Capacity or Coordination? How to Tell What Is Really Slowing a Process Down

Capacity or Coordination? How to Tell What Is Really Slowing a Process Down

Schedule a meeting
  1. Why doesn't adding headcount reduce process cycle time?
  2. What is the difference between touch time and elapsed time in a process?
  3. What are deciding, fetching, and checking work?
  4. How do you calculate coordination share and wait multiple?
  5. How does wait multiple relate to process cycle efficiency?
  6. What do coordination share results mean for agentic process redesign?
  7. Which processes are good candidates for Agentforce Operations?
  8. What can the Coordination Cost Diagnostic not tell you?
  9. How does OSF Digital deliver Agentforce Operations?
Find out where your process time goes. Download the Coordination Cost Diagnostic, walk one process, and send your results to OSF Digital for review.
When a process is slow, leaders usually look at two fixes: add people, or look closely at where the time goes. Which one is right depends on whether the process is capacity-constrained or coordination-constrained, and standard operations reporting rarely shows which. Two ratios answer that question for a single process in about an hour: coordination share and wait multiple. All it takes is a worksheet and someone who knows how the work runs.

Why doesn't adding headcount reduce process cycle time?

Adding headcount often leaves process cycle time unchanged when the constraint sits in the coordination between teams and systems. New hires there add requests rather than capacity. A new coordinator sends more emails to the same upstream teams, and those teams got no extra capacity. The queue moves out of the function where it shows on a dashboard and into functions where it doesn't.This pattern is common in processes that run between systems, such as supplier onboarding, invoice exceptions, claims, complaints, and account changes. Each system does its part. The work between the systems falls to people: chasing a document, keying data from one system into another, rebuilding a file for a reviewer. When that work makes up most of the effort, the process is coordination-constrained, and hiring addresses a different problem from the one slowing it down.When most of the hands-on work is genuine judgment, the process is capacity-constrained. Then the honest answer is more skilled people, better decision support, or clearer decision rules. Telling the two situations apart is the first decision, and it takes two measurements that most operations dashboards don't include.Find out where your process time goes. Download the Coordination Cost Diagnostic, walk one process, and send your results to OSF Digital for review.

What is the difference between touch time and elapsed time in a process?

Touch time is how long the work in a process step takes once someone is actually doing it. Elapsed time is how long the step takes from the moment it is ready to the moment it is finished, including every hour it sits in a queue. A risk review might take 40 minutes of reading and four days of waiting for the reviewer.Operational reporting usually merges the two. Average handle time counts a genuinely complex case and a case that sat three days in an inbox the same way. Both show up as elapsed days, and only one of them represents work. Separating touch time from elapsed time makes the waiting visible, and waiting is usually where cycle time accumulates.

What are deciding, fetching, and checking work?

Every hands-on step in a process falls into one of three types of work: deciding, fetching, or checking. The balance between them shows where a team's effort goes.
  • Deciding is judgment work: assessing facts, applying rules, weighing options, approving or rejecting, and writing down the reasoning behind a decision.
  • Fetching is getting something: requesting records from a system the team cannot access, chasing replies, pulling data from several systems into one place, assembling a file so someone else can follow it, and keying information from one system into another.
  • Checking is verifying or redoing work that has already been done once: second-reviewer passes that re-read rather than sample, reconstructing a file for an escalation, and re-running cases after a rule or policy change.
Two classification rules keep the result conservative. If a step mixes judgment and retrieval and they cannot be separated, it counts as deciding. Waiting always goes into elapsed time and never into touch time. Both rules push the coordination figure down, so whatever number the exercise produces is a floor rather than a ceiling.

How do you calculate coordination share and wait multiple?

Coordination share and wait multiple come from walking one process step by step as it is actually performed, usually 8 to 15 steps. For each step, record the role, the touch time, the elapsed time, and the activity type.
  • Coordination share = (fetching touch time + checking touch time) ÷ total touch time × 100
  • Wait multiple = total elapsed time ÷ total touch time
Here is a partial walk-through of a supplier onboarding process:
#StepRoleTouchElapsedType
1Request arrives by email from the business unitOps coordinator10 min10 minFetching
2Coordinator emails requester for missing tax documentationOps coordinator15 min3 daysFetching
3Supplier checked against sanctions list in a separate systemOps coordinator20 min20 minFetching
4Coordinator compiles submission pack for reviewOps coordinator45 min45 minFetching
5Risk reviewer reads pack and assesses supplierRisk analyst40 min4 daysDeciding
6Second reviewer re-reads same pack per two-person policyRisk analyst35 min2 daysChecking
7Coordinator keys approved supplier into the ERPOps coordinator25 min1 dayFetching
8Coordinator notifies requester and closes the recordOps coordinator10 min10 minFetching
The fragment holds 200 minutes of touch time. Deciding accounts for 40 of those minutes. Fetching (125 minutes) and checking (35 minutes) make up the other 160, so the coordination share is 80%. Elapsed time is just over 10 working days, or about 80 working hours, which gives a wait multiple of about 24. For every hour of work in this process, 23 hours pass in which nothing happens to it.Step 6 deserves a closer look. It is a legitimate policy control. It is also 35 minutes spent re-reading a pack that a qualified colleague read shortly before. The diagnostic is designed to make steps like this visible so they can be discussed. It does not argue for removing them.Across the operational processes OSF Digital works on, roughly 30% to 40% of a typical business process is already automated. The remaining 60% to 70% is human coordination around that automated core. A coordination share in that range or above it reflects a pattern that appears in most operational processes. What is unusual is having measured it.Find out where your process time goes. Download the Coordination Cost Diagnostic, walk one process, and send your results to OSF Digital for review.

How does wait multiple relate to process cycle efficiency?

Wait multiple is closely related to process cycle efficiency, the lean measure that divides value-added time by total lead time. Wait multiple uses all hands-on time as its base and inverts the ratio. A wait multiple of 24 is the same as a flow efficiency of about 4%. Teams that already use lean methods will recognize the logic.Coordination share adds a second cut. Lean analysis would class much of the fetching and checking as non-value-added. Coordination share separates that portion explicitly from deciding. The result shows how much hands-on time exists only because information has to be moved, gathered, assembled, or verified a second time. That is the portion of the work that agents can take on.

What do coordination share results mean for agentic process redesign?

A coordination share result places a process in one of three bands. Each band points to a different response.
Coordination shareBandWhat it suggests
60% or aboveCoordination-constrainedMost hands-on time is spent moving information. Adding people shifts the bottleneck elsewhere. Agents taking on the fetching and checking between systems is worth assessing.
40% to 59%MixedThe process is probably a composite. Split it into sub-processes, calculate each one separately, and start with the sub-process that has the highest share and the clearest measure.
Below 40%Capacity-constrainedMost of the work is genuine judgment. Improvement comes from decision quality, decision support, or the decision rules themselves. Putting agents into this process adds a layer without releasing much time.
Wait multiple sharpens the reading. Above 10 in the coordination-constrained band, the case is unusually clear: the process contains little work and a great deal of waiting, and the time to recover is calendar time. Above 10 in the mixed band, queueing and handoff structure are the more likely constraint. Above 10 in the capacity-constrained band, the process is waiting on the people who make the decisions, which is a scheduling and prioritization issue.For a coordination-constrained process, three checks come next:
  1. Find the three steps with the largest gap between touch time and elapsed time. Those are the queues.
  2. Separate checking that serves as a deliberate control from checking that exists because the first pass could not be trusted or followed. Only the second kind is removable.
  3. Confirm the coordination is repeatable. Agents can coordinate handoffs that happen consistently and in a consistent sequence. Genuinely varied handoffs are harder.

Which processes are good candidates for Agentforce Operations?

Agentforce Operations is a process layer that sits above existing systems and orchestrates AI agents and human approvers inside the same flow, with the same audit trail. It turns unstructured inputs such as emails, spreadsheets, and documents into coordinated workflows. Agents carry out tasks across systems and teams, communicate through the email and Slack channels people already use, and route each decision to the person who owns it. In a coordination-constrained process, agents take on the fetching and checking between systems and people keep the decisions. A process owner sees in one place what the agents are doing, what the people are doing, and where a decision is still owed.Strong candidates share four traits:
  • Volume. The process runs daily or weekly, so improvement compounds.
  • Manual coordination. The work between system steps lives in email, spreadsheets, chat, and phone calls.
  • A measure with consequences. A defined SLA, cost target, or error rate gives the improvement something to move.
  • Clear boundaries. The process has a defined trigger and a defined completion, and it crosses two or three teams at most.
Scoring each trait from 0 to 3 gives a quick read. Processes that score 9 to 12 out of 12 are strong candidates. The most common reason for a low score is choosing something too large. A sub-process inside a broad value chain usually tests better.The candidates look different in every industry. Examples include complaints and controls in financial services, supply chain exceptions in manufacturing, contract and clinical operations in health and life sciences, and dealer and warranty processes in automotive. Supplier onboarding, the process in the worked example above, appears across all of them. In one OSF Digital engagement, vendor onboarding time fell from 38 to 43 days to 7 to 9 days within six weeks.Moving these processes into production takes more than deploying agents. Decision rights, controls, employee roles, data access, and the handoffs between people and agents all need redesigning around the new way of working.

What can the Coordination Cost Diagnostic not tell you?

The Coordination Cost Diagnostic shows where time goes in one process. It does not confirm that the process is ready for agentic execution. Feasibility depends on data availability, the integration surface, and how much genuine variation sits underneath the steps on the worksheet.The diagnostic also stops short of a business case. Multiplying released hours by a loaded cost rate overstates the result, because released time does not convert to saved cost one for one. It produces a directional figure strong enough to justify a closer look.Each result reflects one process at one point in time, as described by whoever filled in the worksheet. Two people from the same team will often produce different worksheets, and that divergence usually tells its own story about how the process runs.

How does OSF Digital deliver Agentforce Operations?

OSF Digital delivers Agentforce Operations through a dedicated global practice and a gated journey that runs from diagnosis to an operating model the customer owns. The journey starts with an assessment of the process landscape, data readiness, architecture gaps, and team maturity. The first domain is chosen where three things meet: high business value, real organizational readiness, and strong proof potential. That phase ends with a prioritized use-case set, an architecture gap analysis, and executive alignment.A forward-deployed pod then builds a running workflow on a real process, in the customer's own environment, with real data and real users, within four weeks. The Forward-Deployed Industry Strategist diagnoses the process from the inside, the Field CTO builds the blueprint live, and the Value Engineer builds the ROI model from the customer's actual baseline. The phase ends with a live process and a quantified business case to support the decision on the next phase.Scaling puts production processes live across the chosen domain and, in parallel, builds the customer's Center of Excellence: the ways of working, the governance model, and the internal capability to keep building. Each engagement starts from OSF Digital's industry process libraries and implementation playbooks. Knowledge transfer runs throughout, and success is measured by whether the customer's team can run the next wave on its own. A completed Coordination Cost Diagnostic is a useful starting point for the first assessment conversation.Find out where your process time goes. Download the Coordination Cost Diagnostic, walk one process, and send your results to OSF Digital for review.More on OSF Digital’s Agentforce Operations practice: Watch the executive briefing on Agentforce Operations with Sean Catlin, EVP and Global Head of Strategy at OSF Digital and one of Channel Insider’s 2026 AI Leaders in the Channel. Or hear one of our customers talk about working with OSF Digital on Agentforce Operations.
Marlena Hij
AuthorMarlena HijContent Lead at OSF DigitalMarlena Hij leads the content function at OSF Digital. She writes about Salesforce solutions and enterprise AI, with a focus on what it takes to move from decision to production, drawing on OSF Digital experts, delivery teams and customer programs. Before OSF Digital, she spent five years at Japan Tobacco International in commercial roles across Spain and Portugal, ending as Market Manager for Travel Retail with P&L ownership and a thirty-person field team, and four years at IBM Europe in marketing automation and campaign management.
Contact: Kateryna Melkomukova
Sign up for the latest news, trends and insightsSUBSCRIBE

Frequently asked questions

Coordination share is the percentage of hands-on time a team spends fetching and checking information rather than making decisions. It is calculated by dividing fetching and checking touch time by total touch time. A coordination share of 60% or more indicates that most effort goes into moving information between people and systems.

No universal target exists, because acceptable waiting depends on the process and its controls. A wait multiple above 10 is a strong signal that cycle time is driven by queues and handoffs rather than by the work itself. The useful comparison is the same process measured again after a change.

Much of the first wave of Agentforce adoption focused on conversational AI and the CRM experience. Agentforce Operations is a process layer that orchestrates AI agents and human approvers across complete end-to-end processes in functions such as finance, procurement, HR, supply chain, and customer operations. It sits above Salesforce, ERP, and communication systems and connects to more than 250 applications, so a single process can span every system it touches.

The Coordination Cost Diagnostic takes about an hour for one process. It needs someone who knows how that process actually runs and a willingness to record it as performed rather than as documented. The output is two ratios, coordination share and wait multiple, with an optional third ratio for rework share.

A high coordination share shows where time goes, and feasibility is a separate question. It depends on data availability, the integration surface, and how much variation sits underneath the process steps. An assessment of the process landscape, data readiness, and architecture tests those factors before anything is built.

A completed worksheet gives OSF Digital a concrete starting point for an assessment that identifies the strongest first domain for Agentforce Operations. Qualifying processes move into a four-week build of a working process in the customer's own environment, with a quantified business case based on the customer's actual baseline.

BROWSE AND READFor More On This Topic
OSF Digital Launches Global Agentforce Operations Practice to Accelerate Enterprise-Scale Agentic TransformationOSF Digital Launches Global Agentforce Operations Practice to Accelerate Enterprise-Scale Agentic Transformation
News
OSF Digital launches a global Agentforce Operations practice to help enterprises move agentic AI from experimentation into governed, scalable production.Read more
Solutions Applied to Government & Public Sector RealitiesAgentforce Operations: Bring the Work Between Your Systems Under Intelligent Control
OSF Digital deploys Agentforce Operations to orchestrate humans and AI agents across your back-office processes, delivering working value in weeks, not years.Read more
Agentforce Operations for Complaints: Deciding vs. FetchingAgentforce Operations for Complaints: Deciding vs. Fetching
Blog
Why more complaints handlers rarely cut cycle time, how Agentforce Operations orchestrates the fetching work around each case, and how to find the constraint.Read more