Touchless rate: the one number that tells you if AI is working in your back office

An AI tool can read a document quickly and still leave your team with nearly all the work. Someone checks its answer, copies the details, asks for the missing attachment and marks the case complete. The demonstration looked useful. The queue still needs a person at every turn.

Touchless rate asks a more practical question: how many cases reached done without anyone touching them? It puts the result in terms an operator can recognise, and gives us a number to improve together.

Count finished cases, not clever tasks

Touchless rate = cases nobody touched ÷ all cases. Express that fraction as a percentage. A case counts as touchless only when it reaches the endpoint we agreed without human handling along the way. Reading a fax or drafting an email does not, by itself, finish the case.

We define the case boundary before measuring. For a claim, that might run from receiving the documents to confirmed lodgement. If payment reconciliation is outside that boundary, we say so. The number must describe a recognisable piece of work, with a consistent start and finish.

The denominator matters too. We agree which case type and intake period we are measuring, include its exceptions and keep unfinished cases visible. A pending case cannot quietly become a success. If someone corrects, approves or rescues a case, it is touched. Reopened cases need to be reflected in the record.

That makes touchless rate more useful than a vague claim of “efficiency”. You can trace it to individual cases. It is still not the whole picture: errors, rework, waiting time and time spent on exceptions matter. A rising percentage only helps when the work is completed correctly.

Move it one rule at a time

We start with the rules your team already uses. A class of cases may be routine when the required records are present, the details match and no exception applies. Once that path is agreed and proved, those cases can finish without someone checking each one by hand.

The next improvement comes from looking at what still needs a person. Perhaps an agreed request can collect a missing document. Perhaps a check can establish that two records match. We test the change against real case outcomes before expanding what can run on its own.

We do not assume every business starts at zero. If existing systems already finish some cases without human involvement, that belongs in the baseline. Each small win needs to show what changed, using the same definition before and after.

Across an operation that includes judgement cases, touchless rate will not reach 100%. Conflicting evidence, unusual circumstances and decisions reserved for qualified staff still need people. Sending those cases to the right person is part of the design. Removing a necessary judgement call to raise a score would defeat the purpose.

The capacity arithmetic

The simple capacity multiplier is 1 ÷ (1 − touchless), with touchless written as a fraction. At 50%, the calculation is 1 ÷ (1 − 0.5) = 2. When half the cases finish on their own, the same team can handle up to twice as many.

That calculation assumes comparable human effort for each remaining touched case, the same available team time and no other bottleneck. It describes capacity, not revenue or guaranteed throughput. If the remaining exceptions take longer, or another part of the business is full, the practical gain will be lower.

Illustrative worked example: imagine a team with 1,000 minutes available for case handling each week. Each touched case takes 10 minutes, so it can handle 100 cases. These are round numbers chosen to explain the calculation, not a client result or a forecast.

At a 50% touchless rate, an intake of 200 cases would leave 100 for people to handle. If each still takes 10 minutes, that uses the same 1,000 minutes. The other 100 finish on their own. Under those assumptions, capacity can double without increasing the team’s case-handling time.

Now suppose the remaining cases take 20 minutes because they are harder. The team can handle 50 touched cases in those 1,000 minutes. At 50% touchless, that supports 100 total cases. This is why we measure the work left with people as well as the percentage that leaves their desks.

Measure before, during and after

1. 01Learn your business: mapping takes days or weeks, depending on your business. We sit with your team, watch the work and map every case type, step, workaround and exception. For each potential improvement, agree the case boundary, current touchless rate, handling effort and what done means.
2. 02Start with small wins: deliver small, contained improvements found along the way without waiting for the whole map. Measure each one before and after, including completed and pending cases, handling effort and rework. We prove each improvement on real cases before going further.
3. 03Transform the company: when the small wins show it is applicable and the results say it is worth it, extend the case engine across the business’s workflows. Keep measuring the outcomes and work still reaching people. If it is not applicable, we say so honestly.

The record behind the number lets us discuss a specific case when something looks wrong. A change in the mix of incoming work can move the percentage even when the rules have not changed. We keep that context beside the result so we can decide what to improve next.

If you would like to work out what touchless rate could mean for your team, start with a free conversation. We can talk through one case type together.