Contents

Chapter 11

The Whole Cost

At 9.18 on Tuesday morning, the CFO placed an invoice on Maya’s desk.

“I can tell you what the knowledge base costs,” she said. “I cannot tell you what this service-credit use costs.”

The invoice covered a system the organisation already had. The retrieval feature sat inside it. There was no separate purchase order for the narrow use Maya had allowed to resume.

Maya could have called the incremental subscription cost zero. Daniel had configured the feature within the existing licence. No new product had been bought.

The CFO turned over the invoice. On the back she had written the names of the people who had worked on the service-credit use: Daniel, Leila, Priya, Nadia, team leads and advisers. She had left space beside each name.

“The recovery alone took time from all of them,” she said. “Was that free?”

Maya thought of the procedure edit, the pause, the case review and the smaller return to live use. The invoice had missed most of the work.

“No.”

“Then bring me the whole cost. And tell me which part we would pay again.”

The invoice was the easy part

Maya asked Daniel and Leila to meet with the CFO that afternoon. They brought the decision record, time notes from the comparison and release, the monitoring review, and the account of the source-change recovery.

Daniel began with the licence. The organisation already paid for the knowledge base. The retrieval feature had not yet caused an additional subscription charge. If the use grew, the contract and usage terms would need checking. For the current small group, he could confirm the charge on the invoice. He could not claim that every future use would sit inside it at no additional cost.

The CFO drew two columns.

In one she wrote cash paid because of this use. In the other she wrote resources used by this use.

The first column was short. The second filled quickly.

Daniel had configured access and sources, tested questions, made a controlled change, investigated the unreviewed procedure edit, disabled the feature and tested the narrower return. Leila had established procedure status, reviewed case effects and changed the publication route. Priya had prepared training, supported advisers and collected doubtful outputs. Team leads had reviewed cases and approvals. Nadia had challenged the comparison and recovery evidence. Advisers had spent time learning, asking, checking and reporting. Maya had made the decisions.

Some of that work would have happened without the feature. The two conflicting procedures needed repair in any event. Reversed decisions needed better records. The CFO would not charge the entire process repair to the retrieval feature merely because the trial exposed it.

She also would not ignore the extra work caused by putting the feature into service.

“What would we have done if we had chosen ordinary search?” she asked.

Leila described the fallback used during the pause. Advisers could find approved procedure copies and refer uncertain cases to team leads. It was slower for some questions and clearer for others. The organisation would still need a procedure owner and a sound record of reversals. It would not need Daniel to test generated answers after each source change.

The comparison became the basis for the cost question. The choice was not between this feature and a world with no service-credit work. It was between ways of doing work the organisation already owed customers.

A cost belongs to the choice only to the extent that the choice creates or changes it.

The CFO underlined changes. She wanted the calculation to show added work, avoided work and work shifted between people. A minute saved by an adviser but added to a team lead’s review was not a minute removed from the organisation.

Time that appeared in different places

Priya brought the Time-Savings Tracker to the meeting. It contained cases where the feature helped an adviser locate the current procedure. It also contained cases where reading the generated answer and then checking the source took longer than opening the known procedure directly.

The entries were uneven. Some advisers used the feature often. Others had few service-credit cases or preferred ordinary search. The narrowed boundary had changed the work again. For approval and eligibility questions, the feature now displayed source material and status without drafting a case outcome. That reduced one kind of risk. It also meant any earlier measurement of time spent reading a full explanation did not describe the current configuration.

“We cannot carry the old saving forward unchanged,” Priya said.

The CFO agreed. She asked what a completed task was. Was it finding a document, giving the customer an accurate answer, recording a decision, or finishing a case without another contact?

Leila wanted the last of those considered, but the team did not yet have evidence that the feature changed repeat contact. They could measure a procedure search in a small set of cases. They could record what happened after it. They could not claim that a faster search caused a better customer outcome.

A difficult reversal case showed why. The feature had helped locate the procedure quickly. The adviser then spent time establishing who had approved the earlier reversal. The customer needed a later answer because the case record lacked a fact. That later work was not caused by the feature. It was also not erased by a faster search.

The CFO asked Priya to separate the steps in the tracker. Search for the source. Read and check the passage. Establish case facts. Obtain approval. Respond to the customer. Record the result. If a step moved or grew, they would see it.

Priya worried that advisers would be asked to run a stopwatch while serving customers.

“No,” the CFO said. “We need a small, credible sample and an honest limit. We do not need false precision on every call.”

Leila agreed to use selected cases reviewed through ordinary team lead work. Priya would gather the adviser account while the case was fresh. The exercise would stop if it began taking more effort than the question warranted.

The CFO also asked what happened to time that was freed. If an adviser completed a search sooner, did the organisation pay less, avoid overtime, handle a waiting case, or simply have breathing room for careful work? Each could be useful. They were not the same financial benefit.

If a task takes less time, what happens to the time released?

Guidance

Follow the work beyond the shorter step. Consider whether the time helps someone handle another case, check more carefully or avoid paid overtime. State what you can observe and leave any unsupported cash saving out of the answer.

Your answers are saved only in this browser on this device. Nothing is sent anywhere.

Maya had heard proposals that turned every saved minute into a salary saving. Nobody in the small group had reduced their hours because of the feature. A claim of cash savings would be false. Freed capacity might still matter if the team could use it well, but it needed its own evidence.

Time released is a benefit only when you can say what the time becomes.

The cost of a correction

Daniel had logged the work during the source-change incident. He had not done so for a finance exercise. He needed to know who had changed what, when access was paused, and what had to be retested.

The CFO used that record to identify recovery work. Daniel had investigated and configured. Leila had established the approved rule and changed the source publication route. Team leads had checked relevant cases. Priya had handled adviser questions and tested the narrowed use. Nadia had reviewed the diagnosis. Maya had paused and resumed the authority.

They could count time spent. They could not take that single event, multiply it by a convenient frequency and present it as an annual incident cost. They did not know how often an event like it would occur.

The CFO proposed two views. The first would show the actual recovery cost to date. The second would show a forward allowance for reviewing source changes and handling reports, with a range based on what the team had now experienced. The allowance was a planning assumption, not a prediction of incidents.

Daniel objected to one line in the first draft. It charged every hour Leila spent repairing procedure ownership to the feature.

“We needed to do that anyway,” he said.

Leila agreed. The conflicting June and December procedures and the missing record of reversals existed before the retrieval trial. The tool made the problems visible and raised the cost of leaving them unresolved. But the base responsibility to keep current procedures and complete case records belonged to customer operations.

They split the line. The ordinary procedure repair remained a service cost. The extra review needed to prepare, test and maintain the retrieval collection was attributed to this use. Where the separation was uncertain, they marked the estimate and did not conceal it inside a total.

The CFO asked Nadia to challenge the benefit side with the same care. If they excluded pre-existing process faults from the feature’s cost, they could not credit the feature with every improvement that followed those repairs.

Which improvements would your team have made even without this feature?

Guidance

Compare the current use with a credible route for doing the same work without it. Separate the benefit of repairing procedures from the benefit of retrieving them differently. Use the same boundary when attributing costs so that one side of the comparison does not flatter the feature.

Nadia nodded. A cleaner procedure might help advisers whether they used retrieval or ordinary search. The organisation would need to avoid claiming the same benefit twice.

A number with its edges showing

The CFO opened the Cost Worksheet. She asked each owner for recorded hours rather than an impression of effort. Where people had not tracked time directly, they reconstructed a range from calendar sessions, support records and the change history. They marked those entries as estimates.

Finance applied its approved internal labour rates to the time. The rates were a way to compare uses of people’s capacity. They were not invoices sent between departments.

The first result surprised Daniel. The direct subscription increment for the small group was nil under the current licence. The labour spent preparing, operating and recovering the use was material. The exact amount depended on how they allocated shared work, but every reasonable version was much larger than zero.

The CFO refused to collapse everything into a single number without labels. She separated costs already incurred from the estimated cost of continuing under the narrowed boundary. She separated cash commitments from internal capacity. She kept the work that would be needed even without the feature outside the comparison.

FIG-11.1

The cost view Finance asked for

Cost or benefitWhat the team could supportHow it would be treated
Existing knowledge-base licencePaid already; no separate charge yet for this small useShow the relevant contract position and any future usage condition
Setup, training and initial testingSessions, configuration records and staff timeRecord as incurred preparation cost
Source review and controlled updatesLeila’s approval work and Daniel’s retrieval testsEstimate as continuing work under the narrowed boundary
Output checks and case reviewAdviser and team lead work seen in selected casesCompare with the ordinary procedure route
Incident and recoveryRecorded pause, investigation, case review and retestShow actual cost to date; use a range for future planning
ExitAccess removal, source separation, record retention and staff guidance if use stopsInclude a forward provision, not an incurred cost
Search time savedSupported in some tracked cases, not allShow as observed capacity, not automatic cash savings
Customer and decision qualityCases and reports show useful checks, but no reliable monetary effect yetDescribe the evidence without pricing an unproved benefit
Source: author’s method and fictional case.

FIG-11.2

The whole cost, above and below the waterline

The subscription

What the invoice shows. Often the smallest part.

waterline

People's time - preparing, operating and checking the use.

Errors and recovery - finding, correcting and repairing.

Monitoring - the continuing care that keeps it honest.

Exit - leaving cleanly if the use stops.

The licence is the visible part. The operating cost sits underneath it. Source: author's method and fictional case.

The CFO then produced a working total for the period already completed, using the recorded and estimated hours with Finance’s rates. She attached the assumptions and ranges. Maya could take the number to Tom, but only with its description: it was the cost of preparing, running and recovering the limited service-credit use to date. It was not a price per answer, an annual forecast or a return on investment.

For the next period, the CFO gave Maya a different figure. It assumed the same small group, the narrower source-display behaviour for approval and eligibility questions, routine source testing, a modest amount of adviser support and continued monitoring. It did not assume another serious incident. It included a separate allowance to investigate and respond if one occurred.

Maya asked which number was the “defensible number” the CFO had wanted.

“Both, for different decisions,” the CFO said. “One tells you what we spent learning. The other tells you what you are authorising to continue. Don’t add them together and call it a subscription cost.”

She was not avoiding a conclusion. She was preserving what each number meant.

Benefits without a convenient conversion

The Benefits and ROI Calculator had a place for money saved. Priya could not fill it from the tracker alone.

Some searches were shorter. Some checks were longer. The feature had helped advisers locate the current procedure and exposed source problems. The ordinary approval route had prevented an unauthorised credit during the source-change incident. None of those observations yielded a sound monetary return on their own.

Daniel suggested that they count avoided wrong credits. Nadia objected.

“We do not know how many would have been wrong without the feature. In one case the feature itself gave a misleading instruction. The team lead control held.”

They could describe that control as valuable. They could not assign the avoided loss to the retrieval feature. The same team lead route existed before it.

Leila suggested a narrower benefit: the feature sometimes made the right source easier to find, particularly when the adviser used the customer’s words rather than the procedure title. Priya had examples in the tracker. They could test whether that benefit held under the narrower configuration and during ordinary work without Priya beside the advisers.

The CFO accepted that as a hypothesis with evidence, not yet a projected saving. She wanted the next review to show how many relevant tasks were observed, how they were selected, what comparison route was used, and whether the search advantage remained after source checking. If the team selected only the cases where the feature helped, the calculation would flatter it.

Maya asked whether a business case required a positive return number.

“A decision requires a reason,” the CFO said. “A made-up return is not a reason.”

They might continue a limited use because it improved access to current instructions, provided a checkable source and supported more consistent work at an acceptable cost. They could also stop if the benefit remained too small for the effort of maintaining it. The calculation had to make that choice visible.

The CFO entered the observed benefits in the Benefits and ROI Calculator without forcing a monetary value. The financial return remained unproven. She kept a separate place for service quality evidence and the costs of checking it.

A benefit you cannot yet price can still matter. It should not be disguised as money.

What happens if they leave

At 2.36 on Thursday afternoon, Maya asked Daniel what it would cost to stop using the feature for service credits.

“Disable access,” he said.

“That is one action.”

Daniel thought about the pause they had already run. Advisers would need a clear route back to ordinary search. Saved answers and screen captures should not continue to circulate as current guidance. Leila would need to confirm that the approved procedures were available without the retrieval view. The incident and decision records would need to remain accessible under the organisation’s normal controls. Any special source collection or configuration would have to be removed or left in a known state. Staff support material would need updating.

The work would be manageable because the use was narrow. It was not zero.

The CFO added exit to the Cost Worksheet. She did not treat it as a penalty for trying the feature. It was part of the choice to operate a system that people might come to depend on. If the benefit did not justify continuation, the organisation needed enough capacity to leave cleanly.

Daniel asked whether a possible exit cost should be included in every forward period. The CFO proposed a provision shown separately. It should not be counted as if the organisation retired the feature every month. It should also not vanish from the decision because retirement was inconvenient to price.

They checked the contract position as well. The feature was part of an existing knowledge base. The team did not need to invent a termination charge. They did need to understand what would happen to configuration, logs and source copies if the organisation changed the broader system later. Daniel marked what he knew and what remained to be confirmed.

Maya saw a second reason to keep the use narrow. A small group with a clear fallback could stop without leaving customer operations unable to find its procedures. If the whole process became organised around the feature, exit would become harder and more costly even if the licence stayed the same.

The case Tom would read

On Monday, Maya drafted the Business Case Template with Leila and the CFO.

The opening question was modest: should the organisation continue the limited service-credit retrieval use for another review period under its narrowed boundary? It did not ask for a company-wide AI strategy or a general endorsement of assistants.

The options were plain. Continue the current small group and measure the remaining benefit. Expand access, which would require more training, source review and support. Or stop and use the approved ordinary search route while keeping the procedure and record repairs.

They did not recommend expansion. The source-change route had only recently been repaired. The small group had not yet shown a stable benefit across ordinary work. Leila and Daniel could support a further limited period, with a stated review date and a cap on the staff time they would commit to monitoring and testing. If the work exceeded that allowance, they would return to Maya rather than let it become invisible labour.

The case showed actual costs to date with their basis, the forward cost of the limited option, and the uncertain benefit. It included the case where a doubtful answer was caught, the source-change pause and the narrower return. It did not describe the incident as proof the feature failed, or the successful recovery as proof it was worth continuing.

Nadia read the draft and found a sentence that said the feature had “improved decision quality”.

“What evidence supports that?” she asked.

Leila could point to advisers locating current sources and reporting doubtful answers. She could not show that customer decisions as a whole had improved. Maya changed the sentence: The limited use has made some procedure searches easier and produced checkable outputs. Its effect on customer decision quality has not yet been established.

The CFO kept that wording.

Tom would receive a business case that gave him less certainty than a colourful return chart. It would give him a clearer decision.

The question beneath the total

At 4.11, Maya met the CFO again.

The working numbers were complete enough to support the next bounded decision. The CFO wanted to know who would approve the continuing staff time. Daniel’s work did not appear as a new invoice. Priya’s floor support could be absorbed for a while. Leila’s source reviews competed with other customer operations work.

If those contributions remained informal, the business case would understate cost as soon as people stopped logging them. It would also place the burden of maintaining the feature on people whose ordinary targets did not leave room for it.

Maya assigned the continuing work in the decision. Daniel would have time for controlled source tests and technical support. Leila would own the source approvals and service review. Team leads would include source checking in their case discussions. Priya’s extra support would taper to defined sessions rather than an open-ended expectation that she be available at every desk. Nadia would challenge the next review. Maya would decide again at its end.

The CFO asked one final question.

“Suppose the next review shows a small improvement in finding procedures, but maintaining the feature still takes all this work. What will you do?”

Maya did not promise to continue because the organisation had already invested in the comparison. Nor did she promise to stop because the first full cost was higher than an invoice suggested.

“I will decide whether that improvement is worth the continuing cost and the authority it requires,” she said. “And if it is not, we can leave.”

The CFO closed the worksheet. That was the decision she had wanted a number to serve.

Use the Cost Worksheet to separate cash, people’s time, continuing care, recovery and exit. Use the Benefits and ROI Calculator for benefits you can support, leaving uncertain gains unpriced. Put the alternatives, assumptions and limits into the Business Case Template so a decision-maker can see what continuing the use commits the organisation to.

Chapter 12, Earning Authority Again, asks when the permission to continue should expire, and what evidence would justify renewing it or retiring the feature.