Slow workflows are easy to misread.
The team says work is stuck. Customers are waiting. Internal deadlines slide. Managers chase updates. Someone opens the project board and sees too many items sitting in "pending", "review", "waiting", or "blocked".
So the business starts looking for a productivity fix.
Do we need better project management software? Should we automate reminders? Can AI summarise the backlog? Do we need more people? Are individuals not moving fast enough? Should we redesign the process?
Sometimes those questions are useful.
But there is a more basic question worth asking first: where is the next decision actually stuck?
In many businesses, the visible workflow problem is not mainly a task execution problem. It is decision latency. Work slows down because the people doing the work do not have the authority, context, confidence, or agreed rules to move it forward without waiting for someone else.
That waiting often looks like process friction.
Underneath, it is usually decision design.
The Work Is Waiting For Permission
Most teams are reasonably good at doing the work they understand and control.
The slowdown tends to appear when a choice is needed.
Can we approve this quote? Which customer exception should we honour? Is this lead worth pursuing? Can this delivery date move? Who signs off the budget? Does this request fit the policy? Should this bug delay the release? Can support offer a goodwill credit? Do we need legal input? Is this a product decision, an operations decision, or a sales decision?
If the answer is obvious, work moves.
If the answer is not obvious, work waits.
That wait might be a Slack message, an email, a meeting agenda item, a comment in a project tool, a task assigned to a director, or an informal "can you just check this?" conversation. One pause does not look expensive. Fifty repeated pauses across the business are very expensive.
The hidden cost is not only the time spent waiting.
It is the context switching, chasing, rework, customer delay, staff frustration, and management attention that accumulates around the wait.
Faster Tools Do Not Fix Unclear Authority
This is where businesses often reach for tooling too early.
A better workflow system can show you where decisions are stuck. It cannot decide who should be allowed to make them.
Automated reminders can prompt approvers. They cannot tell the business whether approval is genuinely needed.
AI can summarise the case for a decision. It cannot safely invent authority where the operating model has not defined it.
Dashboards can expose cycle time. They cannot remove a bottleneck that exists because one overloaded person is the only accepted decision-maker.
The tool may be doing exactly what it was asked to do. The problem is that the process has encoded a weak decision model.
You see this when every non-standard request routes to a founder. When discounting needs approval but nobody has agreed the acceptable range. When support cannot resolve small goodwill cases without escalation. When delivery teams need permission for routine trade-offs. When account managers promise things that operations later has to negotiate. When finance, sales, and delivery all have different definitions of "approved".
The workflow is not slow because the software is slow.
It is slow because the organisation has not made decision rights clear enough.
Decision Latency Has Patterns
Decision latency usually shows up in repeatable patterns.
The first is the missing owner. Everyone can describe the problem, but nobody clearly owns the decision. Work bounces between functions because the business has never decided where that class of decision belongs.
The second is the overloaded approver. One senior person becomes the queue for anything ambiguous. Their judgement may be good, but the business has turned that judgement into an operational dependency.
The third is the invisible rule. Experienced staff know what will probably be approved, but the rule is not written down or built into the system. Newer staff escalate because they do not want to get it wrong.
The fourth is the conflicting metric. Sales wants speed, finance wants margin, delivery wants feasibility, and customer success wants retention. Nobody is wrong. But if the priority is not explicit, every edge case becomes a negotiation.
The fifth is the missing context. The right person could make the decision quickly, but the information they need is scattered across CRM notes, spreadsheets, email threads, project tools, and personal memory.
None of these are solved by telling people to be more proactive.
They are solved by making the decision easier, clearer, and closer to the work.
The Fix Is Often Smaller Than Expected
The practical fix starts with mapping the waits.
Pick one workflow that feels slow and inspect the last ten examples. Look for the points where work stopped moving. Then ask what decision was needed, who made it, what information they needed, whether the decision was routine or genuinely exceptional, and whether the same pause keeps repeating.
That exercise usually separates three different problems.
Some decisions should be removed. They are approvals that no longer add control, checks that exist because of an old incident, or sign-offs that create delay without reducing risk.
Some decisions should be delegated. The business can define thresholds, rules, guardrails, or escalation criteria so the team closest to the work can move without waiting.
Some decisions should be supported. They still need judgement, but the approver needs better context, cleaner data, a prepared recommendation, or a faster way to compare options.
This is where technology can help, but only after the decision problem is understood.
A small internal tool might gather the relevant customer, margin, delivery, and risk context into one decision view. A CRM change might expose the approval threshold before a quote is sent. An AI assistant might summarise the case, flag missing information, and draft a recommendation. Automation might route only true exceptions to a senior person while routine cases move automatically.
The important point is that the automation should serve the decision model.
If you automate the existing ambiguity, you just create faster ambiguity.
Move Decisions Closer To The Work
Healthy operations do not remove every decision.
They put the right decisions in the right place.
Routine decisions should be handled by the people doing the work, with clear rules and visible guardrails. Higher-risk decisions should escalate with the right context already attached. Strategic decisions should not be disguised as individual task approvals. Exceptions should be tracked so the business can see whether the rule needs to change.
That is what good operational design looks like in practice.
It is not a prettier process map. It is a business where people know which choices they can make, which choices need escalation, what information matters, and what happens next.
At Intelligent Marmalade, an operational friction review looks for these delays directly. We trace where work waits, what decision is missing, who currently carries the burden, and whether the right fix is targeted AI automation, a small internal tool, better system architecture, or a cleaner process rule.
If a workflow keeps stalling, do not only ask whether the team needs a better tool.
Ask which decision is arriving too late.
That is usually where the useful fix starts.