- Most of a complaints handler's day goes to fetching, not deciding. Chasing records, re-checking reviewed cases, and rebuilding files for escalation crowd out the judgment work, and average handle time hides the split.
- More handlers rarely means faster cycle times. Extra handlers send more requests to upstream teams with the same capacity, so the bottleneck moves somewhere no dashboard tracks it.
- Orchestration runs the process from a single blueprint. Routing and triage follow fixed rules, every action is timestamped for escalation, and a rule change updates once instead of requiring retraining.
- The decision stays with a person, and the core systems stay in place. Agentforce Operations coordinates the work around the case without replacing case management or the core platform.
- The judgment-to-fetching ratio is the fastest diagnostic. Tracing one complaint category from receipt to decision shows whether the constraint is capacity or coordination.
Your Complaints Team Isn't Slow; It's Waiting.
Most of a handler's day is spent chasing, waiting, and checking. With orchestration, the task of fetching is eliminated, and more time is left over for judgment and critical thinking.If you ask a complaints handler about what they did today, the majority of their answer isn’t about investigation work per se.They sent an email to the payments team asking for a transaction history. They waited for a reply. They followed up on it. They discovered that the person responsible for that data is currently on leave. Since it's company policy to have two people look at a case, they re-read their colleague's case file. They compiled a chronology from four different systems into a document which a reviewer could follow. After that, they drew up a response.A handler's day consists of two types of work: deciding and fetching. Deciding involves assessing the facts, applying the rules, and reaching an outcome, while fetching means obtaining the items that the decision requires.Almost all complaint handling operations assess the total figure and not the individual components. The average handle time is calculated in the same way whether the handler is dealing with a complicated case or is waiting three days for a document.Which is why the capacity problem in the area of complaints shows such odd behaviour since companies add more handlers but the cycle time hardly changes at all.Where the time actually goes; deciding vs fetching
In all the complaint operations the same six activities take up most of the handler's time and none of these is investigation.Gathering evidence by obtaining the records from systems that the handler cannot access directly, which involves asking a person who can access them and then waiting.Checking work. Quality assurance and second-pair-of-eyes review, both of which require re-reading the entire case rather than sampling it.Putting together the file by turning the notes, screenshots, and email threads into a form that a reviewer, an approver, or an external body can follow.Drafting. Writing a response that states a decision, gives reasons, references the evidence, and explains the escalation route.Building on the escalation: when a case is referred to an ombudsman or to some other external body, the file generally has to be reconstructed rather than exported since it was never kept in a single location.When a rule changes, the process is carried out again. This involves applying a new procedure to a population of cases and currently consists of retraining people and rewriting the guidance.Two of the tasks—carrying out the check and carrying out the rebuilding—are completely redundant since the work had already been done once.Why more handlers doesn't fix it
If one person were to carry out the task of handling complaints, the output would increase roughly in proportion to the number of people employed.That is not the case. When a complaint involves multiple teams, each interaction involves a handoff accompanied by a queue. If you add handlers, this results in more requests being sent to the same upstream teams, who were not given additional resources to cope with the increase. As a result, the bottleneck shifts away from the complaints function, where it can be seen on a dashboard, and into other functions where it cannot be seen.We should also check the scales. Having more cases requires more quality assurance, more approvals, and more review, all of which involves people reading what other people have written.The same proportions apply to all processes. About 30 to 40 per cent of a typical business process is already automated, the remaining 60 to 70 per cent consisting of human coordination. When more people are added, capacity is increased for the part which was not the bottleneck.There is a fundamental aspect of the way complaints operations were structured. The process was learned by individuals and retained in their minds before being passed on. Because of this, half of the process knowledge is held by one person, a quarter by another, and no one is able to put the entire process together when asked to do so.What orchestrating the coordination layer means
The other option is not a faster case management system; for case management keeps the record of the decision, it does not keep the work that led to the decision.The term 'orchestration' has four specific meanings.The process is described once, as a blueprint. The stages, the tasks inside each stage, what data each task needs, who or what executes it, how long it should take, and the conditions under which a human has to be involved. Not distributed as guidance, held as a single executable definition.Execution is deterministic. The same inputs produce the same sequence of steps, every time. Routing, triage, scoring, and structured data validation run to fixed rules rather than to a model's discretion, which is what makes this usable in a regulated process.Every action is logged. Each step, whether performed by an agent or a person, is recorded with a timestamp. That record is what an external body asks for when a case escalates. Built during the case, it becomes an export rather than a reconstruction.People are brought in at judgment and at exceptions. The complaint decision stays with an accountable person. What changes is that they stop spending judgment time on chasing a document.There's a fifth property that matters more in complaints than in most processes. When a rule changes, you change the blueprint, and the change applies to cases from that point forward. The alternative is a change request, an IT release cycle, and a retraining exercise, none of which fit inside a regulator's implementation window.What this is not
It doesn't decide complaints. The outcome stays with a person accountable for it.It doesn't replace your case management system. The case record still holds the case.It doesn't require rebuilding your core platform. The argument runs the other way: the core stays as it is, and the work around it changes. That's what makes this a shorter piece of work than a platform program.And it's separate from customer-facing AI. Agentforce Operations runs middle and back office coordination. It doesn't interact with service or sales agents a firm already has, and it's applied only to the specific processes where manual coordination is the constraint.What comparable processes show
This is the loan manufacturing operation of a lender based in North America. A team of 280 people was working under a 10-day service level agreement that in reality took 15 days to fulfil. The coordination of their work was carried out by means of email, process documents, and institutional memory. Following the prioritisation of the various sub-processes and the introduction of agentic workflows, the outcome was the elimination of over 55,000 hours of manual work and the freeing up of about 35 FTE, which amounted to approximately 2.16 million USD in the first year.The procurement department of a global software company had a fragmented way of taking in requests, with the associated manual work distributed among various systems. As a result, more than 10,000 hours were saved and approximately 314,000 USD was achieved in the first year. The initial working blueprint was put into action within 24 hours of the workshop in which it was produced.What they establish is what happens to the coordination layer of a document-heavy, deadline-bound process when it stops being run by people remembering it.The situation with the loan manufacturing case is the more appropriate analogy, and the point worth considering is the SLA. A commitment of ten days when it actually takes fifteen days represents a 50 per cent overrun on a regulated clock and is something that a team of 280 people has been unable to resolve, since the people themselves were not the limiting factor.How to find out whether this applies to you
Look at a single complaint category and trace how the time is distributed between the date it is received and the date a decision is made. Not the process as it is documented, but the process actually carried out, including the waiting period.Divide each step into two columns—one for steps that involve judgment and one for those that involve having to go and get something or check something that has already been done.The second column is typically bigger than anybody in the room anticipates, since it shows how the deadline is being used. More than any other indicator, that ratio reveals whether the limitation you're facing is one of capacity or coordination.You will also, when carrying out this exercise, discover that you already possess most of what is required by a blueprint. Complaints teams have process maps, standard operating procedures, and DISP or similar rule mappings since regulators require them. These documents form the input.What delivers it
Agentforce Operations orchestrates the work around the case. It builds the blueprint from process documentation your team already holds, runs the sequence deterministically, timestamps every agent and human action, and routes exceptions to a person. Handlers interact with it through email, Slack, or Teams, so the change management cost is low.How we work with clients on this
An Agentforce Operations Proof of Value carries out this approach; it involves a structured and time-limited process made up of five stages: agreeing on the metrics that are important, scoring and narrowing down the candidate processes, designing the blueprint based on the documentation already available, testing it using real process data, and then presenting the quantified value together with a roadmap.You end up with a tested blueprint, a return on investment projection, and a prioritized set of recommendations, and feasibility is addressed before any commitment is made to the budget.Questions worth asking internally
How many days there being between the time your last hundred complaints were received and the time they reached the person who made the decision?What share of a handler's day is spent deciding, and what share is spent fetching or checking? If nobody can answer that, that's the finding.When a case reaches a higher level, is the file exported from one system or is it reconstructed from several?When was it last time the rules were altered, and how long did it take before all the handlers were carrying out the new version consistently, and how do you know?Start with the ratio
The first thing you should do is to find out how much of your handlers' time is spent on judgment and how much on retrieval.We will examine one complaint process as it is carried out, compare the blueprint with your actual data, and provide you with a quantified indication of the costs associated with the coordination layer. Either way you will keep both the blueprint and the roadmap. Book an Agentforce Operations Proof of Value.About OSF Digital
OSF Digital, an AI-forward, data-driven, Salesforce-centric consulting services company, helps global enterprises navigate technology complexity and modernize their operations through expert-led strategy, multi-cloud execution, and measurable outcomes.We combine deep technical expertise, industry insight, and a distinctive partnership mindset to simplify complexity and deliver real business value, from advisory through implementation to managed services. We work with Fortune 500 and regulated enterprises where the cost of getting it wrong is high, demonstrating measurable growth, building on trusted data, and putting intelligent agents into production.
Contact: Kateryna Melkomukova
Sign up for the latest news, trends and insightsSUBSCRIBE
FAQ
Most complaint cases depend on records and approvals from other teams, and each handoff adds a queue. More handlers send more requests to those same upstream teams, so the delay shifts out of the complaints function instead of shrinking. OSF Digital helps firms trace where that waiting time actually sits before deciding whether headcount or coordination is the fix.
Orchestration means the complaints process runs from a single executable blueprint that defines stages, tasks, data needs, owners, and the points where a person must step in. Routing, triage, and validation follow fixed rules, and every agent and human action is timestamped. OSF Digital builds that blueprint with complaints teams, starting from the process documentation they already hold.
No. The complaint outcome stays with an accountable person, and Agentforce Operations handles the coordination around the case, such as gathering evidence and assembling the file. OSF Digital designs the blueprint so the human checkpoints sit at judgment points and exceptions.
The case management system continues to hold the case record, and the core platform stays as it is. Agentforce Operations changes the work around the case rather than the system of record. OSF Digital scopes the work so it fits alongside current systems instead of turning into a platform program.
A rule change updates the blueprint once and applies to cases from that point forward, instead of moving through a change request, an IT release, and a retraining cycle. That makes it easier to work within a regulator's implementation window. OSF Digital maps rule changes, such as DISP updates, into the blueprint so handlers work to the current version consistently.
Trace one complaint category from receipt to decision, including waiting time, and sort each step into judgment or fetching and re-checking. If the fetching column is larger, the constraint is coordination, and more handlers will not solve it. OSF Digital runs this analysis on real process data as part of an Agentforce Operations Proof of Value.
The Proof of Value is a structured, time-limited engagement in five stages: agreeing on metrics, scoring candidate processes, designing the blueprint from existing documentation, testing it on real process data, and presenting quantified value with a roadmap. The firm keeps the tested blueprint, an ROI projection, and prioritized recommendations. OSF Digital addresses feasibility before any budget commitment is made.
What a Complaint Costs a Financial Services FirmHandling, escalation, redress, prevention, and lost customers: the five costs behind every complaint, with regulator data from seven markets.Read more
What Agentforce Operations Is and What It Isn't Agentforce Operations runs the coordination work between systems, not the work inside your CRM. What it does, how it differs from RPA, and which processes fit.Read more
From Pilots to Pattern: The Operating Model for the Agentic EnterpriseThe capability question has been answered. The harder question is what to do with it. Most enterprises have succeeded at the demo. Some at the pilot. Very few have built the repeatable pattern that takes one working process to many in production, a thousand users, and a self-sufficient Center of Excellence inside twelve months. This whitepaper sets out the missing layer of the agentic architecture and the operating model that turns it into compounding business value.Download