---
title: "Transcript: Shipping Production AI Inside Government — William Tarr, Ministry of Justice (DO NOT PUBLISH)"
category: "transcripts"
videoId: "qlHaO6laBlM"
sourceLabels: ["YouTube transcript", "Cached transcript markdown"]
wordCount: "11459"
---

# Transcript: Shipping Production AI Inside Government — William Tarr, Ministry of Justice (DO NOT PUBLISH)

## Source Video
- [YouTube](https://www.youtube.com/watch?v=qlHaO6laBlM)

## Local Cache
- `raw/sources/youtube-transcripts/qlHaO6laBlM.txt`
- 11,459 words

## Transcript

Hello. Hello. Um I'm here Sorry to bring the tone down Um I'm here Sorry to bring the tone down Um I'm here Sorry to bring the tone down to about government. Um so I'm what's to about government. Um so I'm what's to about government. Um so I'm what's called a forward deployed engineer in called a forward deployed engineer in called a forward deployed engineer in the UK government. So I'm currently the UK government. So I'm currently the UK government. So I'm currently working in the Ministry of Justice in working in the Ministry of Justice in working in the Ministry of Justice in the Justice AI unit. Uh some of you the Justice AI unit. Uh some of you the Justice AI unit. Uh some of you might have heard of it. It's It's the might have heard of it. It's It's the might have heard of it. It's It's the closest thing you can find to a startup closest thing you can find to a startup closest thing you can find to a startup in government, basically. Uh before in government, basically. Uh before in government, basically. Uh before this, I was at Harvard, dropped out, had this, I was at Harvard, dropped out, had this, I was at Harvard, dropped out, had a YC company. And then while I was a YC company. And then while I was a YC company. And then while I was there, I was reading the Times, as one there, I was reading the Times, as one there, I was reading the Times, as one does, and there was a puzzle in the does, and there was a puzzle in the does, and there was a puzzle in the Times. It was like a math problem. Times. It was like a math problem. Times. It was like a math problem. Uh and this, after solving it, led me to Uh and this, after solving it, led me to Uh and this, after solving it, led me to apply for apply for apply for the number 10 innovation fellowship, the the number 10 innovation fellowship, the the number 10 innovation fellowship, the leader of which is is over here, Owen, leader of which is is over here, Owen, leader of which is is over here, Owen, who's talking tomorrow, so I recommend who's talking tomorrow, so I recommend who's talking tomorrow, so I recommend everyone to go and watch that. everyone to go and watch that. everyone to go and watch that. Um from here, I ended up in the Justice Um from here, I ended up in the Justice Um from here, I ended up in the Justice AI unit, so I was actually seconded AI unit, so I was actually seconded AI unit, so I was actually seconded there. there. there. Uh and um thoroughly impressed by by Uh and um thoroughly impressed by by Uh and um thoroughly impressed by by what's going on there. So I joined in what's going on there. So I joined in what's going on there. So I joined in November, I've been here for roughly 6 November, I've been here for roughly 6 November, I've been here for roughly 6 months. Um months. Um months. Um and our our key principle is that we are and our our key principle is that we are and our our key principle is that we are shipping AI products. Um not these shipping AI products. Um not these shipping AI products. Um not these pilots and slides that government is pilots and slides that government is pilots and slides that government is used to with bloated bureaucracy and used to with bloated bureaucracy and used to with bloated bureaucracy and infrastructure. We're building things infrastructure. We're building things infrastructure. We're building things people actually want to use. people actually want to use. people actually want to use. So when I first joined, this was the So when I first joined, this was the So when I first joined, this was the headlines every day. headlines every day. headlines every day. Prisoners being released. Prisoners being released. Prisoners being released. Um for some reason, Wandsworth. You can Um for some reason, Wandsworth. You can Um for some reason, Wandsworth. You can see them walking out, the news cameras see them walking out, the news cameras see them walking out, the news cameras are stationed there 24/7 watching people are stationed there 24/7 watching people are stationed there 24/7 watching people being released, and there was a good being released, and there was a good being released, and there was a good chance they weren't meant to be chance they weren't meant to be chance they weren't meant to be released. released. released. Uh so this is my first introduction to Uh so this is my first introduction to Uh so this is my first introduction to um why don't things work in government. um why don't things work in government. um why don't things work in government. Um so looking at all of these Looking at Um so looking at all of these Looking at Um so looking at all of these Looking at the systems that guide government, it the systems that guide government, it the systems that guide government, it was pretty clear. You have a combination was pretty clear. You have a combination was pretty clear. You have a combination of like terrible factors. You have of like terrible factors. You have of like terrible factors. You have disjointed systems, like different disjointed systems, like different disjointed systems, like different numbers for prisoners and courts and numbers for prisoners and courts and numbers for prisoners and courts and probation and uh and and in prison probation and uh and and in prison probation and uh and and in prison themselves. themselves. themselves. Poor organization. So there's a lot of Poor organization. So there's a lot of Poor organization. So there's a lot of people who were working within the people who were working within the people who were working within the end-to-end system didn't even know each end-to-end system didn't even know each end-to-end system didn't even know each other, didn't know the process. other, didn't know the process. other, didn't know the process. Uh bad contracts. You have layers and Uh bad contracts. You have layers and Uh bad contracts. You have layers and layers and layers of government tech layers and layers of government tech layers and layers of government tech that have been bought over the years, that have been bought over the years, that have been bought over the years, layering on one another with dodgy layering on one another with dodgy layering on one another with dodgy connections, with weird APIs, and with connections, with weird APIs, and with connections, with weird APIs, and with weird governance between them. weird governance between them. weird governance between them. Uh and uh legacy tech that they work Uh and uh legacy tech that they work Uh and uh legacy tech that they work together. together. together. So why do we exist? Um about a year and So why do we exist? Um about a year and So why do we exist? Um about a year and a half ago, we didn't exist. Uh and then a half ago, we didn't exist. Uh and then a half ago, we didn't exist. Uh and then Dan James, our director, so I'm deputy Dan James, our director, so I'm deputy Dan James, our director, so I'm deputy director. Dan James director. Dan James director. Dan James uh uh uh started this about a year and a half ago started this about a year and a half ago started this about a year and a half ago with just one person in the Ministry of with just one person in the Ministry of with just one person in the Ministry of Justice, you know, fighting a lot of Justice, you know, fighting a lot of Justice, you know, fighting a lot of bureaucracy to get this done. And now bureaucracy to get this done. And now bureaucracy to get this done. And now we're about 40 people. And this This is we're about 40 people. And this This is we're about 40 people. And this This is a team of 40 people who in government a team of 40 people who in government a team of 40 people who in government probably require usually about 300 probably require usually about 300 probably require usually about 300 people, given the extent of our output. people, given the extent of our output. people, given the extent of our output. Um as it says, the traditional model is Um as it says, the traditional model is Um as it says, the traditional model is centralized and often doesn't find work. centralized and often doesn't find work. centralized and often doesn't find work. So a lot of a lot of uh So a lot of a lot of uh So a lot of a lot of uh uh tech that's built in the government uh tech that's built in the government uh tech that's built in the government is so disjointed from people on the is so disjointed from people on the is so disjointed from people on the front line. People who build tech for front line. People who build tech for front line. People who build tech for prisoners, prison officers, probation prisoners, prison officers, probation prisoners, prison officers, probation officers rarely see prison officers, officers rarely see prison officers, officers rarely see prison officers, rarely interact with the system itself, rarely interact with the system itself, rarely interact with the system itself, and it's a it's created all this like and it's a it's created all this like and it's a it's created all this like kind of mess of technology. kind of mess of technology. kind of mess of technology. Uh we're here to close the gap on that. Uh we're here to close the gap on that. Uh we're here to close the gap on that. So the Ministry of Justice is now 95,000 So the Ministry of Justice is now 95,000 So the Ministry of Justice is now 95,000 people, which is a huge organization. people, which is a huge organization. people, which is a huge organization. About 30,000 of those are About 30,000 of those are About 30,000 of those are like operational people, so people that, like operational people, so people that, like operational people, so people that, you know, officers themselves. you know, officers themselves. you know, officers themselves. So what makes us different? We have So what makes us different? We have So what makes us different? We have we've adopted this forward deployed we've adopted this forward deployed we've adopted this forward deployed model. So a lot of my work is myself model. So a lot of my work is myself model. So a lot of my work is myself going into prison, sitting with the going into prison, sitting with the going into prison, sitting with the officers themselves, watching them work, officers themselves, watching them work, officers themselves, watching them work, not kind of road maps Q1 to Q2, Q3, not kind of road maps Q1 to Q2, Q3, not kind of road maps Q1 to Q2, Q3, sitting there asking what are your sitting there asking what are your sitting there asking what are your problems, watching them work, what problems, watching them work, what problems, watching them work, what hasn't been fixed in this many years, hasn't been fixed in this many years, hasn't been fixed in this many years, what did you just to do. So we build what did you just to do. So we build what did you just to do. So we build MVPs in It says weeks here, but I I MVPs in It says weeks here, but I I MVPs in It says weeks here, but I I think it's probably more like days. Uh think it's probably more like days. Uh think it's probably more like days. Uh we get things working, we show the we get things working, we show the we get things working, we show the officers on site, we get them seeing officers on site, we get them seeing officers on site, we get them seeing what we can do with AI. what we can do with AI. what we can do with AI. Uh we work again directly in for Uh we work again directly in for Uh we work again directly in for prisons, probation, and courts. Louis prisons, probation, and courts. Louis prisons, probation, and courts. Louis here is mainly in probation. here is mainly in probation. here is mainly in probation. I'm mainly in prisons. I'm mainly in prisons. I'm mainly in prisons. Um again, we have this a mostly Um again, we have this a mostly Um again, we have this a mostly engineering team, so we have engineering team, so we have engineering team, so we have we have a professor of sociology in our we have a professor of sociology in our we have a professor of sociology in our team who's there to kind of think about team who's there to kind of think about team who's there to kind of think about the future of the MOJ. We're very the future of the MOJ. We're very the future of the MOJ. We're very forward looking. We think What will the forward looking. We think What will the forward looking. We think What will the MOJ What will justice look like in 20 MOJ What will justice look like in 20 MOJ What will justice look like in 20 years' time, 10 years' time? years' time, 10 years' time? years' time, 10 years' time? Uh we have built this messy environment. Uh we have built this messy environment. Uh we have built this messy environment. So we have all of our builders come from So we have all of our builders come from So we have all of our builders come from uh an array of places. Me from YC. We uh an array of places. Me from YC. We uh an array of places. Me from YC. We have startup people. We have people from have startup people. We have people from have startup people. We have people from Palantir, people from lots of different Palantir, people from lots of different Palantir, people from lots of different companies with different mindsets. companies with different mindsets. companies with different mindsets. Okay, now pass it to Louis. Okay, now pass it to Louis. Okay, now pass it to Louis. Sweet. So whilst I was in the prisons Sweet. So whilst I was in the prisons Sweet. So whilst I was in the prisons context, I was sitting in probation. I'm context, I was sitting in probation. I'm context, I was sitting in probation. I'm Louis Hogarth, an AI product builder. Louis Hogarth, an AI product builder. Louis Hogarth, an AI product builder. So our mission was quite simple. We know So our mission was quite simple. We know So our mission was quite simple. We know there's roughly 12,000 probation there's roughly 12,000 probation there's roughly 12,000 probation officers who join this work because officers who join this work because officers who join this work because they're really excited to do something they're really excited to do something they're really excited to do something impactful with the front line, impactful with the front line, impactful with the front line, face-to-face contact with people who face-to-face contact with people who face-to-face contact with people who have come out of prison needing help have come out of prison needing help have come out of prison needing help reoffending. reoffending. reoffending. Our mission was quite simple. Our mission was quite simple. Our mission was quite simple. Feel their pain, and ultimately do Feel their pain, and ultimately do Feel their pain, and ultimately do whatever it takes to build something whatever it takes to build something whatever it takes to build something they need, something they want. they need, something they want. they need, something they want. So in our context, what normally we've So in our context, what normally we've So in our context, what normally we've seen government for a digital solution, seen government for a digital solution, seen government for a digital solution, you have a team of roughly you have a team of roughly you have a team of roughly 40 people, uh mostly different roles, 40 people, uh mostly different roles, 40 people, uh mostly different roles, different types of people. different types of people. different types of people. Me and one other engineer built within Me and one other engineer built within Me and one other engineer built within weeks MVP, within months the pilot, and weeks MVP, within months the pilot, and weeks MVP, within months the pilot, and then we finished our national rollout then we finished our national rollout then we finished our national rollout within a few months. This is something within a few months. This is something within a few months. This is something that's unheard of in government, but it that's unheard of in government, but it that's unheard of in government, but it was simply using the best practice in was simply using the best practice in was simply using the best practice in the context that most needs it. Falling the context that most needs it. Falling the context that most needs it. Falling in love with the problem, not the in love with the problem, not the in love with the problem, not the product. And in our context, uh we spend product. And in our context, uh we spend product. And in our context, uh we spend two, three days a week on site with two, three days a week on site with two, three days a week on site with users. users. users. So we look at the feedback, we look to So we look at the feedback, we look to So we look at the feedback, we look to the things people love and hate about the things people love and hate about the things people love and hate about the product, and we also find the things the product, and we also find the things the product, and we also find the things that people have for new ideas. We go on that people have for new ideas. We go on that people have for new ideas. We go on site, build with them. If it looks good, site, build with them. If it looks good, site, build with them. If it looks good, we feature flag it. If they don't find we feature flag it. If they don't find we feature flag it. If they don't find it works well at a little bit of scale, it works well at a little bit of scale, it works well at a little bit of scale, yeah, we kill code, whatever it takes. yeah, we kill code, whatever it takes. yeah, we kill code, whatever it takes. Um and yeah, based on real feedback at Um and yeah, based on real feedback at Um and yeah, based on real feedback at pace in a way that's most useful. pace in a way that's most useful. pace in a way that's most useful. Oh, have I skipped a slide? Oh, have I skipped a slide? Oh, have I skipped a slide? After you're done. After you're done. After you're done. Um yeah, no, he's right. So So my Um yeah, no, he's right. So So my Um yeah, no, he's right. So So my actually my first second day of being in actually my first second day of being in actually my first second day of being in the team, I was in a probation office, the team, I was in a probation office, the team, I was in a probation office, and on the fourth day, I was in a and on the fourth day, I was in a and on the fourth day, I was in a prison. So we get engineers who like prison. So we get engineers who like prison. So we get engineers who like mainly have been at desks their entire mainly have been at desks their entire mainly have been at desks their entire life, and we get them in these places. life, and we get them in these places. life, and we get them in these places. So this is my ugly face in front of a So this is my ugly face in front of a So this is my ugly face in front of a prison uh holding the keys that I just prison uh holding the keys that I just prison uh holding the keys that I just got, so I can now walk around freely in got, so I can now walk around freely in got, so I can now walk around freely in a prison. I don't know who allowed that a prison. I don't know who allowed that a prison. I don't know who allowed that to happen. Uh but it's Yeah, that's so I to happen. Uh but it's Yeah, that's so I to happen. Uh but it's Yeah, that's so I can now walk around the prison, talk to can now walk around the prison, talk to can now walk around the prison, talk to staff, go into the wing, see how they're staff, go into the wing, see how they're staff, go into the wing, see how they're doing, see what they want, build doing, see what they want, build doing, see what they want, build products for them on site, which is kind products for them on site, which is kind products for them on site, which is kind of a a surprise to them as it is to as of a a surprise to them as it is to as of a a surprise to them as it is to as it is to me. it is to me. it is to me. Um the environment in a prisons is is Um the environment in a prisons is is Um the environment in a prisons is is like very different to what you expect, like very different to what you expect, like very different to what you expect, and also to what like the MOJ as a whole and also to what like the MOJ as a whole and also to what like the MOJ as a whole expects. And all these policy documents expects. And all these policy documents expects. And all these policy documents that outline perfect you best use and that outline perfect you best use and that outline perfect you best use and and perfect operation is just kind of and perfect operation is just kind of and perfect operation is just kind of uh tacitly followed in a prison. You uh tacitly followed in a prison. You uh tacitly followed in a prison. You know, these officers are doing whatever know, these officers are doing whatever know, these officers are doing whatever it takes to get the job done, and it's it takes to get the job done, and it's it takes to get the job done, and it's often very different to what's expected often very different to what's expected often very different to what's expected in the MOJ. So building our products is in the MOJ. So building our products is in the MOJ. So building our products is all about working around how they all about working around how they all about working around how they currently work and how they operate, currently work and how they operate, currently work and how they operate, rather than this kind of mytholized rather than this kind of mytholized rather than this kind of mytholized What's the word? Rather than this What's the word? Rather than this What's the word? Rather than this idealized version of what a prison how idealized version of what a prison how idealized version of what a prison how prison should run. prison should run. prison should run. Um Um Um Yeah, so we're seeing workflows, Yeah, so we're seeing workflows, Yeah, so we're seeing workflows, pressures, edge cases firsthand. A lot pressures, edge cases firsthand. A lot pressures, edge cases firsthand. A lot of the tech doesn't work. A lot of of the tech doesn't work. A lot of of the tech doesn't work. A lot of people using old tech, and you've seen people using old tech, and you've seen people using old tech, and you've seen this all firsthand. It's very shocking this all firsthand. It's very shocking this all firsthand. It's very shocking the first time you go there. the first time you go there. the first time you go there. Uh you can't design this stuff from a Uh you can't design this stuff from a Uh you can't design this stuff from a desk. Challenges. So So uh what What Challenges. So So uh what What challenges with government? Uh so being challenges with government? Uh so being challenges with government? Uh so being a startup in government comes with a lot a startup in government comes with a lot a startup in government comes with a lot of pressure uh pushing back on you. A of pressure uh pushing back on you. A of pressure uh pushing back on you. A lot of the governance is very outdated. lot of the governance is very outdated. lot of the governance is very outdated. Um so I I I I uh Um so I I I I uh Um so I I I I uh There's a difference in my mind between There's a difference in my mind between There's a difference in my mind between governance for policy governance for policy governance for policy and governance for products. and governance for products. and governance for products. All the governance we go through is All the governance we go through is All the governance we go through is designed for policy. Everything we designed for policy. Everything we designed for policy. Everything we decide to do, every product we decide to decide to do, every product we decide to decide to do, every product we decide to build is all about what's the impact build is all about what's the impact build is all about what's the impact it's going to have on the broader MOJ. it's going to have on the broader MOJ. it's going to have on the broader MOJ. It's long, It's long, It's long, you know, you know, you know, uh committees, long processes, months uh committees, long processes, months uh committees, long processes, months and months and months of chatting of and months and months of chatting of and months and months of chatting of people in offices who never been to people in offices who never been to people in offices who never been to prison in their lives talking about what prison in their lives talking about what prison in their lives talking about what we're doing. In my mind, we're building we're doing. In my mind, we're building we're doing. In my mind, we're building products for staff. We trust these staff products for staff. We trust these staff products for staff. We trust these staff to operate in a prison. We trust prison to operate in a prison. We trust prison to operate in a prison. We trust prison officers to make the best decision to officers to make the best decision to officers to make the best decision to use the tools they think are best for use the tools they think are best for use the tools they think are best for them. them. them. So in that way, the government's So in that way, the government's So in that way, the government's outdated. The government should reflect outdated. The government should reflect outdated. The government should reflect the way that these prison officers will the way that these prison officers will the way that these prison officers will use use these tools. use use these tools. use use these tools. Obviously, hesitancy from HQ. A lot of Obviously, hesitancy from HQ. A lot of Obviously, hesitancy from HQ. A lot of people in the MOJ and the government people in the MOJ and the government people in the MOJ and the government more broadly just don't understand the more broadly just don't understand the more broadly just don't understand the tech, or they have kind of a tech, or they have kind of a tech, or they have kind of a uh uh embedded like mistrust of it. uh uh embedded like mistrust of it. uh uh embedded like mistrust of it. A lot of people in government now who A lot of people in government now who A lot of people in government now who have added to their job titles AI or have added to their job titles AI or have added to their job titles AI or innovation who have rarely touched an AI innovation who have rarely touched an AI innovation who have rarely touched an AI tool in their lives. tool in their lives. tool in their lives. Uh and I think Uh and I think Uh and I think the worst of which is this this this gap the worst of which is this this this gap the worst of which is this this this gap I've been talking about. So a lot of I've been talking about. So a lot of I've been talking about. So a lot of There's a There's a huge distance There's a There's a huge distance There's a There's a huge distance between what people on the front line between what people on the front line between what people on the front line experience and what people in HQ and MOJ experience and what people in HQ and MOJ experience and what people in HQ and MOJ actually do. And we're here to bridge actually do. And we're here to bridge actually do. And we're here to bridge that. I'll jump in. I think I'm at Oh, that. I'll jump in. I think I'm at Oh, that. I'll jump in. I think I'm at Oh, that No, that works out. Yeah, stay that No, that works out. Yeah, stay that No, that works out. Yeah, stay there. So one of the things we found there. So one of the things we found there. So one of the things we found most effective, most effective, most effective, rather than getting into a position of rather than getting into a position of rather than getting into a position of having discussion about a theoretical having discussion about a theoretical having discussion about a theoretical product with theoretical thoughts on product with theoretical thoughts on product with theoretical thoughts on what that outcome might look like, the what that outcome might look like, the what that outcome might look like, the observable change in behavior. observable change in behavior. observable change in behavior. Because we've been prioritizing safe to Because we've been prioritizing safe to Because we've been prioritizing safe to fail piloting, testing on site, fail piloting, testing on site, fail piloting, testing on site, it means that we're going to these it means that we're going to these it means that we're going to these meetings, and we're not giving them an meetings, and we're not giving them an meetings, and we're not giving them an idea with something that we think, we idea with something that we think, we idea with something that we think, we give them something with a bit more give them something with a bit more give them something with a bit more clear signal. It's not the strongest clear signal. It's not the strongest clear signal. It's not the strongest signal until we get real users and real signal until we get real users and real signal until we get real users and real product giving us real data, for sure. product giving us real data, for sure. product giving us real data, for sure. But it's been something that's not been But it's been something that's not been But it's been something that's not been seen in this context. Rather than we go seen in this context. Rather than we go seen in this context. Rather than we go and say, That's dramatic. Um so yeah, in our That's dramatic. Um so yeah, in our context, we found it really effective to context, we found it really effective to context, we found it really effective to say to people, here's an idea. This is say to people, here's an idea. This is say to people, here's an idea. This is what we're learning so far. And now we what we're learning so far. And now we what we're learning so far. And now we can start navigating bureaucracy with can start navigating bureaucracy with can start navigating bureaucracy with ruthless efficiency. So we code fast, we ruthless efficiency. So we code fast, we ruthless efficiency. So we code fast, we also build fast in a way that's working also build fast in a way that's working also build fast in a way that's working across the system for governance across the system for governance across the system for governance ruthlessly efficiently. ruthlessly efficiently. ruthlessly efficiently. So in our context, one of the things So in our context, one of the things So in our context, one of the things that's been big on our radar, it'd be that's been big on our radar, it'd be that's been big on our radar, it'd be really stupid for us to exist and not really stupid for us to exist and not really stupid for us to exist and not have safety and security of the have safety and security of the have safety and security of the uh products in our minds from day one. uh products in our minds from day one. uh products in our minds from day one. So all of the basic best practice in So all of the basic best practice in So all of the basic best practice in terms of where's the data going, how we terms of where's the data going, how we terms of where's the data going, how we process it, it keeps us busy, process it, it keeps us busy, process it, it keeps us busy, but it's part of the national job. but it's part of the national job. but it's part of the national job. We work with people who are hungry to We work with people who are hungry to We work with people who are hungry to make a change, and we know that part of make a change, and we know that part of make a change, and we know that part of it, you have to make sure from the it, you have to make sure from the it, you have to make sure from the get-go, you can't be too scrappy in the get-go, you can't be too scrappy in the get-go, you can't be too scrappy in the context of doing something dangerous. context of doing something dangerous. context of doing something dangerous. So we engage with all sorts of people, So we engage with all sorts of people, So we engage with all sorts of people, be it unions, be it our ethics be it unions, be it our ethics be it unions, be it our ethics colleagues, uh be it policy as well. So colleagues, uh be it policy as well. So colleagues, uh be it policy as well. So from day one, we're finding by building from day one, we're finding by building from day one, we're finding by building something that people want and also something that people want and also something that people want and also being transparent in a really effective being transparent in a really effective being transparent in a really effective way, this helps us navigate quite a slow way, this helps us navigate quite a slow way, this helps us navigate quite a slow process previously. Yeah. I think also process previously. Yeah. I think also process previously. Yeah. I think also having, as we do, amazing platform having, as we do, amazing platform having, as we do, amazing platform engineers to basically bat away any any engineers to basically bat away any any engineers to basically bat away any any any questions or concerns from MOJ is is any questions or concerns from MOJ is is any questions or concerns from MOJ is is a really big thing. So, all of our tech a really big thing. So, all of our tech a really big thing. So, all of our tech is built on really strong foundations is built on really strong foundations is built on really strong foundations and all our data security is is is up to and all our data security is is is up to and all our data security is is is up to the scratch. Mhm. the scratch. Mhm. the scratch. Mhm. Let's have a look what's on this one. Let's have a look what's on this one. Let's have a look what's on this one. Yes, a feature of AI across justice. So, Yes, a feature of AI across justice. So, Yes, a feature of AI across justice. So, this is the exciting thing. This is this this is the exciting thing. This is this this is the exciting thing. This is this is the bit that's away from the the the is the bit that's away from the the the is the bit that's away from the the the boring bit about governance. So, so boring bit about governance. So, so boring bit about governance. So, so a lot of our job is thinking about what a lot of our job is thinking about what a lot of our job is thinking about what will justice be like in the future. We will justice be like in the future. We will justice be like in the future. We we we're here to basically empower our we we're here to basically empower our we we're here to basically empower our colleagues who are on the ground who are colleagues who are on the ground who are colleagues who are on the ground who are doing the jobs. We don't want prison doing the jobs. We don't want prison doing the jobs. We don't want prison officers at a desk typing away reports. officers at a desk typing away reports. officers at a desk typing away reports. We want them on the wing. We want them We want them on the wing. We want them We want them on the wing. We want them looking after prison prisoners. We want looking after prison prisoners. We want looking after prison prisoners. We want them looking after other colleagues. We them looking after other colleagues. We them looking after other colleagues. We want them doing the job they signed up want them doing the job they signed up want them doing the job they signed up to do. Same with probation. We need to do. Same with probation. We need to do. Same with probation. We need these officers talking to people rather these officers talking to people rather these officers talking to people rather than making notes or rather than doing than making notes or rather than doing than making notes or rather than doing these boring processes that take them these boring processes that take them these boring processes that take them away from the thing that they're there away from the thing that they're there away from the thing that they're there to do. to do. to do. So, obviously big part of that is So, obviously big part of that is So, obviously big part of that is transcription. It's kind of low-hanging transcription. It's kind of low-hanging transcription. It's kind of low-hanging fruit of government, which is which is fruit of government, which is which is fruit of government, which is which is saving many many hours as justice saving many many hours as justice saving many many hours as justice transcribers done. transcribers done. transcribers done. Uh then we go into kind of more Uh then we go into kind of more Uh then we go into kind of more broader things. So, for example, my broader things. So, for example, my broader things. So, for example, my colleague David over there is working on colleague David over there is working on colleague David over there is working on CCTV analytics and predictive CCTV analytics and predictive CCTV analytics and predictive intelligence. So, how can we use CCTV in intelligence. So, how can we use CCTV in intelligence. So, how can we use CCTV in prisons, which is currently for example prisons, which is currently for example prisons, which is currently for example quite a reactive measure, to to to build quite a reactive measure, to to to build quite a reactive measure, to to to build out a better uh analytics tool for us to out a better uh analytics tool for us to out a better uh analytics tool for us to use? use? use? Um of course, assistance and agents to Um of course, assistance and agents to Um of course, assistance and agents to to replace some of that paperwork, which to replace some of that paperwork, which to replace some of that paperwork, which is outdated and messy. is outdated and messy. is outdated and messy. Um Um better data integration. This is this is better data integration. This is this is better data integration. This is this is the releasing error problem. the releasing error problem. the releasing error problem. Uh education and learning. How can we Uh education and learning. How can we Uh education and learning. How can we empower our colleagues to to um empower our colleagues to to um empower our colleagues to to um to use AI in a in a proper way apart to use AI in a in a proper way apart to use AI in a in a proper way apart from these kind of from these kind of from these kind of uh uh boring AI seminars some of them are boring AI seminars some of them are boring AI seminars some of them are forced to go to. forced to go to. forced to go to. Um yes, okay. So, I I'm wrapping up Um yes, okay. So, I I'm wrapping up Um yes, okay. So, I I'm wrapping up here. So, if you want to do work in here. So, if you want to do work in here. So, if you want to do work in government, build it with users. So, we government, build it with users. So, we government, build it with users. So, we are a small embedded team. Four deployed are a small embedded team. Four deployed are a small embedded team. Four deployed engineers at the core and and uh the engineers at the core and and uh the engineers at the core and and uh the pace of our impact has been has been pace of our impact has been has been pace of our impact has been has been really huge. And this model does work. really huge. And this model does work. really huge. And this model does work. And importantly, we are hiring in just And importantly, we are hiring in just And importantly, we are hiring in just in the in the just say I unit. We're in the in the just say I unit. We're in the in the just say I unit. We're looking for four deployed engineers to looking for four deployed engineers to looking for four deployed engineers to work on site to work uh with with with work on site to work uh with with with work on site to work uh with with with the broader MOJ and to improve our the broader MOJ and to improve our the broader MOJ and to improve our improve our AI technology. So, here's my improve our AI technology. So, here's my improve our AI technology. So, here's my um QR code. LinkedIn's here. Uh and get um QR code. LinkedIn's here. Uh and get um QR code. LinkedIn's here. Uh and get in touch if you if you'd like to in touch if you if you'd like to in touch if you if you'd like to connect or chat about it or just have connect or chat about it or just have connect or chat about it or just have any questions or or suggestions. And any any questions or or suggestions. And any any questions or or suggestions. And any questions would be would be great now if questions would be would be great now if questions would be would be great now if anyone has any. Sorry. Sorry. Uh Okay. Um Okay. Um thanks, guys. I think this must be like thanks, guys. I think this must be like thanks, guys. I think this must be like a fresh a fresh a fresh a breath of fresh air in a lot of areas. a breath of fresh air in a lot of areas. a breath of fresh air in a lot of areas. Um Um how often do you find that you go out how often do you find that you go out how often do you find that you go out part of the organization and say part of the organization and say part of the organization and say How often do you go out to a part of the How often do you go out to a part of the How often do you go out to a part of the organization and say there's a great organization and say there's a great organization and say there's a great opportunity here if we added opportunity here if we added opportunity here if we added transcription or if we added analytics transcription or if we added analytics transcription or if we added analytics or if we were able to automate this or if we were able to automate this or if we were able to automate this workflow, but foundational workflow, but foundational workflow, but foundational technology is not there. Like the staff technology is not there. Like the staff technology is not there. Like the staff might not have laptops or they don't might not have laptops or they don't might not have laptops or they don't have have have or they have like a a legacy proprietary or they have like a a legacy proprietary or they have like a a legacy proprietary system or um so, you don't have the system or um so, you don't have the system or um so, you don't have the tools that are built on. Yeah, no, good tools that are built on. Yeah, no, good tools that are built on. Yeah, no, good question. So, that's that's definitely a question. So, that's that's definitely a question. So, that's that's definitely a a problem. So, I'll give an example. So, a problem. So, I'll give an example. So, a problem. So, I'll give an example. So, in in prisons, a lot of prisons don't in in prisons, a lot of prisons don't in in prisons, a lot of prisons don't have Wi-Fi. A lot of them are like big have Wi-Fi. A lot of them are like big have Wi-Fi. A lot of them are like big big Victorian buildings with kind of 6 big Victorian buildings with kind of 6 big Victorian buildings with kind of 6 7-ft walls. Wi-Fi doesn't travel through 7-ft walls. Wi-Fi doesn't travel through 7-ft walls. Wi-Fi doesn't travel through it. And also in some of them it's banned it. And also in some of them it's banned it. And also in some of them it's banned as well cuz you know, prison prisoners as well cuz you know, prison prisoners as well cuz you know, prison prisoners can use can use Wi-Fi. Um so, how do we can use can use Wi-Fi. Um so, how do we can use can use Wi-Fi. Um so, how do we build around that? And again, it's build around that? And again, it's build around that? And again, it's sitting with the staff. The people in HQ sitting with the staff. The people in HQ sitting with the staff. The people in HQ don't really know this stuff, right? We don't really know this stuff, right? We don't really know this stuff, right? We talk to them and we say you can use talk to them and we say you can use talk to them and we say you can use this, but they're like we don't have this, but they're like we don't have this, but they're like we don't have Wi-Fi. We wrote they wrote out this Wi-Fi. We wrote they wrote out this Wi-Fi. We wrote they wrote out this tool, we can't use it, but no one in MOJ tool, we can't use it, but no one in MOJ tool, we can't use it, but no one in MOJ HQ knows about that. So, we go there and HQ knows about that. So, we go there and HQ knows about that. So, we go there and we talk to the people on the front line we talk to the people on the front line we talk to the people on the front line and we know these problems. And so, for and we know these problems. And so, for and we know these problems. And so, for example, we've got offline mode to all example, we've got offline mode to all example, we've got offline mode to all the stuff we do in prison. You know, the stuff we do in prison. You know, the stuff we do in prison. You know, like you know, caching in the browser like you know, caching in the browser like you know, caching in the browser kind of thing. Um so, yeah, so it's a kind of thing. Um so, yeah, so it's a kind of thing. Um so, yeah, so it's a good it's a good point. I'm sure there's good it's a good point. I'm sure there's good it's a good point. I'm sure there's more information that Louie can attest more information that Louie can attest more information that Louie can attest to. to. to. >> Yeah, loads. So, we find if you look at >> Yeah, loads. So, we find if you look at >> Yeah, loads. So, we find if you look at like infrastructure of our technology like infrastructure of our technology like infrastructure of our technology across England and Wales, it's a bit of across England and Wales, it's a bit of across England and Wales, it's a bit of a mess. Like some places have good a mess. Like some places have good a mess. Like some places have good Wi-Fi. Some places you have to do the Wi-Fi. Some places you have to do the Wi-Fi. Some places you have to do the Lion King impression waiting for the Lion King impression waiting for the Lion King impression waiting for the Wi-Fi to connect. And it just keeps it Wi-Fi to connect. And it just keeps it Wi-Fi to connect. And it just keeps it simple in ours. We don't go that's a simple in ours. We don't go that's a simple in ours. We don't go that's a shame, we'll try again next time. We shame, we'll try again next time. We shame, we'll try again next time. We quickly pivot to be like, this feels quickly pivot to be like, this feels quickly pivot to be like, this feels like an on-prem context. And this is how like an on-prem context. And this is how like an on-prem context. And this is how we make it work. Or in our context, just we make it work. Or in our context, just we make it work. Or in our context, just fully offline. And this is when they fully offline. And this is when they fully offline. And this is when they have connectivity here here and here. have connectivity here here and here. have connectivity here here and here. It means that the job is quite fun in It means that the job is quite fun in It means that the job is quite fun in that way because it's not simply like a that way because it's not simply like a that way because it's not simply like a gnarly problem. It's a gnarly problem in gnarly problem. It's a gnarly problem in gnarly problem. It's a gnarly problem in a gnarly context. Like you go there and a gnarly context. Like you go there and a gnarly context. Like you go there and then you find something that's really then you find something that's really then you find something that's really unexpected. But yeah, unexpected. But yeah, unexpected. But yeah, keep it simple. Just pivot, test keep it simple. Just pivot, test keep it simple. Just pivot, test something new, and if that lands, something new, and if that lands, something new, and if that lands, perfect. And so, the last point on that perfect. And so, the last point on that perfect. And so, the last point on that as well is that is that is the is the as well is that is that is the is the as well is that is that is the is the reason that we we can solve these reason that we we can solve these reason that we we can solve these problems is and Louie's a great example problems is and Louie's a great example problems is and Louie's a great example of this is that of this is that of this is that I have officers in my teams like I have officers in my teams like I have officers in my teams like officers on the job. Louie has a million officers on the job. Louie has a million officers on the job. Louie has a million probation officers that have his teams probation officers that have his teams probation officers that have his teams and just chat to him during the day and just chat to him during the day and just chat to him during the day saying, "Oh, Louie, this is broken. Can saying, "Oh, Louie, this is broken. Can saying, "Oh, Louie, this is broken. Can you help me fix it?" And then we have you help me fix it?" And then we have you help me fix it?" And then we have this like this like this like instant fix these problems or these instant fix these problems or these instant fix these problems or these instant kind of responses and and we we instant kind of responses and and we we instant kind of responses and and we we know what's going on at any given time, know what's going on at any given time, know what's going on at any given time, which is quite rare in government. which is quite rare in government. which is quite rare in government. Anything else from anyone? Yeah. I have Anything else from anyone? Yeah. I have Anything else from anyone? Yeah. I have just a quick follow-up. Thank you. Just quick a follow-up on Thank you. Just quick a follow-up on that kind of scenario in that setting, that kind of scenario in that setting, that kind of scenario in that setting, particularly when you come to solve the particularly when you come to solve the particularly when you come to solve the problem, uh a lot of the environment may problem, uh a lot of the environment may problem, uh a lot of the environment may link into beyond just digital AI type of link into beyond just digital AI type of link into beyond just digital AI type of thing. In your FDT, do you have like thing. In your FDT, do you have like thing. In your FDT, do you have like more like you know, engineering type more like you know, engineering type more like you know, engineering type like electrician everything part of your like electrician everything part of your like electrician everything part of your again to actually go on site almost like again to actually go on site almost like again to actually go on site almost like work with as well as the digital work with as well as the digital work with as well as the digital deployment, but looking at some of the deployment, but looking at some of the deployment, but looking at some of the physical world fixing as well. Yeah, no, physical world fixing as well. Yeah, no, physical world fixing as well. Yeah, no, definitely, yeah. So, David's example of definitely, yeah. So, David's example of definitely, yeah. So, David's example of this so so, for example, CCTV is is is this so so, for example, CCTV is is is this so so, for example, CCTV is is is half like a hardware problem as well, half like a hardware problem as well, half like a hardware problem as well, right? Um right? Um right? Um Yeah, thinking about three solutions. Yeah, thinking about three solutions. Yeah, thinking about three solutions. The problem is a lot of this stuff is The problem is a lot of this stuff is The problem is a lot of this stuff is it's guided by contracts that we don't it's guided by contracts that we don't it's guided by contracts that we don't control. Uh and so, a lot of the control. Uh and so, a lot of the control. Uh and so, a lot of the hardware in for example, prisons are hardware in for example, prisons are hardware in for example, prisons are bought by either the prisons themselves bought by either the prisons themselves bought by either the prisons themselves or by regional bodies that manage those or by regional bodies that manage those or by regional bodies that manage those prisons. Some of these contracts are prisons. Some of these contracts are prisons. Some of these contracts are profoundly horrendous. Like they're profoundly horrendous. Like they're profoundly horrendous. Like they're poorly poorly uh negotiated or they're poorly poorly uh negotiated or they're poorly poorly uh negotiated or they're just kind of For example, there's a just kind of For example, there's a just kind of For example, there's a there's a there's providers that layer there's a there's providers that layer there's a there's providers that layer on multiple different types of software on multiple different types of software on multiple different types of software for the same type for the same camera for the same type for the same camera for the same type for the same camera and it's a mess. Um so, I mean that's a and it's a mess. Um so, I mean that's a and it's a mess. Um so, I mean that's a kind of a a longer-term problem for us kind of a a longer-term problem for us kind of a a longer-term problem for us to solve. At the moment, we're kind of to solve. At the moment, we're kind of to solve. At the moment, we're kind of getting the low-hanging fruit of the um getting the low-hanging fruit of the um getting the low-hanging fruit of the um of the of the kind of software uh AI uh of the of the kind of software uh AI uh of the of the kind of software uh AI uh side of things, but it's definitely side of things, but it's definitely side of things, but it's definitely something we're looking at. Um we're something we're looking at. Um we're something we're looking at. Um we're making progress in CCTV, for example. making progress in CCTV, for example. making progress in CCTV, for example. But that's where like in our context, But that's where like in our context, But that's where like in our context, we're like truly obsessed about we're like truly obsessed about we're like truly obsessed about problems, not products. Cuz in our problems, not products. Cuz in our problems, not products. Cuz in our context, we're like this is really cool, context, we're like this is really cool, context, we're like this is really cool, but but but it won't land. it won't land. it won't land. And in our context, we don't go sorry, And in our context, we don't go sorry, And in our context, we don't go sorry, we promised in a year's time we'll we promised in a year's time we'll we promised in a year's time we'll deliver that. We go, "Hey, we tried. deliver that. We go, "Hey, we tried. deliver that. We go, "Hey, we tried. This is what we encountered. But this is This is what we encountered. But this is This is what we encountered. But this is the big impact we want to make." Yeah. the big impact we want to make." Yeah. the big impact we want to make." Yeah. But yeah, cool question. Thank you. But yeah, cool question. Thank you. But yeah, cool question. Thank you. Probably one more if there's if there's Probably one more if there's if there's Probably one more if there's if there's one more. If not, we can wrap up. one more. If not, we can wrap up. one more. If not, we can wrap up. Yep. Um you're going into all these like Um you're going into all these like prisons, everything like that, talking prisons, everything like that, talking prisons, everything like that, talking to all these people. to all these people. to all these people. How much time are you actually spending How much time are you actually spending How much time are you actually spending maybe educating who you're talking to maybe educating who you're talking to maybe educating who you're talking to so that at least you're sort of talking so that at least you're sort of talking so that at least you're sort of talking the same language uh about what you want the same language uh about what you want the same language uh about what you want to build, how you're going to build it, to build, how you're going to build it, to build, how you're going to build it, >> Yeah. how you at least get them to the >> Yeah. how you at least get them to the >> Yeah. how you at least get them to the level Yeah. to talk with you, basically? level Yeah. to talk with you, basically? level Yeah. to talk with you, basically? Yeah, so like uh sometimes when I'm on Yeah, so like uh sometimes when I'm on Yeah, so like uh sometimes when I'm on site building new features or updating site building new features or updating site building new features or updating pre-existing product, sometimes their pre-existing product, sometimes their pre-existing product, sometimes their worst team is like IT support. It's worst team is like IT support. It's worst team is like IT support. It's quite sweet. They're like, "Oh, but my quite sweet. They're like, "Oh, but my quite sweet. They're like, "Oh, but my phone won't work." And I have to look phone won't work." And I have to look phone won't work." And I have to look and I go, "Oh, okay. So, this is how it and I go, "Oh, okay. So, this is how it and I go, "Oh, okay. So, this is how it makes sense." makes sense." makes sense." Um as a manager for the team, Um as a manager for the team, Um as a manager for the team, we've got staff who need to think more we've got staff who need to think more we've got staff who need to think more about the reason they joined, like the about the reason they joined, like the about the reason they joined, like the analytical part of the work. They don't analytical part of the work. They don't analytical part of the work. They don't need to think more about the how do I do need to think more about the how do I do need to think more about the how do I do a thing. a thing. a thing. So, we always optimize to have like the So, we always optimize to have like the So, we always optimize to have like the product two times simpler to make it 10 product two times simpler to make it 10 product two times simpler to make it 10 times better. times better. times better. So, all of the language we use, like So, all of the language we use, like So, all of the language we use, like we're using today, you would never say we're using today, you would never say we're using today, you would never say to a member of staff, "Oh, we're going to a member of staff, "Oh, we're going to a member of staff, "Oh, we're going to do something cool. We're going to to do something cool. We're going to to do something cool. We're going to cache it locally." And then you just cache it locally." And then you just cache it locally." And then you just means nothing to them. But we make sure means nothing to them. But we make sure means nothing to them. But we make sure we always use an approachable like we always use an approachable like we always use an approachable like language. So, it's like, "Hey, you're language. So, it's like, "Hey, you're language. So, it's like, "Hey, you're struggling. Could this go offline? I'm struggling. Could this go offline? I'm struggling. Could this go offline? I'm sorry about that. sorry about that. sorry about that. We'll try this next time. We'll keep it We'll try this next time. We'll keep it We'll try this next time. We'll keep it simple. Keep your phone on. Make sure simple. Keep your phone on. Make sure simple. Keep your phone on. Make sure you've got VPN on." Such right. This is you've got VPN on." Such right. This is you've got VPN on." Such right. This is VPN. So, we also do things like instead VPN. So, we also do things like instead VPN. So, we also do things like instead of just sending an email to be like this of just sending an email to be like this of just sending an email to be like this is what the thing is. We are big big on is what the thing is. We are big big on is what the thing is. We are big big on videos. We are big big on visual. Um and videos. We are big big on visual. Um and videos. We are big big on visual. Um and we also are happy to do like on site to we also are happy to do like on site to we also are happy to do like on site to show them the thing. So, if they're show them the thing. So, if they're show them the thing. So, if they're trying to struggle to understand it, we trying to struggle to understand it, we trying to struggle to understand it, we jump on quick calls. We we keep it jump on quick calls. We we keep it jump on quick calls. We we keep it simple as can be, you know? simple as can be, you know? simple as can be, you know? Um but yeah, fair question. Um but yeah, fair question. Um but yeah, fair question. Cool. Cool. Cool. I think it's probably it. I think it's probably it. I think it's probably it. Um yeah. Um yeah. Um yeah. Thank you. Thank you very much.