving records.
List-based. A workflow adds a record to an active list; another workflow uses that list as its enrollment trigger. The two workflows never reference each other directly. The connection lives in the list.
Property-based. A workflow sets a property value; other workflows trigger or branch on that value. This is the most common and the most dangerous, because a single property change can alter behavior across the entire portal.
To see how these compound, follow one realistic chain. An operational workflow called "Set as MQL" sets Lifecycle Stage to Marketing Qualified Lead. A second workflow, "MQL Welcome and Sales Follow-up," enrolls every contact whose Lifecycle Stage becomes Marketing Qualified Lead and creates a task for the sales rep. A third workflow adds those same contacts to an active list, and a fourth uses that list as its enrollment trigger to start a nurture sequence. Four workflows, three connection types, one original cause. Turn off the first workflow to "clean up," and the task creation stops, the nurture sequence quietly stops receiving anyone, and the sales team starts asking why their MQL queue went dry. Nothing in the workflow list would have warned you, because none of those four workflows names the others.
Every workflow and every connection in the portal, mapped onto a single canvas you can actually read.
Tracing these three connection types manually across 220 workflows is the step that turns a one-day audit into a one-week audit. You are opening each workflow, reading its triggers and actions, noting every list and property it touches, then cross-referencing all of it to find which workflows share a list or a property. The cross-referencing is where errors creep in, and a missed connection is exactly the kind of thing that breaks a client portal after you have signed off on the audit.
This is the stage a dedicated mapping tool exists to solve. Howly connects to the portal read-only via OAuth and detects all three connection types automatically, rendering direct enrollment, list-based, and property-based connections as a single visual canvas. It loads a full portal workflow map in 10 to 25 seconds depending on portal size, so you see the enrollment chains, the shared lists, and the shared properties at a glance instead of reconstructing them by hand.
The tool does not change anything, and it cannot. It reads the portal and draws the map.
Once you can see the connections, you can find the problems. A health audit looks for four specific structural issues:
Orphaned workflows: active workflows with no upstream or downstream connections. An orphan is either doing something the client forgot about or doing nothing at all, and both need a decision.
Stale workflows: active workflows not modified in 6+ months. Staleness is not automatically a problem, but a stale workflow that no one remembers building is a prime cleanup candidate and a common source of unexpected enrollments.
Broken enrollments: workflows that enroll based on a list or property that no longer behaves as the original builder expected, often because an upstream workflow was changed or turned off.
Naming and folder collapse: the structural disorder that makes everything else harder. This is qualitative, but it is real, and it is the first thing a client notices when they take the portal back over.
A single health score for the portal, backed by the specific workflows that are dragging it down.
The output of this stage is a prioritized list: what is safe, what needs review, and what should be archived or deleted. Howly's Health Checker produces a 0 to 100 portal health score and flags these issues directly, so the audit starts from evidence rather than from opening every workflow one by one looking for trouble.
For the property question specifically, "what happens if I change this field?", the Impact Analyzer shows the property-level blast radius before a change is made: every workflow that references the property, and how. That single view answers the question that causes the most accidental breakage on large portals.
The blast radius of one property, showing every workflow that would be affected before you change a thing.
An audit that lives in your head is worth nothing the moment you hand the portal back. The deliverable is the point.
Good workflow documentation for a client includes the full inventory, the dependency map, the health findings, and a plain-language summary of what is connected to what, written for someone who owns the portal but did not build it. For a pre-migration or pre-handoff engagement, this document is the thing that lets the client safely make changes after you are gone.
Howly exports a branded PDF report, white-label for agencies, so the dependency map and health findings go to the client under your name rather than as a screenshot dump. For agencies, the documentation is often the billable deliverable, and a clean, branded report is what justifies the engagement.
The first four stages tell you what exists and how it connects. The fifth stops you from recreating the problem the day after you finish cleaning it up.
Every workflow you add to a mature portal is a new set of connections, and the danger is identical to the danger of changing an existing one: the new workflow enrolls off a property something else sets, or writes to a list something else reads, and you do not find out until a record behaves unexpectedly in production. On a large portal you cannot hold the whole connection graph in your head while you build, which is exactly why new automation is where fresh dependency problems get introduced.
The fix is to plan against the map before you build. You sketch the workflow you intend to create, configure its trigger and actions the way you would in HubSpot, and look at how it would connect to everything already in the portal before any of it goes live. If the new enrollment trigger collides with an existing one, or the property you plan to set is read by six workflows you forgot about, you see it at the planning stage rather than after launch.
This is what Howly's planning workflows feature does. You add a planned workflow to the canvas and configure its triggers and actions using your real portal data, and Howly draws every connection it would have, before you build a thing. You see the downstream connections before anything goes live, and you catch conflicts at the canvas rather than in HubSpot. For the agencies that use it most, this is the step that turns the audit from a one-time cleanup into an ongoing discipline: the portal stays mapped, and every addition is checked against the map before it ships.
A planned workflow configured against real portal data, with every connection it would create drawn before it is built in HubSpot.
Before evaluating a dedicated tool, it is worth being precise about where native HubSpot stops, because the gap is the entire reason this category of tooling exists.
Native HubSpot is strong at the things it was built for. The workflow list shows every workflow with its status, folder, object type, and last-modified date, and you can filter and search it. Inside a single workflow, the editor shows that workflow's enrollment triggers, its branches, and its actions clearly. Folders let you impose some organization, and the enrollment history on a given workflow tells you what has entered it. For building and editing one workflow at a time, the native experience is absolutely fine.
What native HubSpot does not do is show you the relationships between workflows. There is no view that says "these nine workflows all read Lifecycle Stage," no view that traces an enrollment chain across four workflows, and no view that warns you a list one workflow populates is the enrollment trigger for another. The list is a directory, not a map. To reconstruct the relationships, you open each workflow and hold the connections in your head or in a spreadsheet, which is the manual process that does not scale.
This is not a criticism of HubSpot. The workflow tool is built to create and run automation, not to audit a portfolio of it across hundreds of workflows and dozens of client portals. The visibility gap is real, it widens as the portal grows, and it is precisely the gap a dependency-mapping tool fills. Knowing exactly where native capability ends is what lets you judge whether a tool is adding something real or just re-skinning what HubSpot already shows you.
If you are managing more than a handful of portals, you will eventually evaluate tooling. Native HubSpot is the baseline, and the question is what a dedicated tool adds on top of it. Use these criteria.
This is the first and most important question. A tool that maps direct enrollments but misses list-based and property-based connections gives you a map with holes in it, and the holes are exactly where the breakage happens. The property-based connection is the hardest to detect and the one most tools skip. Confirm it handles all three.
For an agency connecting to client portals, write access is a liability. A read-only tool cannot break anything, cannot be blamed for a change, and is far easier to get a client to approve. Confirm the tool connects via OAuth and confirm, explicitly, that it cannot write to the portal. Howly is read-only by design: it cannot write, edit, or change anything in a portal.
Speed matters because you will run the audit repeatedly: before every migration, handoff, and review. A tool that takes an hour to assemble a map you will regenerate weekly is a tool you will stop using. The benchmark to hold tools against is a full portal map in well under a minute. Howly loads in 10 to 25 seconds depending on portal size.
This is the agency-specific requirement that consumer-oriented tools ignore. You need to connect, switch between, and audit many client portals without separate logins, separate billing, or per-seat friction. Howly supports unlimited client portal connections under one agency account.
The map is for you, but the report is for the client. A tool that only shows you the map internally still leaves you building the deliverable by hand, so look for branded, white-label export that goes out under your agency's name. Howly produces a white-label branded PDF report for exactly this.
The single most valuable thing a tool can tell you is "here is everything that will be affected if you change this property." That is the question clients ask most and the one that causes the most damage when answered wrong. Howly's Impact Analyzer shows property-level blast radius before the change is made.
A quick way to weigh tools against each other:
| Criterion | Why it matters for agencies |
|---|---|
| All three connection types | Missed connections are where breakage hides |
| Read-only access | No write access means no liability on client portals |
| Full-portal map speed | You run the audit repeatedly, not once |
| Multi-portal under one account | You manage many clients, not one |
| Client-ready branded export | The documentation is the billable deliverable |
| Property blast-radius view | Answers the question that causes the most damage |
The audit is not a one-time event. There are four moments when running it is non-negotiable:
Before a migration. Migrating a HubSpot portal is one of the highest-risk operations a RevOps team can execute, not because HubSpot makes it hard, but because most portals contain years of accumulated automation logic that nobody has fully documented. You cannot migrate what you have not mapped.
Before a handoff. When the engagement ends and the client takes the portal back, the dependency map is what lets them make changes safely without you.
Before any property change on a mature portal. Check the blast radius first. Always.
On a recurring cadence. A quarterly health check catches staleness and orphaning before they accumulate into the 220-workflow mess that started this guide.
Most failed audits fail the same handful of ways. Knowing them in advance is the cheapest way to avoid them.
Auditing only the active workflows. Inactive workflows are where the surprises live. A paused workflow can be reactivated by anyone with access, and when it comes back on it brings whatever enrollment logic and property writes it had when it was switched off. An audit that ignores inactive workflows is an audit with a blind spot the size of every workflow someone "turned off for now."
Treating the workflow list as the dependency map. The list tells you what exists, not what connects. Reading down a list of 220 names and folders feels like understanding the portal, but it tells you nothing about the enrollment chains, shared lists, and shared properties that determine what breaks. The list is the starting inventory, not the finish line.
Ignoring property-based connections. Direct enrollment is the connection people remember to check because it names another workflow explicitly. Property-based connections are invisible by comparison, and they are the most common kind. An audit that maps enrollment chains but never asks "what else reads this property?" will miss the connections most likely to cause damage.
Auditing once and calling it done. A portal is not static. New workflows get built, properties get repurposed, and the map you produced in January is stale by April. The agencies that avoid recurring fire drills treat the audit as a cadence, not a project, and check every new workflow against the map before it ships.
Documenting for yourself instead of the client. A map you understand because you built it is not documentation. The deliverable has to make sense to someone who owns the portal but did not build any of it, which means plain-language summaries of what connects to what, not a screenshot of a canvas with no explanation.
When the engagement ends and the client takes the portal back, this is the minimum the documentation should contain. Each item answers a question the client will eventually ask.
A complete workflow inventory, active and inactive, with object type, status, trigger, action count, and last-modified date for every workflow.
The full dependency map, showing direct enrollment, list-based, and property-based connections, so the client can see what connects to what before they change anything.
A health summary listing the orphaned workflows, stale workflows, and broken enrollments you found, with a recommendation for each.
A property reference identifying the high-traffic properties, the ones read or written by many workflows, so the client knows which fields carry the most blast radius.
A plain-language overview of the portal's main automation flows, written for someone who did not build them.
A "change safely" note explaining how to check the impact of a change before making it, so the client does not break the portal the first week you are gone.
A tool that produces this as a branded export turns what would be a multi-day documentation effort into a generated deliverable. Howly's white-label PDF report covers the inventory, the dependency map, and the health findings in one document under your agency's name.
Work in five stages. Build a complete inventory of every workflow, active and inactive, capturing object type, trigger, status, action count, and last-modified date. Map the dependencies between them across all three connection types: direct enrollment, list-based, and property-based. Run a health audit to flag orphaned and stale workflows and broken enrollments. Export the inventory, the dependency map, and the health findings as a single document the client can read and act on without you. Then, for anything you build before handing over, plan the new workflow against the existing map so you do not introduce a fresh dependency problem on your way out. For portals past roughly 50 workflows, a read-only mapping tool like Howly builds the dependency map automatically rather than requiring you to trace every connection by hand.
Not natively. The HubSpot workflow list shows names, statuses, and folders, but it does not show connections between workflows. To see how workflows enroll into each other, share lists, or depend on the same properties, you either open and cross-reference every workflow manually, or you use a dedicated mapping tool. Howly detects all three connection types automatically and renders them as a single visual canvas, read-only, in 10 to 25 seconds.
Use a read-only tool that connects via OAuth and cannot write to the portal. Read-only access means the tool physically cannot edit, disable, or delete anything, which removes the liability of touching a client's live automation and makes the connection far easier for a client to approve. Howly is read-only by design and is available on the HubSpot Marketplace as a featured app, so the connection goes through HubSpot's own OAuth approval.
An orphaned workflow is an active workflow with no upstream or downstream connections: nothing enrolls into it and it enrolls into nothing, which usually means it is either doing something forgotten or doing nothing at all. A stale workflow is an active workflow that has not been modified in six or more months. Neither is automatically a problem, but both are prime cleanup candidates, since an orphan needs a decision about whether it still serves a purpose and a stale workflow nobody remembers building is a common source of unexpected enrollments.
Trace every workflow that references the property, in enrollment triggers, branch conditions, and set-property actions, before you make the change. This is the property's blast radius. Doing it manually means opening every workflow and checking each one for the property, which is slow and error-prone on a large portal. Howly's Impact Analyzer shows the property-level blast radius in one view: every workflow affected by a property change, and how, before the change is made.
Manually, mapping a portal of 200+ workflows across all three connection types takes days, because the dependency-tracing and cross-referencing has to be done by reading each workflow individually. With a dedicated mapping tool, it is seconds. Howly loads a full portal workflow map in 10 to 25 seconds depending on portal size, then keeps it current so the audit can be re-run before every migration, handoff, or property change.
Plan it against the existing dependency map first. Sketch the workflow you intend to create, configure its trigger and actions, and look at how it would connect to the workflows, lists, and properties already in the portal before any of it goes live. Howly's planning workflows feature does this directly: you add a planned workflow to the canvas, configure its triggers and actions using your real portal data, and it draws every connection the workflow would create before you build it in HubSpot. You see the downstream connections and catch conflicts at the canvas rather than discovering them in production.
Because every new workflow adds connections, and on a portal with hundreds of workflows you cannot hold the full connection graph in your head while you build. A new workflow that enrolls off a property something else sets, or writes to a list something else reads, creates a dependency you will not notice until a record behaves unexpectedly. The way to prevent it is to plan new workflows against the existing map so the new connections are visible before launch, rather than auditing the damage afterward.
It can, if there is no audit discipline behind it. HubSpot's Breeze AI lets users build workflows from natural-language prompts, which dramatically lowers the effort of creating automation and therefore raises the rate at which workflows accumulate. The workflows it produces connect to the same lists and properties as hand-built ones, but they tend to arrive without consistent naming, documentation, or awareness of what already exists, which is the definition of workflow debt. AI accelerates creation without improving visibility, so the faster a portal generates workflows, the more important it becomes to map dependencies, score health, and plan new workflows against the existing map. The audit is what keeps AI-accelerated creation from becoming AI-accelerated sprawl.
A properly designed mapping tool is read-only and changes nothing. Read-only means the tool can read your workflows, lists, and properties to build the map, but it physically cannot edit, disable, or delete anything in the portal. This is what makes it safe to connect to a client's live environment. Howly is read-only by design, connects via OAuth, and is available on the HubSpot Marketplace as a featured app.
Auditing a large HubSpot workflow setup comes down to four things.
First, the connections are the audit, not the list. A flat inventory of 220 workflows tells you what exists, but it tells you nothing about what breaks when you change something. The dependency map across direct enrollment, list-based, and property-based connections is the part that matters, and it is the part the native interface hides.
Second, manual mapping does not scale past about 50 workflows. The cross-referencing required to trace every connection by hand is slow and error-prone, and a missed connection is exactly what breaks a client portal after you have signed off.
Third, for agencies, the deliverable is the documentation. A clean, client-ready map and health report, produced read-only, under your name, is what justifies the engagement and what lets the client take the portal back safely.
Fourth, the audit is a discipline, not a one-time cleanup. Planning new workflows against the existing map keeps the portal from sliding back into the mess you just fixed, because every addition is checked against the connection graph before it ships.
A dedicated, read-only mapping tool collapses the slowest stage of the audit from days to seconds, and lets you plan new workflows against the map instead of building blind. Connect the portal to Howly and see the full dependency map before your next migration, handoff, or property change. The free trial is 7 days, no credit card required.
Howly is a read-only HubSpot workflow mapping and audit tool. It maps workflow connections, flags structural issues, and shows the impact of property changes before you make them. Used by RevOps teams and HubSpot agencies managing complex portals at scale.
