Saved time is only the beginning.

02Count the complete workflow, not only model usage.

03Measure review, exceptions, integration, monitoring and ownership.

04Separate released capacity, avoided spending, incremental contribution and risk reduction.

05These can all matter, but adding them without clear boundaries can double-count the same benefit.

06The business case is a conditional model that should improve as evidence arrives.

07It is not a guarantee of savings.

08This explanation supplies a framework and explicit illustrative arithmetic, not investment advice or a forecast of dotSuper client results.

02Their 2025 publication reports an average 15% increase in issues resolved per hour, with substantial variation across worker experience.

03That denominator is task productivity, not firm profit, staffing reduction or manufacturing output.

04Generative AI at Work, Quarterly Journal of Economics 140(2), 2025, author's published-paper copy (source 1).

05A different experiment points the other way.

06METR's 2025 randomised study involved 16 experienced open-source developers and 246 tasks on familiar repositories; allowing the tested tools increased completion time by 19%.

07This is evidence about that setting and period, not proof that all coding assistance slows work.

08METR, 10 July 2025 (source 2).

09METR's February 2026 update cautions that a later experiment produced an unreliable signal of current effects, including participation and task-selection issues.

10Therefore the 2025 finding should not be presented as a current universal productivity estimate or as settled evidence that newer tools have the same effect.

11METR, 24 February 2026 update (source 3).

12Sculley and co-authors describe how machine-learning systems accumulate maintenance obligations through data and surrounding software.

13The paper provides an engineering explanation for hidden work, not an empirical percentage to add to every budget.

14Hidden Technical Debt in Machine Learning Systems, NeurIPS 2015 (source 4).

02A cheap model call can generate an expensive unresolved exception.

03Track cost per completed, acceptable outcome alongside cost per request.

04One-time work includes process discovery, data preparation, integration, permissions, evaluation, training and documentation.

05Recurring work includes model or software fees, hosting, monitoring, review, exception resolution, support and periodic revalidation.

06Exit work includes export, replacement, access removal and handover.

07Separate internal time from external expenditure.

08Both are costs, but they affect cash differently.

09A salaried employee spending less time on data entry does not automatically lower payroll.

10If that person uses the time to clear a backlog, the value depends on what the backlog is worth and whether the organisation captures that benefit.

11Define benefit categories without overlap.

12An avoided hire is different from released capacity in an existing team.

13Incremental contribution should account for variable costs and demand constraints.

14Avoided error cost should be based on credible incidence and consequence estimates, with uncertainty shown.

15Never value the same hour as both payroll saving and extra output unless the model explains why those are separate effects.

02Assume 2,000 documents, six minutes of baseline handling each, and a fully loaded labour value of AED 80 per hour.

03Baseline effort is 200 hours.

04With assistance, every document still needs two minutes of handling.

05Twenty percent also require eight additional exception minutes.

06Total effort is (2,000 × 2 + 400 × 8)/60 = 120 hours.

07Released capacity is 80 hours, valued at AED 6,400 if all of it can be usefully captured.

08Assume only 60% of that capacity value is captured through a specified operational use.

09Captured benefit is AED 3,840 monthly.

10Assume recurring non-labour and support cost of AED 2,500, after ensuring the support estimate does not duplicate the handling labour already counted.

11Net monthly economic benefit is AED 1,340.

12With an illustrative one-time cost of AED 30,000, simple payback is 30,000/1,340, or about 22.4 months, if benefit and costs remain stable.

13It excludes discounting, taxes, financing and implementation ramp-up.

14It is an economic-value payback, not necessarily cash payback.

15If no spending is avoided and no extra contribution is captured, payroll-valued capacity alone cannot repay a cash outlay.

16Now increase the exception share to 30%.

17Handling becomes (4,000 + 600 × 8)/60 = 146.7 hours.

18Released capacity is 53.3 hours, with gross value of about AED 4,266.7.

19At 60% capture, benefit is AED 2,560; after recurring cost, only AED 60 remains.

20Under these assumptions, the business case is highly sensitive to exceptions.

02This is sensitivity analysis, not a probability distribution.

03It does not imply that these outcomes are equally likely or that a provider has achieved any of them.

04At the assumed 20% exception rate, the capture share required to cover AED 2,500 recurring cost is 2,500/6,400 = 39.1%.

05That threshold still excludes recovery of the initial investment.

06A recurring break-even claim is therefore different from an acceptable total-project return.

07A useful pilot measures the uncertain inputs most likely to change the decision.

08If exception handling dominates the sensitivity, test difficult documents early.

09If value capture is uncertain, agree what the released time will be used for and how that use will be verified.

02Select comparable work, track task mix and preserve unsuccessful attempts.

03Report who participated, how long the evaluation ran and what changed besides AI.

04A before/after comparison can be useful operationally while remaining weak evidence of causation.

05Measure adoption, rework and review quality.

06People can underuse a helpful tool or overtrust a flawed one.

07A short demonstration rarely captures support burden, seasonal complexity or staff turnover.

08Schedule a review after the workflow has encountered real exceptions, with a named person responsible for the evidence.

09Set explicit stop and continue criteria.

10A pilot can correctly conclude that a simpler rule, better data or no change is preferable.

11Document that decision rather than treating every pilot as the first instalment of an inevitable rollout.

02Each source supports only the scope stated beside the claim.

02Test the exception rate, capture share and recurring work before approving a broader rollout.

02A narrow time-saving model can miss those benefits.

03Equally, vague strategic benefits should not become a blank cheque.

04Name the outcome and the evidence that could establish it.

05The two productivity studies involve different tasks, workers, tools and research designs.

06They should never be averaged into a single “AI productivity effect.”

07Neither directly estimates a Saudi manufacturer's returns or a UAE accounts-payable team's performance.

08No model price, vendor quote or market-wide ROI benchmark is asserted.

09Prices and contracts change; collect current quotes for a real decision.

10No search-volume or popularity claim is made.