MysteryShopper.gov
Using AI to insert a rocket where required
Back when the mighty northern glaciers were freshly melted, I had a casual job in retail. There were three main parts to my job – register, stocking shelves and moving boxes out the back. This list is in order from my least to most favourite.
I actually enjoyed moving boxes and getting sweaty. On register I was socially awkward, shy of my savage pimples under the glaring fluorescent tube lights, and I made mistakes.
Adding to this pressure was the possibility of receiving a Mystery Shopper. If you haven’t done retail, this is a person paid by the company to act like a shopper in order to ensure they are offered a basket if they are holding two or more items, are addressed by name if they use a credit card, that change is counted back, and of course that they get a big shit-eating grin from the poor spotty register monkey.
I never ended up getting a review from a mystery shopper but it was yet another source of terror for a timid young man.
Nevertheless, if you’ve ever had trouble at a shop or restaurant, you can see how mystery shoppers can be a good thing if properly deployed.
Today’s idea is a DOGE-like concept for mystery shoppers to analyze government services.
Town Hall mystery shopper
The first type of mystery shopper would be an ordinary person, ideally with limited education and experience with bureaucracy, sent into a government office to attempt an ordinary transaction.
The person would wear a hidden camera and microphone.
It would be important to continually rotate mystery shoppers so that none become familiar.
An AI system would analyze the recorded data for basics like time spent, tasks completed, fees paid, bribes paid, tone of voice, assistance received, and so on. A few human experts could also review as required.
Feedback would be more carrot than stick. A manager of a branch that receives good reviews could expect bonuses and promotions. There might also be prizes for staff – all staff at the branch, not just those who happened to deal with the mystery shopper on the day.
Everyday problems could be dealt with in an anonymized way. For example, if the process took too long, a regional manager might be sent out to help streamline the system to get it up to speed with other branches.
After this, there would be future chances to win those yummy bonuses if improvement is achieved.
Reprimands for individuals would only be given in extreme circumstances such as lying (“We can’t do that at this branch, sir!”), failing to complete assigned tasks, allowing vital applications and so on to fall into a black hole, rudeness, and demanding/accepting bribes.
In some countries these are not ‘extreme’, but they soon will be once my mystery shoppers are rolled out.
Overall, the system should act as a de facto pay rise for the public service as they could expect regular bonuses so long as they are doing their jobs.
The cost would be worth it for the increase in government efficiency. Imagine how much more tax would be generated if business could actually get things done without all the red tape – that is, with civil servants being paid to make things happen, and fast.
Of course, a few mystery shoppers should also have an unacceptable proposal (unsound septic tank design proposal, etc.) to check that things are rejected as required.
Currently, this is technically possible or soon will be.
In the future, we might add in mini-drones or bots monitoring .gov devices for further data.
As for privacy, have government employees sign a contract agreeing to be monitored in this way.
Online mystery shopper
Another type of mystery shopper would be set the task of trying to complete an online application for a government service. This could be a bot instructed to mimic how a human would attempt the process if they had limited education.
As with the face-to-face mystery shopper, the online bot would record time taken and make note of any unclear instructions, hard-to-find pages, doomloops, and eternal error messages or other problems.
The bot could also make recommendations on how essential each required document and step really is.
Web designers and departmental heads could look forward to rewards for implementing systems that work well.
Bonus points could be given for government services that can communicate with each other, especially in a quick, automated way that saves the user time. For example, a passport uploaded as ID for a drivers license should automatically be verified by the department that issued it.
The overall effect should be to optimize the process by minimizing unnecessary human involvement. Human time would be better devoted to marginal or unusual cases that require judgement – the kind of cases that present bureaucrats hate and reflexively reject without thought so that they don’t get into trouble.
Giving such cases due considerations would become the main part of their job.
I guess this process might lead to fewer staff or to more staff doing nothing, but even the latter would be preferable to them doing something and getting in the way (as seems to happen at present).
Independent agency
Rather than departments assessing themselves, the mystery shoppers would be deployed by a government efficiency agency like DOGE. This would help guarantee independence.
Of course, this agency itself would require feedback and review. Perhaps it might be assessed by surveys of real citizens on how satisfied they are with government service provision.
International comparison
When assessing anything, you need something to compare it against. Otherwise you’ll have no way of knowing, for example, how long is a reasonable time for someone to get a building permit or a business licence.
Each country participating in this scheme could assign a bot to consult with other bots around the world. They could produce tables comparing relative performance across all areas, perhaps by income level or geographic region.
If you can sign your new employee up for tax, pension, insurance etc. in 20 minutes in Denmark, maybe it can also be done that fast in New Zealand. Maybe a bit longer for poor countries with less advanced IT.
The bots could also share tips on how the world’s best practice nations manage to do it.
Experience
Mystery shoppers like these have been tried, with mixed results.
Not surprisingly, many public employees hated them, seeing them as spies or agents of vindictive bosses.
A few programs used hidden cameras but most were trained people, making the process slower and more expensive.
A lot of the efforts were disjointed and temporary.
The feedback was often focused on the negative, sparking union and departmental resistance.
To overcome these problems, the mystery shopper program should be:
Broad, covering all areas of government except sensitive areas where other approaches are required
Focused on positives, with bonuses for satisfactory performance
Permanent
Automated to reduce costs
Final thougths
Of course, another reason some departments wouldn’t like this is that they profit from being inefficient and corrupt.
In less developed countries, such a program would interrupt cozy patronage networks.
At present, extractive states use the public service to take money from the people and distribute it to the friends, family and supporters of those in power. No doubt they would attempt to subvert the mystery shopper program to extend rather than break apart these patronage networks.
However, I offer it as a potential tool for any strong leader who wants to whip the public service into shape, like a somewhat smarter Duterte. If one or two heroes succeed, others may follow.
Once it’s been designed, there might be a cheap, off-the-plan version that works anywhere.
Rather than a future world ruled by AI, we should use it cleverly to fix age-old problems in cost-effective ways.
Worth a shot, hey?