Very little writing about robots discusses stopping. Yet a meaningful share of projects should stop, and stopping after three months is far better than sustaining something for two years to avoid admitting it did not work.
Signals a project is declining
Caught early it is recoverable; caught late only stopping remains.
Interaction numbers falling steadily month on month. After the novelty period the figure should stabilise, not continue declining.
Nobody updating content. The earliest and clearest signal. Static content means nobody considers the robot theirs.
Staff avoiding it or positioning themselves in front of it. There is usually a real reason worth asking about.
The robot flat for days without anyone noticing.
Nobody asking about the numbers. It has left leadership attention.
It has been moved somewhere quieter. Usually the penultimate step.
The unanswered-question list keeps growing. Nobody is working it.
The first three appear earliest. Seeing all three leaves a few weeks to intervene before the situation becomes hard to reverse.
Before stopping: try changing direction
Many projects are not fundamentally broken but doing the wrong job or standing in the wrong place.
Move it. The cheapest action with the largest effect. If interaction numbers are low, try this first.
Change the job. A robot greeting visitors ineffectively might work internally — guiding new staff, moving items, patrolling in the evening.
Change the hours. It may be useful at quiet times rather than peak times.
Reinvest in content. If the problem is poor answers, this is where money should go rather than anywhere else.
Change the owner. If the current one lacks time or fit.
Narrow the scope. Have it do less but do it well. A robot answering five questions correctly is more useful than one attempting everything and getting half wrong.
Give each change a defined period — usually six to eight weeks — then reassess.
Deciding objectively
Difficult because money and reputation are attached.
Ignore what has been spent. That money is gone whether you continue or stop. The right question is whether, from today, continuing creates more value than the cost of maintaining it.
Cost the real maintenance. Owner time, content, servicing, power. Compare against the value currently produced.
Ask an outsider. Someone not involved in the original decision sees it more clearly.
Use the criteria written beforehand. If criteria were set at the start, this is when they earn their keep. If not, that is a lesson for next time.
Separate the decision from the people. A project not working does not mean someone did something wrong. If stopping is read as assigning blame, nobody will ever propose it.
Set a decision deadline. Review on a specific date. Without one, projects simply drift.
If the decision is to stop
Do it cleanly and extract what value remains.
Export all data first. Conversation history, the list of questions customers asked, contacts captured. This is an asset with value independent of the robot.
Record the lessons. What worked, what did not, what would be done differently. One or two pages, filed.
Tell the staff. With the reason. Letting the robot silently disappear creates a poor impression of how the organisation makes decisions.
Deal with the equipment. Resell, sublet, transfer to another department, or return if rented. Service robots retain some value while under warranty and supported.
Keep the content. The knowledge base built is usable for a website chatbot, for training material, for a customer support channel. This is the largest remaining asset.
Do not treat it as total failure. The organisation learned how to work with this technology, learned what customers ask, and learned where it was not ready.
Returning later
Stopping a project does not close the door permanently.
What may change. Technology improves, prices fall, new products appear. A problem unsolvable today may be solvable in two years.
What to maintain meanwhile. Keep the knowledge base current. Keep procedures documented. Keep tracking what customers ask. These start the next attempt considerably further along.
Track the market lightly. No investment needed, just awareness of what is new.
Talk to organisations that succeeded. After trying and failing, your questions become specific and far more useful.
Reread your recorded lessons. Before starting again.
Many organisations succeed on the second attempt precisely because the first taught them something concrete — and that only holds if the lessons were written down rather than allowed to disappear with the project.
Frequently asked questions
What is the earliest signal a project is declining?
Nobody updating the content. Combined with steadily falling interaction numbers and staff avoiding or standing in front of the robot, these three appear earliest and leave a few weeks to intervene.
What should be tried before stopping?
Moving it is the cheapest action with the largest effect. Then changing the job, changing the hours, reinvesting in content, changing the owner, or narrowing the scope so it does less but does it well.
How do you decide objectively?
Ignore what has already been spent since it is gone either way, then ask whether from today continuing creates more value than the cost of maintaining it. And separate the decision from any individual's performance.
What value remains after stopping?
The knowledge base built — usable for a website chatbot, training material or a support channel — plus the data on what customers actually ask and the written lessons for a future attempt.
More in Robots for real estate and showrooms and Reception and event robots.
