A calculation on hypothetical assumptions. It is not the result of an implementation, a HumGu price offer or a forecast. All the numbers serve to illustrate the method of assessment; they have not been confirmed by measurement on a real process.
Task
Assess the analysis of inquiries and a draft for checking. We compare work of the same required quality, including control, corrections and exceptions. A positive calculation does not replace checking quality and feasibility.
How do we understand the measures?
- Working time
- the sum of people's active work: reading, preparation, control, corrections and exceptions. Parallel work by several people is added up.
- Lead time
- calendar time from receipt of the case to the agreed completion, including waiting.
- Throughput
- the number of correctly completed cases in a given period, for the agreed scope.
- Cost per completed case
- the allocated work, tools and other costs of the agreed process, including cases requiring corrections. The average also includes the cost of failed attempts, allocated to correctly completed cases. The launch outlay is shown separately so that it is not counted twice.
Minutes in this model mean people's working time. We do not derive the client's waiting time or the throughput of the process from them. The parameters are hypotheses. The conditions of quality and permissibility of use must be met before the benefit is assessed; a positive result does not replace them.
Base assumptions
- 1200 cases per month; currently an average of 10 minutes of active work per case.
- 75% of cases covered by assistance. For these, 2 minutes of work and control + 20% × 5 minutes of corrections = an expected 3 minutes. The remaining cases still take 10 minutes.
- Internal hourly rate: PLN 90. Additional operational supervision: 8 hours per month, beyond the control of individual cases.
- Expenses: PLN 0.80 per assisted case + PLN 1,800 per month in fixed costs of tools/infrastructure and support.
- Launch: PLN 18,000 in cash and 20 hours of internal preparation. These are assumptions, not HumGu prices.
In the model, qualification is included in the working times, and cases handled the existing way do not generate additional AI calls. A different mechanism requires adding its cost and work. The assumptions need to be checked on representative cases.
The calculation step by step
- Currently: 1200 × 10 / 60 = 200 hours per month.
- With assistance: 900 × 3 / 60 + 300 × 10 / 60 = 95 hours of work on cases.
- The difference of 105 hours, less 8 hours of additional supervision, gives 97 hours of net capacity.
- Valuation of this capacity: 97 × 90 = PLN 8,730.
- Additional operating expenses: 900 × 0.80 + 1800 = PLN 2,520.
- Potential economic result: 8730 − 2520 = PLN 6,210 per month.
Check by another route: the current work is valued at PLN 18,000 per month. Work on cases, supervision and expenses in the assisted variant come to 95 × 90 + 8 × 90 + 2520 = PLN 11,790. The difference is again PLN 6,210.
Three scenarios
- Cautious: 600 cases, 60% assisted, 6 minutes per such case → 16 hours of net capacity and a negative economic result of −PLN 648 per month.
- Base: 1200 cases, 75%, 3 minutes → 97 net hours and PLN 6,210 per month.
- Favourable: 1600 cases, 85%, 2.5 minutes → 162 net hours and PLN 11,692 per month.
- Other parameters as in the base. These are sensitivity variants, not their probabilities.
Why is this not yet A cash saving?
If real expenses do not fall, 97 hours do not automatically turn into a financial inflow. They can be used for other work, but we do not add unconfirmed revenue. In cash terms, the launch and PLN 2,520 of monthly operating expenses still have to be covered.
If it were separately confirmed that half of the valuation of the freed-up capacity allows a real expense to be avoided, the cash balance would be 50% × 8730 − 2520 = PLN 1,845 per month. Without such a mechanism, there is no such saving. The percentage values are solely an illustration of sensitivity.
Outlay and threshold
Economic outlay: PLN 18,000 + 20 × PLN 90 = PLN 19,800. At the base PLN 6,210, this gives about 3.19 months of potential economic coverage of the outlay. This is not a cash payback period. A full monthly effect from the start was assumed, without a ramp-up period, taxes, discounting or a valuation of the risk of errors.
With the other base parameters unchanged, a positive monthly economic result starts from 347 cases. This is the threshold of this calculation, not the minimum volume for HumGu services. A change in quality, the cost of checking, fixed costs or the share of cases changes the assessment.
From trial to effect over time
- Preparation: data, reference point, criteria and outlay.
- Trial: quality, share of assisted cases and the full working time including control and exceptions.
- Implementation: integrations, user preparation and responsibility.
- Ramp-up: actual use, the burden of supervision and how the hours are put to use.
- Maintenance: ongoing expenses and checking the parameters after changes.
We match the schedule to the engagement. The model above assumes a full effect from the start; it is not a schedule for achieving benefits. For an investment decision, outlays and effects have to be assigned to real periods, instead of multiplying the target month by the whole implementation period. The volume, the distribution of cases, the share of assistance and all costs also need to be examined. The process owner indicates how the hours are used; cash requires a specific avoided expense.
When to postpone or change the idea?
When the required quality cannot be achieved, permitted data is lacking, control consumes the benefit, or the required kind of financial effect has no real mechanism of achievement. It is also worth comparing tidying up the rules and using the existing tools.