Workflow Management Software: Run the Leaver Test First

Workflow management software is the system that holds your team's repeatable processes: intake, routing, approvals, handoffs and the automations that move work between them. Choose it on four control facts, not feature lists: who an automation runs as, what happens when that person leaves, what the platform records, and what control costs.

Every serious platform in this category can draw a board and fire a rule. The thing they differ on, and the thing almost nobody compares, is what your process layer does the day the person who built it is deprovisioned. Our position is that this single question separates the platforms better than any feature grid, because it is the one place where the vendor's identity model becomes visible.

It is the question a Smartsheet customer asked in September 2024, in the vendor's own community, with an account already queued for deactivation: "the person who created those workflows has left the company." The accepted answer opened with three words no operations lead wants to read under a workflow they depend on: "I'm pretty sure".

The short answer: Pick workflow management software by running the Leaver Test. Pick your most business-critical automation, name the human account it runs as, and confirm in the vendor's own documentation what happens to it when that account is deactivated. Platforms differ enormously here, the answer is rarely in the feature comparison, and getting it wrong means silent process failure rather than a visible error.

Last updated: July 30, 2026.

The Leaver Test in one picture: one automation, one human account, three very different vendor behaviours.

A note on method. We did not run a controlled side-by-side trial of these platforms in a paid tenant, and will not pretend otherwise. Everything in the platform sections below is sourced from the vendors' own current documentation and pricing pages, fetched and quoted on July 30, 2026, with the links inline so you can check each claim against the same page we read.

What Workflow Management Software Actually Does

Workflow management software turns an implicit process into an explicit one. It stores the steps, the order, the owner of each step, and the conditions that move an item forward, then fires automations when those conditions are met. Boards, timelines and dashboards show you the state; rules and integrations change it.

The distinction that matters for buying is between the two halves. The display half is commoditised. Every platform reviewed here gives you a table view, Kanban boards, a timeline, custom fields, and a rule builder that reads when this, then that. Task tracking, team collaboration and visual dashboards are table stakes across the whole category in 2026. If your shortlist is built on those, every option passes and you have not narrowed anything.

The state-changing half is where platforms genuinely differ, in ways that never appear in a features table. An automation that moves an item between groups is trivial. One that posts to a customer, writes to a finance sheet, files a ticket in another system or grants access to a folder is software acting inside your company with somebody's permissions attached. Workflow management tools have quietly become one of the largest populations of unattended software in the average company, and they are administered by operations teams rather than IT.

What it is not

Workflow management software is not business process management (BPM) in the heavyweight sense. BPM suites model processes formally, version them and treat compliance controls as a first-class concern. Work management platforms come at the same territory from the collaboration side, with a rule builder bolted to a task database. That origin explains most of what follows: the identity model was designed for humans clicking, and automations were fitted into it afterwards.

It is also not integration middleware. Native automations run inside the platform's own data model, so cross-system orchestration still needs an integration platform or the vendor's connector marketplace, and those connectors carry their own credentials and consent scopes.

The three artefacts every platform creates

Whatever you buy, the platform will accumulate three things that outlive the person who created them:

The Leaver Test: A 20-Minute Diagnostic

The Leaver Test is a four-question diagnostic you can run in one sitting, on a platform you already own or one you are evaluating in trial. It does not need a security team, a spreadsheet, or a vendor call. It needs one automation that matters and twenty minutes.

Pick the automation your business would notice within a day if it stopped. The invoice routing rule. The customer-response SLA alert. The nightly status roll-up that the ops meeting depends on. Then answer these four in order, and write down where you found each answer.

Question 1 — Whose account does it run as? Open the automation and find the identity attached to it. Some platforms show this plainly; some require you to hunt. If you cannot determine the answer from the interface within five minutes, that is your answer for governance purposes: nobody in your company knows either.

Question 2 — What happens when that account is deactivated? This one must be answered from the vendor's documentation, not from a guess. You are looking for one of three behaviours: the automation stops, the automation transfers to someone else, or the automation continues under a non-human identity.

Question 3 — Who finds out? If the behaviour is stops, does anyone get told? A silent stop is materially worse than a loud failure, because the process appears to be running until someone downstream notices the absence of something.

Question 4 — What is written down afterwards? If a rule ran, or was disabled, or was edited, does the platform's audit log contain that event, and for how long is it retained? A rule that fired and moved money leaves a trail in exactly one place, and if that place holds thirty days you have thirty days of provable history.

Scoring it

Score each answer green, amber or red:

Two reds on your most critical automation is a procurement finding, not a nitpick. It means the process you depend on is attached to an employment relationship, and your platform will not tell you when that relationship ends.

If your current tool fails the test

Failing is common and it does not mean migrating. Do three things in order. Nominate a durable owner for every automation that matters, using whatever mechanism the platform offers: a bot actor if it has one, a default-owner setting if it has that, a named administrator if neither. Add "list and reassign automations, integrations and shares" to the offboarding checklist as an explicit step with a named person responsible, because no platform here will prompt you. Then export the audit log somewhere you control, since the vendor's retention window is not yours to set.

On the shared-account objection this raises: a nominated fallback owner should be a named service or administrator account with a documented human accountable for it, recorded outside the tool. It should not be a shared login several people use, which trades one governance failure for a worse one.

The test is deliberately cheap, because expensive evaluations do not get run. It also has a useful property: the answers are almost always in public documentation, so you can run it on platforms you have not bought yet. The rest of this article is that test, already run, on six of the most widely deployed options.

Why Automations Inherit A Person, Not A Role

Here is the mechanism. Work management platforms grew out of collaboration tools, where every action had a human behind it. Permissions were modelled per user, on boards, sheets and pages. When automation arrived, the cheapest correct implementation was to have the rule act with the permissions of whoever created it — that way the rule can never do more than its author could, which is a genuinely sound security default.

Notion states the consequence plainly in its own documentation: if you want an automation to add or edit pages and databases, you need permission to edit those pages and databases, and automations will not take action on any pages whose access is restricted. The automation is bounded by its creator's reach. That is good design, and it has an unavoidable corollary. The automation is also bounded by its creator's continued existence.

monday.com documents the failure mode with unusual clarity. If the automation's creator is unsubscribed from a board with permissions enabled, has permissions revoked from a column the rule touches, or is downgraded to a viewer, the rule is deactivated and the error reads: "The user who created this automation doesn't have permissions to make changes on this board." A routine permission tidy-up in one board can therefore stop an automation that a different team depends on.

Atlassian is the interesting counterexample, because it built the abstraction the others mostly lack. In Jira, the default flow actor is a "user" called Automation for Jira, a non-human identity that does not resign. Admins can set anyone as the actor if they want the change attributed to a real person. In Confluence automation, by contrast, the actor is the person who created the flow, and per the same page you can set the actor as yourself but cannot change it to someone else. One vendor, two products, two identity models.

The pattern this produces at scale

Multiply one automation by three years of operations work and you get a specific shape of risk that has nothing to do with hacking:

• A population of unattended jobs, each attached to an individual employee.
• No inventory of that population outside the platform, because the platform is administered by an operations team and not by IT.
• A deprovisioning process built for humans (disable SSO, revoke the laptop, close the mailbox) that does not know these jobs exist.
• Integrations installed per-user, each holding a token to a second system.

Public breach data is a reasonable proxy for how much this category of standing access matters. In the 2026 Data Breach Investigations Report, published by Verizon on May 19, 2026, breaches involving a third party accounted for 48% of all breaches, and 31% of breaches began with vulnerability exploitation. Third-party involvement is not the same thing as an orphaned workflow automation, and we are not going to stretch the statistic into one. What it does establish is that the connections between your systems, rather than the systems themselves, now carry a large share of real-world breach exposure, and a workflow platform's connector list is exactly that kind of connection.

The standards vocabulary here is older than the problem. NIST's SP 800-53 Rev. 5 account management control (AC-2) lists the account types an organisation is expected to manage, and the list explicitly includes shared, group, guest, temporary and service accounts alongside individual ones. The same control expects a list of active system accounts with the name of the individual associated with each. Applied to workflow management software, that is a straightforward instruction: every automation should have a named owner recorded somewhere outside the tool, and that record should be checked when people leave.

We have argued a related case for AI agents before, in our earlier analysis of non-human identity and why every agent needs an owner, a scope and an expiry. The workflow automation is the same object, ten years earlier and with a friendlier user interface.

For a wider view of how movers and leavers create orphaned access across a whole SaaS estate, this short vendor explainer covers the lifecycle problem directly:

Platform By Platform: What Happens When The Builder Leaves

Six platforms, run through the Leaver Test using their own current documentation. Prices are list prices in USD, fetched July 30, 2026, and every platform prices per seat with the governance features concentrated at the top tier. That pattern is the subject of a later section.

At a glance

Three of the seven rows above say "not documented". That is a finding in itself, and it is the honest state of the evidence rather than a gap in our reading. Each cell reflects the vendor's own current help content, linked in the sections below.

monday.com

Best for: operations teams that want the failure mode written down, and admins who will actually use the default-owner setting.

monday.com is the only platform in this set that documents the leaver scenario end to end. Its help centre lists "Deactivated User" as a named cause of automation failure, gives the exact error text, "The automation owner was deactivated from your account", and states plainly that you will not receive a notification that the automation was deactivated. It also documents the awkward remedy: to transfer ownership of a deactivated user's automations, the user has to be reactivated in the account first.

Crucially, it also documents the fix. There is an account setting to keep automations running when users are deactivated, by nominating a default owner. That single toggle converts a red Leaver Test result into an amber one, and most buyers do not know it exists.

Control facts: the audit log is Enterprise-only, admin-only, and the documentation states that audit log records will not be deleted after a certain amount of time. The documented event set is security-centric (logins, failed logins, attachment downloads, board exports) and the page does not describe automation configuration events.

Pricing: per the pricing page, Work Management runs Basic $9, Standard $12 and Pro $19 per seat per month billed annually, with Enterprise on custom pricing. Automations and integrations are only available from Standard upward, and the allowance is metered in actions: 250 per month on Standard, 25,000 on Pro and 250,000 on Enterprise. monday.com's own note on that page is more useful than any third-party review: 250 actions a month is "typically exhausted within days" for accounts that notify several people or map several columns per automation.

Pros: documented leaver behaviour; a default-owner setting that actually solves it; indefinite audit retention. Cons: SSO, audit log and advanced permissions all sit behind Enterprise; the 250-actions-per-month Standard cap is low enough that a single busy board can exhaust it. Bottom line: the most transparent vendor in this comparison on the exact question that matters, and the only one whose documentation lets you pass the Leaver Test before you buy.

Asana

Best for: companies that want app and token control early, without waiting for an Enterprise contract.

Asana's rule ownership behaviour is documented, though we found it stated most plainly in Asana's community forum, where the deprovisioning section of the company's own guide on rules is quoted and restated by an Asana moderator. When a user is deprovisioned, their rules are paused and ownership of rules on projects passes to the project owner; where there is no project owner, ownership passes to the admin who performed the deprovisioning, and only the new owner can reactivate the rule. That is a better default than a hard stop, but it still requires a human to notice and act.

The practitioner experience of this is worth reading. In a long-running Asana forum thread on offboarding, the consultant Larry Berger describes a client whose IT manager was leaving after six years of heavy use, owning large numbers of custom fields and projects, and whose support requests for a simple list of what she owned came back empty. His summary: "My client strongly encourages Asana to address this gap in the product making offboarding painful." The thread has been maintained with updates through January 2026, which tells you how durable the problem is.

Control facts: the Audit Log API is restricted to service accounts on Enterprise+, legacy Enterprise, or an Enterprise domain with the compliance management add-on. Its event schema does include OAuth app names and rule names, so automation and integration activity is at least representable in the log, which is better than several competitors. Notably, app management controls that govern which third-party apps, personal access tokens and service accounts can connect are available from the Starter tier, per the pricing page, rather than being an Enterprise-only privilege.

Pricing: Starter $10.99 and Advanced $24.99 per user per month billed annually ($13.49 and $30.49 billed monthly), with Enterprise and Enterprise+ on contact-sales pricing. SAML SSO and SCIM provisioning begin at Enterprise.

Pros: automatic ownership transfer rather than a dead stop; app and token controls low in the range; audit schema covers apps and rules. Cons: the transferred rule stays paused until someone acts; audit log access is API-only and gated at the very top of the range. Bottom line: the best mid-market answer on connected-app control, with a leaver behaviour that degrades gracefully instead of failing silently.

Jira and Confluence (Atlassian)

Best for: organisations that want automations to run as software rather than as staff.

Atlassian is the only vendor here that gives you a first-class non-human identity for automations by default. In Jira, changes performed by a flow are attributed to the flow actor, and the default actor is the "Automation for Jira" user; app administrators can set anyone as the actor when human attribution is preferable. If you leave the default in place, deactivating an employee has no effect on the rules they wrote. The Leaver Test comes out green on question two without any additional configuration.

Confluence automation does not inherit this. The actor there is the person who created the flow, and the same page states you can set the flow actor as yourself but cannot change it to someone else. Buying the Atlassian suite therefore gets you two different answers depending on which product the automation lives in.

Control facts: the audit story has a sharp edge. Atlassian's documentation states that audit log activities are stored for up to 180 days and anything older is removed and cannot be recovered, and, more importantly, that organisation-level insights require an Atlassian Guard subscription or at least one Enterprise plan, without which "organization-level logs will not store or display any events". Not degraded. Not shorter. Not stored at all.

There is also a documented operational cliff on automation volume. Each plan tier carries a higher monthly automation limit, with Enterprise unlimited, and when you hit the limit Atlassian states that all rules for that product stop until the limit resets at the start of the next calendar month, with unused runs not rolling over. A busy month can therefore switch off your entire automation layer until the first of the following month.

Pros: a real bot actor as the default; the only platform here that solves the ownership problem structurally. Cons: no organisation audit events at all without Guard or Enterprise; a hard monthly automation ceiling that stops every rule; Confluence does not share Jira's actor model. Bottom line: structurally the strongest identity model in the category, paired with the strictest commercial gate on evidence.

ClickUp

Best for: teams that need cheap breadth and are prepared to compensate for a short audit window.

ClickUp publishes the most detailed event catalogue here. The workspace audit log includes User, Task, Fields, Hierarchy, Agents and Other logs, and the user log explicitly records "App connected" and "App disconnected" events, including external MCP server connections, alongside SCIM provisioning and deprovisioning events, SSO configuration changes and custom role changes. There is a dedicated Agents log tracking Super Agent activity by user, agent and activity type. For a 2026 buyer, that MCP line matters. It is the only mainstream work platform in this set that names the protocol in its audit events, and we have written separately on why MCP connectors need governing rather than trusting.

The catch is retention and reach. Audit logs are Enterprise-only, visible to Owners by default, and historical data is available for 30 days. Owners can grant admins access via a custom role permission. Thirty days is shorter than most internal incident review cycles and shorter than several regulatory minimums.

We could not find, on either audit page, an event type for automation rule creation or modification, and neither page documents what happens to an automation when its creator is deprovisioned.

Pricing: per the pricing page, Unlimited $7 and Business $12 per user per month billed annually, with Enterprise on custom pricing. Business includes 5,000 automations per month; Enterprise includes 250,000, plus SAML SSO, audit logs and unlimited custom roles.

Pros: the most granular published event catalogue, including app connections and MCP servers; low list price at the working tiers. Cons: 30-day audit retention is the shortest in this comparison; automation ownership behaviour on deprovisioning is undocumented; audit access defaults to Owners only. Bottom line: excellent visibility while the window lasts, and the window is short.

Notion

Best for: documentation-led teams whose automations are mostly internal state changes.

Notion's model is permission-bounded rather than ownership-bounded. Database automations can be created, edited or deleted by paid-plan users with full access to the database, and the platform will not act on pages whose access is restricted. Slack automations are the exception and can only be edited by the automation creator. Because editing rights are attached to database access rather than to the original author, a colleague with full access can usually pick up an orphaned automation without an admin transfer, which is the practical outcome most teams want.

Control facts: the audit log is available to organisation owners on the Enterprise plan, retains history for up to 365 days, and, unusually, captures integration lifecycle events including integration created, integration deleted and integration installation toggled, plus guest removals and guest-capability changes. That is the most integration-aware audit log in this comparison, and 365 days is the longest retention among the vendors that publish a number.

Pricing: per the pricing page, Plus $10 and Business $20 per seat per month, with Enterprise on custom pricing. SAML SSO starts at Business; SCIM and the audit log are Enterprise.

Pros: longest published audit retention; integration and guest events logged; automations are portable between colleagues with database access. Cons: no documented behaviour for automations when a creator is deprovisioned; the audit log itself is Enterprise-only. Bottom line: the strongest audit content in the set, sitting behind the same Enterprise gate as everyone else.

Smartsheet

Best for: structured, sheet-centric operations where automation volume is high and the Business tier's unlimited automations are the draw.

Smartsheet is where the evidence runs out. We could not find help content stating what happens to an automated workflow when its creator is deactivated; the workflow management article covers who can manage workflows by permission level but not ownership succession. The community thread quoted at the top of this article is the clearest public answer, and its accepted response is explicitly a guess about automations running under the sheet owner's context.

Control facts: Event Reporting is the audit mechanism. It requires the Enterprise plan, covers events going back up to six months, and is accessed through the Smartsheet Events API rather than an admin screen. That last detail is a real operational cost. Retrieving your own audit trail is a development task, not a filter and export.

Pricing: per the pricing page, Pro is $9 and Business $19 per member per month billed yearly, with Enterprise on custom pricing. Pro includes 250 automations per month; Business includes unlimited automations and unlimited free guests. SAML-based SSO and directory integrations begin at Enterprise.

Pros: unlimited automations from the $19 tier, which is the best automation-per-dollar position in this set; a six-month audit window. Cons: audit is API-only and Enterprise-gated; ownership succession for workflows is not documented publicly. Bottom line: strong economics on automation volume, weakest published answer on the leaver question.

Airtable, briefly

Airtable belongs on many shortlists here, and we checked it. Its automation management documentation sets out the permission split. Creator-level roles can create, delete, duplicate, configure and rename automations, while Editors can view an automation's configuration or copy its URL. But the page we read does not state whose permissions an automation executes with, or what happens when the automation's owner is removed. We are not going to fill that gap from third-party summaries, so Airtable sits outside the main comparison with an unresolved Leaver Test.

The Audit Question: What Gets Recorded, And For How Long

Retention is where the workflow category collides with regulation, and almost no buying process checks it. The pattern across the six platforms is consistent: the audit log is an Enterprise feature, its retention is set by the vendor rather than by you, and the published windows vary by more than an order of magnitude.

Now put a legal number next to that column. Article 26(6) of the EU AI Act requires deployers of high-risk AI systems to keep the logs automatically generated by those systems, where the logs are under their control, for at least six months unless other Union or national law says otherwise. That obligation attaches to high-risk AI systems, not to workflow tools generally, so it does not apply to your invoice-routing rule as such. It matters because of where the category is heading: the same platforms are shipping AI agents into the same boards. ClickUp already logs Super Agent activity. Once an agent inside a workflow tool touches a high-risk use such as recruitment screening, worker evaluation or access to essential services, a 30-day audit window is below the floor the regulation sets, and the retention decision was made by your vendor's plan tier rather than by your compliance team.

Three practical consequences follow.

Export is not optional. If your platform's retention is shorter than your incident review cycle or your regulatory window, the audit log has to be streamed or exported into something you control. Atlassian says this explicitly, recommending regular export for longer record-keeping. Treat the vendor's log as a live buffer, not as your archive.

API-only audit has a staffing cost. Smartsheet's Event Reporting and Asana's Audit Log API both require someone to write and maintain a client. That is a real line item, and it is invisible in a per-seat price comparison. Budget for it or accept that the log will be consulted only after an incident, by whoever can be found to write the script that day.

Check what is actually logged, not just that logging exists. monday.com's audit log is a security log of logins, exports and downloads, and its documentation does not describe automation configuration events. ClickUp's catalogue covers app connections and SCIM lifecycle events but we found no automation rule events. Notion logs integration creation and deletion. Asana's schema carries rule names and OAuth app names. "Has an audit log" is not a yes or no answer; it is a question about which of your three risky artefacts leave a trace.

We have written at more length on what a defensible action trail looks like when software rather than people is doing the acting, in our analysis of AI observability and audit trails. The short version applies here without modification: a log that records what ran but not what was refused, changed or disabled will not answer the question you eventually need to ask it.

The Real Cost Of Control: Seat Price Is Not The Price

The published per-seat number tells you very little about what a governed deployment costs. Three separate mechanisms move the real figure, and all three are verifiable before you sign.

Mechanism one: the governance cliff

Across every platform in this comparison, the controls that make the tool safe to run at company scale (SAML SSO, SCIM provisioning, audit logs, custom roles, advanced permissions) arrive together at a single tier, and that tier is almost always contact-sales.

Notion is the only platform here that publishes a price for a tier including SAML SSO. For every other option, the moment you require single sign-on you leave published pricing entirely, and the delta between the highest published tier and the negotiated Enterprise price is the actual cost of governing the tool. Ask for that number in the first vendor call, not the third.

Mechanism two: automation limits as a forced upgrade

Automation caps are the least-examined line in this category and the most likely to surprise you mid-year.

Two things follow. First, 250 actions a month is roughly eight a day across an entire account, and monday.com itself says that allowance is typically exhausted within days once automations notify several people or touch several columns. Treat the 250-action tiers at monday.com and Smartsheet as evaluation tiers, not operating tiers. Second, Atlassian's behaviour at the ceiling is the one to plan for: hitting the monthly limit stops every rule for that product until the calendar month turns over. If your automation layer is load-bearing, a usage cap is an availability risk rather than a billing detail.

Smartsheet's Business tier at $19 with unlimited automations is, on this axis alone, the strongest published position in the comparison.

Mechanism three: the seat you did not count

Guests, contributors and viewers are priced differently on every platform, and the differences are large. Smartsheet includes unlimited free contributors on Pro and unlimited free guests from Business, while capping Pro at ten members and requiring at least ten members for Enterprise. A per-seat comparison between two vendors stays close to meaningless until you model your actual roster, including externals and read-only stakeholders, against each vendor's seat classes and minimums.

A worked total

Take a 40-person operations team, moderately automation-heavy, needing SSO because IT requires it for anything holding customer data.

The arithmetic is simple, and that is the point: at the tier where these tools become buyable for a governed team, published pricing stops being available for almost every vendor, and the comparison you thought you were making evaporates. The correct response is not to give up on comparison. It is to make each vendor put four numbers in writing during evaluation, before you shortlist:

1. The Enterprise per-seat price at your exact seat count, including any minimum.
2. The monthly automation allowance at that price, and the behaviour when it is exceeded.
3. The audit log retention at that price, and whether streaming or export to your own storage is included or extra.
4. The price of every guest, viewer or contributor seat class you will actually use.

Four answers make two vendors comparable. A feature checklist does not.

Choose This If: A Decision Tree That Commits

The Leaver Test does not produce a single winner, because different organisations fail differently. Here is where each option is the defensible choice.

Choose Jira if the automations must survive personnel change without anybody remembering to intervene. The default bot actor is the only structural answer to the ownership problem in this comparison. Accept in exchange that you must buy Atlassian Guard or an Enterprise plan to have any organisation-level audit events at all, and that Confluence automations do not inherit the same model.

Choose monday.com if you want the failure mode documented and a switch that fixes it. It is the only vendor that publishes the leaver behaviour, the exact error text, and a default-owner setting. Set that default owner on day one, before you have twenty automations rather than after.

Choose Asana if third-party app and token control matters more than audit depth. App management controls sit from the Starter tier rather than at Enterprise, rule ownership transfers automatically on deprovisioning, and the audit schema names both OAuth apps and rules. Accept API-only audit at the top of the range.

Choose Notion if your evaluation is weighted toward evidence retention and integration visibility. 365 days is the longest published window here, integration create, delete and toggle events are logged, and SAML SSO is available at a published $20 seat price rather than behind a sales call.

Choose Smartsheet if automation volume is your binding constraint and you have engineering capacity. Unlimited automations from $19 per member is the best automation economics in the set. Budget explicitly for the Events API client, and run the Leaver Test with the vendor directly, because the public documentation will not answer it.

Choose ClickUp if you need broad functionality at the lowest published price and can live with a 30-day window. The event catalogue is the most detailed published in this category, including MCP server connections. Pair it with log export from the first week, because 30 days will not cover a quarterly review.

Choose none of them yet if you cannot answer question one of the Leaver Test on your current tool. Migrating before you understand your existing ownership map moves the problem rather than solving it, and you will rebuild the same orphaned automations in a new interface.

When The Feature Comparison Still Wins

This article has argued for a control-first evaluation, and it would be dishonest to pretend that is always the right frame. Three situations where the ordinary feature comparison is the better guide.

When nothing the automations touch leaves the tool. If your rules only move items between statuses, assign owners and send in-app notifications, the blast radius of an orphaned automation is a stalled board that somebody will notice in a day. Usability, view types and mobile parity genuinely matter more than actor models in that world, and the competing roundups that rank on those criteria are giving you the right advice.

When the team is small enough to be its own inventory. In a company of eight, everyone knows who built the rule. Governance controls are a real cost: Enterprise tiers, SSO licences, admin time. Buying them for a team that can reconstruct its whole automation map in a standup is over-engineering.

When adoption is the actual risk. The most common workflow platform failure is not an orphaned automation; it is a beautifully governed platform nobody uses, with the real process still running in a spreadsheet and a group chat. If your last two rollouts stalled, weight ease of adoption heavily and revisit control once the tool has stuck. A governed system of record only helps if it is the system of record.

There is also a fair objection to the Leaver Test itself: it optimises for a low-frequency event. Most companies deprovision a workflow builder a handful of times a year. The reason we still put it first is asymmetry. The test is cheap and repeatable, and the failure it detects is silent. Most procurement criteria fail loudly, in a demo, in front of the buyer. This one fails quietly, three months after go-live, in a process the buyer has stopped watching.

Where AI Agents Change The Question

Everything above concerns rules, meaning deterministic logic that a person wrote and a platform executes. The category is now shipping something different into the same boards. ClickUp's audit log has an Agents section for Super Agent activity. monday.com ships AI agents. Every major vendor in this comparison has an AI roadmap that puts non-deterministic software inside the workflow, with the same identity model underneath.

That changes the four questions in a specific way. With a rule, "who does it run as" determines what it can reach. With an agent, it also determines what it can reach after being persuaded to do something its author never wrote down. The ownership question stops being an offboarding hygiene issue and becomes the primary control.

This is the layer our own platform is built for. LeapForce is a governance layer that sits in front of the AI tools, connectors, models and agents a company uses, so that access, policy, cost and audit are enforced in one place rather than configured separately in each SaaS product. Our Workflows product chains agents and connectors with explicit human approval gates, and our Access and Identity layer treats non-human identities as first class, with an owner, a scope and an expiry attached to every agent. That is precisely the record the workflow platforms in this comparison do not keep. Our rollout model for the gateway is deliberately incremental: observe first, enforce second, optimize third. Point traffic at the gateway in observe mode, learn what is actually running, and only then write rules. Per our published build-status convention, gateway endpoints, tracing and SSO are live today, while capabilities such as shadow-AI discovery and compliance evidence packs are still in development.

To be clear about what this does and does not do: LeapForce is not a workflow management platform and will not replace monday.com, Asana or Jira for your process boards. If you are choosing a tool to hold your team's projects, choose one of the platforms above.

Honest Limits: What We Could Not Verify

Four limits on this analysis, stated plainly.

We did not run these platforms. There is no paid tenant behind this article and no controlled deactivation test. Every platform claim comes from vendor documentation and pricing pages fetched on July 30, 2026, quoted with links. Documentation and behaviour can diverge, and the correct next step for a serious evaluation is to reproduce the Leaver Test in a trial workspace with a throwaway account. If you do, the result will be better evidence than anything on this page.

Three platforms have undocumented leaver behaviour and we left it undocumented. ClickUp, Notion and Smartsheet do not publish what happens to an automation when its creator is deprovisioned, on the pages we checked. Third-party guides state answers for some of them. We are not repeating claims we could not trace to a primary source, which is why those cells read "not documented" rather than carrying a number.

Enterprise pricing is unpublished almost everywhere, so the cost section is incomplete by construction. We can tell you which tier a control lives in and what the highest published tier costs. We cannot tell you the Enterprise delta, because no vendor in this set publishes it. Any total-cost model that claims to is estimating.

We cannot quantify the downside, so we have not tried. A fair challenge to this whole argument is "what does an orphaned automation actually cost?" There is no credible public dataset on losses attributable to workflow automations failing after an owner's departure, and inventing a figure would undermine everything else here. What we can say is that the failure is silent by the vendors' own documentation, and that silent failures are found late by definition.

Documentation goes stale fast. monday.com's audit log article was last modified in May 2026; Atlassian's automation limits changed materially in 2023; ClickUp's event catalogue already names MCP servers, which did not exist as a category three years ago. Re-check any figure here against the linked page before you put it in a business case.

One further limit on scope: this article is about the identity and audit behaviour of general-purpose workflow management software. It is not a review of how pleasant these tools are to use, and it is not an AI-first comparison. For that side of the decision, see our earlier piece on choosing AI workflow tools for the failure mode rather than the demo, which asks what each step is permitted to do when the model is wrong.

Ready to Govern Your AI?

Talk to LeapForce — one controlled layer for every AI tool, connector, model, and agent.