The question we usually put to automation is almost always the wrong one. For any commercial operation, automating part of the work has been a given for years; what’s worth arguing over is what to automate, and what to leave well alone. Automating a process makes it faster and more rigid. That rigidity pays off on a task built for repetition, but on a task that needed a human reading it quietly turns expensive: nothing announces the cost until the same mishandling starts repeating, now on a production line.
Automating Means Freezing—So Choose What Deserves It
An automated task always runs the same way, at the same speed, without fatigue or oversight. That’s exactly what you want for chasing an unpaid invoice, entering an order, or sending a confirmation email: repetitive, rule-bound, high-volume actions that call for no judgment. The duller and more predictable a task is, the better a candidate it makes. Automating it doesn’t just free up time; it also removes the slips of attention that creep into any repetitive manual work.
The criterion flips the moment a task requires reading a context. Replying to an unhappy customer, qualifying an ambiguous lead, negotiating an exception: these are actions where the rule never covers every case, and where the « right » answer depends on what you sense in the moment. Automating them amounts to swapping judgment for a rigid decision tree—one that will handle the anticipated cases correctly and mishandle all the others, which are usually the ones that actually matter. The same rigidity that makes automation dependable on a routine task starts costing you the moment the job calls for discernment.
A sales cycle illustrates the dividing line well. Enriching a prospect record, sending a follow-up reminder at the right time, generating a quote from a catalog, pushing a form into the CRM: all stable, rule-based actions that gain from running on their own. Conversely, the discovery call where you pin down a poorly stated need, the call on whether to grant a discount, the moment you sense it’s time to pick up the phone rather than send one more email—none of that reduces to an equation. The useful boundary separates what follows a rule from what requires reading a situation, and it cuts straight through both « sales » and « admin ». It’s that sorting, not the number of tools deployed, that determines the real return on automation.
| Hand to the machine | Keep in human hands | |
|---|---|---|
| The task | Follows a stable rule, repeats identically | Requires reading a situation |
| Examples | Invoice follow-up, order entry, confirmation email, quote from the catalog | Unhappy customer, ambiguous lead, discount call, discovery call |
| What rigidity produces | A dependable action, free of slips of attention | Mishandled edge cases, the ones that matter |
What Resists Automation, and Why That’s Good News
The most useful figure on the subject is also the most misread. According to the McKinsey Global Institute, about 60% of occupations have at least 30% of tasks that are automatable, but fewer than 5% can be fully automated. In other words, automation nibbles at tasks inside occupations, almost never whole occupations. The fear of « replacement » gets the scale wrong: what disappears are fragments of work—the most mechanical ones—not the role that held them.
This reading changes how you frame a project. The point is to take off someone’s plate what never needed their presence, so their attention can concentrate where it is irreplaceable: the relationship, the exception, the decision. A sales team whose data entry and follow-ups have been automated doesn’t vanish—it spends the time saved selling and understanding its customers. That’s also why trying to automate everything in the customer relationship quickly backfires: past a certain point, automation stops serving customer satisfaction and starts eroding it, the day a customer simply needs to talk to someone.
The Trap: Automating a Process You Haven’t Fixed
That leaves the costliest mistake, and the most common: automating a process that’s still wobbly. As long as an approval chain is muddled, a step gets done « by hand because it’s easier », or an exception comes up one time in three, it isn’t ripe for automation. Freezing it in that state fixes nothing—it industrializes the mess, making it faster and harder to untangle. The order of operations matters: clarify and stabilize the process first, automate it second. It’s the direct corollary of the principle that technology amplifies the process it touches, good or bad.
Taken in that order, the effort truly pays off. A well-automated repetitive task doesn’t just save hours in the moment: it changes the cost structure for good, where a simple manual reorganization unravels the moment the pressure rises again. That’s what separates a process saving from a one-off cut in costs: the saving lasts because it took out the underlying cause rather than papering over a symptom that resurfaces under the next round of pressure.
Conclusion
Automating isn’t an end in itself, and it isn’t inherently virtuous; what it really comes down to is how you allocate effort. Sorted well, it gives the machine the stable, judgment-free tasks and hands back the time those tasks were eating, so a team can spend it where a context still has to be read. The right question to ask before launching any tool: among our current tasks, which ones truly call for no decision on our part?
If you’re unsure what deserves automating in your sales processes—and what’s better left in human hands—get in touch: it’s often by sorting the tasks that the real seam comes into view.
