On this page
Connie and I have been out kicking through the leaves this weekend. As another year marches towards its close, I wanted to share something we've been thinking about and debating for a long time.
We could have built SAI Revisit in months, but we took years. This is why.
Anyone who has worked in retail knows that a small number of repeat offenders cause most of the problems in store, both the losses and the threats, abuse and violence that colleagues face. Behind every one of those incidents is a person, a colleague who should be able to work safely or a customer who should be able to shop without feeling intimidated.
Over the last few years, I have spent a lot of time talking to retailers about colleague safety, organised retail crime and repeat offenders. At the same time, SAI has been helping retailers identify theft, violence and aggression in real time. I have also watched hundreds of hours of video from our platform, and we kept seeing the same people coming back. It is distressing to see.
The question for us has always been, could we help retailers deal with prolific offenders in a way that was fair, proportionate and useful? Technically, we knew we could easily build it. The harder question was whether we should.
Som Sinha, Abhijit Sanyal, Chris Bell and I returned to that question many times over the years.
What about someone stealing food because they cannot feed their family? What about a teenager who makes one mistake? What about someone who has turned their life around? Those questions shaped the system we eventually built.
So we waited, talked it through and decided on three principles we would not compromise on.
Proportionality
SAI Revisit should only be used where there is clear evidence of repeated and serious offending. Not suspicion, not one isolated incident.
Due Process
AI should not decide that someone belongs on a watchlist. That decision should be based on evidence and human review.
The Right to Challenge
If someone is challenged in a store, there should already be evidence explaining why. The technology is not the evidence; it simply helps spot someone where the evidence already exists.
Those principles became the foundation of SAI Revisit.
So Why Now?
For a long time, we did not think we had the right combination of process and technology, so we worked on both. We built the process around how someone could be added to a watchlist, what evidence was needed and how decisions would be reviewed. We worked with Hannah Crowther, Privacy and AI Lawyer at Bristows, to challenge our approach and make sure privacy, proportionality and due process were built in from the start.
At the same time, SAI One meant we could see the full contextual picture of what was happening in store, which is how we always believed it should be done.
Eventually, the pieces came together. We had the process, the safeguards and the technology. That was when we were confident the system was proportionate and the time was right, and we started working with a leading UK retailer.
We've also made a commitment: revenue from SAI Revisit will help fund the Safer Stores Initiative. Technology like this may have a role to play, but it is only one part of the answer. Education and rehabilitation matter just as much. Lots of people are already working hard on this, and through Safer Stores we hope to play our part alongside them.
We could have built SAI Revisit sooner, but we chose not to. We kept working on the technology, the process and the legal safeguards until we were comfortable that the approach was proportionate. It has taken some time, but I think it was the right thing to do.
If you're a retailer thinking about these questions, or you see it differently, I'd genuinely like to hear from you.
Helen Wilde
Co-founder and CCO, SAI Group
