Most writing about robots concerns which one to buy. This concerns when to buy none — because that is also an answer, and for a substantial number of situations it is the right one.
Five signals that it is premature
You cannot state specifically what it will do. If the answer is to be more modern or to keep up, the requirement has not been identified yet.
Nobody is named to look after it. Charging, cleaning, content, unjamming. Without a specific person, it will be unused within months.
The task occurs occasionally rather than constantly. An idle robot is a forgotten robot, and idle is the normal state for infrequent tasks.
The space does not physically suit it. Narrow passages, steps, confined layout. This is a hard constraint that software does not solve.
Something more urgent is unresolved. Cash flow, product quality, staffing gaps. A robot does not address these and it consumes resources that would.
Two of these five warrants pausing. Three or more and the money almost certainly produces more elsewhere.
Reasons for buying that deserve scrutiny
Common enough to name directly.
A competitor has one. Different premises, different customers, different operation. What works there may not transfer, and the visible part of their deployment is not the part that determines whether it works.
Something seen online. Filmed in ideal conditions and usually the successful attempt among many.
A discount is available. A reduced price for something not needed is still money spent.
To appear modern. Legitimate if budgeted as marketing and measured with marketing metrics. Not legitimate if you convince yourself it will reduce operating cost.
Staff are hard to hire. A robot replaces a narrow slice of a role. If the problem is insufficient people, the solution is about staffing rather than equipment.
You have already told people you would. The worst reason and a real one.
What to do instead
Actions that typically produce more, for less.
Measure where the time actually goes. One week of recording. The result usually differs from the impression and it points precisely at where intervention would help.
Do the software before the hardware. Enquiry handling, scheduling, management systems. Cheaper by an order of magnitude and addressing problems that genuinely exist.
Rearrange the space. A surprising share of flow and waiting problems resolve by moving things.
Write down the process. Who does what, when. Many operations lack this, and its absence is the source of much of the disorder that automation is expected to fix.
Ask the staff. Which tasks feel like wasted effort. The answers are usually direct and cheap to act on.
Improve what the customer notices. Product, space, service. These retain customers, which is a more fundamental problem than efficiency for most businesses.
If you still want to test the idea
There is a low-risk way.
Rent for a month. A fraction of purchase cost and it answers the question with your own customers in your own space.
Time it to a peak period. When the value would be most evident.
Write criteria first. After one month, which number indicates it is worth buying. Written before the rental begins.
Measure. Interactions, staff time saved, new customers, whatever the objective is.
Ask staff and customers afterwards.
Accept the result. If the figures say it is not worthwhile, return it. A month of rental is inexpensive tuition.
This converts a large irreversible decision into a small reversible experiment, which for anyone genuinely uncertain is the sensible move.
And the signals that it is the right time
For balance, the opposite list.
You can state the task specifically, and it repeats many times daily.
You have measured how long it currently takes. A baseline exists to compare against.
Someone is named to own it, and knows they will.
The space has been checked physically, with a trolley along the intended route.
Cheaper alternatives have been tried and found insufficient. Kiosk, software, rearrangement.
The budget covers running costs, not only purchase.
You have visited a working installation.
Meeting six of these seven puts the decision on solid ground, and the success rate for deployments made this way is markedly different from those made on enthusiasm.
Frequently asked questions
What most clearly signals a premature purchase?
Not being able to state specifically what the robot will do, and having nobody named to look after it. Both almost always end with the equipment unused within months.
Which reason for buying deserves the most scrutiny?
That a competitor has one. Different premises, customers and operations mean what works there may not transfer, and the visible part of their deployment is not the part determining whether it works.
What should be done before considering equipment?
Measure for one week where the time actually goes, do the software before the hardware, rearrange the space, write down the process, and ask staff which tasks feel like wasted effort.
How can the idea be tested cheaply?
Rent for a month during a peak period, with success criteria written before starting, then accept the result. This converts a large irreversible decision into a small reversible experiment.
More in Humanoids at work and Voice and language.
