Six months after its launch, Thames Valley Police and Hampshire & Isle of Wight Constabulary say that Bobbi, the UK’s first police AI agent, is changing how communities access policing and how contact center teams manage non-emergency demand.
Built on Salesforce Agentforce, Bobbi was introduced as an additional digital contact channel for citizens who need guidance, updates or support, but do not require an emergency 999 response. The agent is now helping manage the equivalent of 14,000 citizen contacts annually across both forces, while freeing up more than 3,200 operational hours a year.
Chief Superintendent Simon Dodds from the Joint Operations Unit for Thames Valley Police and Hampshire & Isle of Wight Constabulary, said:
“Bobbi has allowed us to rethink how we manage non-emergency demand, ensuring our officers and staff can focus on the people and situations that need them most.”
Related articles
Salesforce Panel Highlights Cautious Public Sector AI Adoption in Frontline Use Cases
Salesforce Futures VP: Customers Face Three Core AI Challenges as Agentic Enterprise Takes Shape
Salesforce Expands London AI Campus as Agentforce Adoption Moves “Beyond Pilots”
The Challenge: A Digital Front Door for Modern Policing
Policing teams across Thames Valley, Hampshire and the Isle of Wight handle a large volume of public contact every year. At Salesforce’s Agentforce World Tour London event, Tom Kempster, Director of Digital and Innovation at Thames Valley Police, explained the scale of that challenge, explaining: “We typically receive between 1.3-1.4 million contacts per year.”
A significant proportion of those contacts are emergency calls. But much of the remaining contact is non-emergency, with citizens asking for updates, trying to contact an officer, seeking advice, or looking for the right public service to handle an issue.
Before the Salesforce program, wait times for calls to 101 were placing pressure on citizens and staff. “We were under extreme pressure when we started this pre our Salesforce journey. Average wait times were about 26 minutes on our 101 service. That number is doing a really good job of hiding some really high wait times in our peak times.”
The forces also recognized that citizens’ expectations had changed. People are now used to contacting banks, retailers, healthcare providers and councils through multiple digital channels. Policing, by contrast, has historically relied on 999 for emergencies and 101 for non-emergencies, with some online reporting options.
As Rob Brind, Superintendent and Head of Operational Delivery for Contact Management, Thames Valley Police and the Hampshire and Isle of Wight Constabulary, noted in a panel discussion: “We knew that the challenge was the public is shifting in the way that they want to contact us or how they want to speak to us, so we knew that we needed to move with the times.”
The forces’ first Salesforce transformation focused on the service stack: public portals, automation journeys, and more sustainable routes for citizens to contact officers or receive updates. That work produced measurable gains, including a 15 percent increase in customer satisfaction and a major reduction in average wait times.
Bobbi was then layered on top as the agentic AI capability.
The Solution: Bobbi, Built on Salesforce Agentforce
Bobbi is designed to support non-emergency interactions, answer common questions, guide citizens to appropriate services and escalate cases where risk is detected. Bobbi now handles around 200 conversations per day and fully resolves approximately 45 percent of non-emergency enquiries without human intervention, according to Salesforce.
The AI agent is grounded in 91 verified knowledge articles. It does not access the open internet or unverified sources, and its responses are capped at 350 words to ensure they remain concise, professional, accurate and empathetic.
Under the hood, Bobbi uses Salesforce Agentforce 360, Data 360, Experience Cloud, Agentforce for Public Sector and MuleSoft. This setup connects into legacy policing systems, reduces duplicate data entry and enables replies to reach citizens without unnecessary manual intervention.
At the London event, Salesforce positioned Bobbi as an example of agentic AI moving beyond pilot deployments and into real operational use. Zahra Bahrololoumi, CEO of Salesforce UK&I, said: “Bobbi proves that the Agentic era is here, protecting lives right now. With Agentforce, our police forces have shown what it means to put communities first.”
More than a generic chatbot, Bobbi operates in one of the most sensitive contact environments possible, where even an apparently routine interaction may reveal a safeguarding issue, a crime or an emergency.
As Kempster pointed out, “any channel that the police open could potentially be an emergency.”
That principle shaped the implementation, because Bobbi had to do more than deflect low-value demand. It had to understand when not to automate.
From Contact Deflection to Safeguarding
The forces initially expected Bobbi to help with low-level, non-policing enquiries such as dog fouling, parking complaints, loud music or issues better handled by local councils.
While that has happened, the more significant finding has been that Bobbi is reaching people who may not have contacted the police at all.
As Brind explained:
“What we’re finding is that people who perhaps aren’t confident enough to pick up the phone and talk to a police officer and speak to us in the traditional way will actually engage with the agent and have the conversation with the agent.”
That behavior has opened a new route into policing for vulnerable people. Some citizens use Bobbi before deciding whether to speak to a person and others use it because they cannot safely make a phone call.
“If we identify a crime or something specific, we can deflect it to an agent on a digital desk, but people don’t always want to do that,” Brind said. “They want to test the water, they want to understand, they want to find out a bit more, and we’re finding that in some of the most harmful cases that we get reported, so we’re seeing it in serious sexual offenses cases and also domestic abuse, where people will speak to the agent, get the advice, build up their confidence and then we can start dealing with that in a very, very different way.”
“That was a real eye opener to us, because we didn’t expect that contact to come through, and it came through literally within days of us going live.”
Salesforce cited one example involving a 16-year-old boy whose father was making threats at home. Unable to risk an audible call, he used Bobbi to type out what was happening. Bobbi detected the risk, flagged it to a human supervisor, and offered a handoff to a live web chat operator. Officers were dispatched and de-escalated the situation.
The forces say that every day since launch, Bobbi has identified at least one high-harm offence and escalated it to a human operator. On average, two cases of violence against women and girls are identified and routed for human intervention daily.
As Chief Superintendent Dodds said: “What’s been particularly powerful and unexpected is seeing how these services can help people access support in ways they may not have felt able to before.”
Guardrails, Testing and Trust
Deploying AI into policing raises obvious questions around trust, safety, accuracy and accountability. The forces were conscious of that from the outset.
“Al in policing, is quite controversial,” Brind said. “So, we were really, really mindful of that when we went live with Bobby as agentic Al, and the first one in the UK, possibly globally, so we put some quite strict guardrails around that solution.”
Those guardrails include keeping Bobbi focused on approved internal knowledge sources, continuously reviewing responses, monitoring feedback, and tuning the system as new patterns emerge.
Implementation was relatively fast, but the forces emphasized that much of the work was in testing rather than simply switching on the technology.
“We spent three to four weeks implementing it, and then the rest of it was all testing and checking and putting guardrails in,” Brind said. “That was the time-consuming part of our goal, which was making sure that the messaging Bobby was giving out was consistent and also factory accurate, especially in a policing environment, because we need to make sure that it was correct.”
That approach aligns with Salesforce’s broader message around agentic AI, that agents need data, governance, observability and human supervision.
Joe Inzerillo, President, Enterprise and AI Technology at Salesforce, said:
“These agents are not here to replace people, they’re here to do work, so people can focus on more impactful things. But you also have to mind them, you can’t just let them go off.”
For Bobbi, that means maintaining a clear line between AI-supported service and human decision-making. The agent can guide, answer, route, and escalate, but it does not replace officers, call handlers or safeguarding professionals.
The Results: Hours Saved, Demand Managed, Access Expanded
After six months, the forces say Bobbi is now an established part of their digital front door. Citizens have rated the experience 4.6 out of 5, and the agent is handling enquiries in multiple languages, helping reduce barriers for communities that may otherwise struggle to access policing services.
Bobbi helps manage around 14,000 citizen contacts annually across the two forces and frees an estimated 3,266 operational hours each year. It handles 200 conversations a day, with 45 percent fully automated.
“What we’re not looking to do… is to save money using Bobby,” Brind said. “What we’re trying to do is we invest that in value-added service to the public.”
The freed-up time is being reinvested into higher-value work, particularly crime-related and safeguarding interactions that require human attention.
Brind added: “What now we can do is focus on the meaningful crime, and focus on those people that are true victims of crime, and how do we then deliver a much better service through that.”
What Comes Next
The forces are now exploring additional internal AI agents, including HR tools that could support officers and staff behind the scenes. The goal is to apply what has been learned from Bobbi externally to internal processes, where policing complexity, legislation and operational knowledge create heavy information burdens for staff.
Mike Lattanzio, Chief Digital & Information Officer at Thames Valley Police, described Bobbi as a marker of broader change: “The success of Bobbi over the last six months marks a paradigm shift in how citizens interact with vital public services. This is a first for UK policing, and I’m confident other forces will follow.”
For Salesforce, the case study is also an example of responsible AI in public services. Bobbi shows how agentic AI can reduce avoidable demand while strengthening human response where it is needed the most.
The bigger lesson for policing may be that in addition to making responses via digital channels more efficient, AI agents can reveal needs that were previously invisible.
“We’ve had multiple examples of people reaching out to us when they can’t get to the phone, or they can’t communicate with us in any way, or they don’t feel comfortable with communicating with us in any other way, and we’ve been able to make arrests off the back of that contact,” Kempster explained.
That is where Bobbi’s real impact sits, in creating another safe route to reach policing.
