Thames Valley Police (TVP) and Hampshire & Isle of Wight Constabulary (HIOWC) introduced Bobbi, a public-facing AI agent built with Salesforce Agentforce, to ease pressure on the 101 non-emergency line, according to a Salesforce customer story from May 2026. The line received more than 800,000 calls a year across both forces, nearly half about unresolved issues and 20% asking for updates on active cases, and waits could exceed 24 minutes at peak. Cutting waits to four minutes through hiring would have needed 33 more staff at £8 million a year.
Bobbi works on the forces' websites and in their Citizen Portal as a 24/7 first point of contact. It collects the person's name, contact details and issue, answers in natural language in under 350 words, and routes the chat to an operator when it cannot help. For matters such as noise or parking it explains the situation and points to the right reporting page, and for incidents such as car break-ins it gives step-by-step guidance on reporting and preserving evidence.
Answers come only from 91 approved knowledge articles stored in Salesforce Data 360 and indexed in a vector database, not from the open internet, and staff can update the articles themselves. Bobbi checks each message for keywords and language signalling vulnerability, drawing on articles on domestic abuse, sexual violence and honour-based abuse. When it detects risk it creates a case in Agentforce Service, flags it for staff and offers an immediate handoff to a trained operator by live chat, passing the transcript, metadata and a summary so the person does not have to repeat themselves. The forces say it identifies around two potential violence against women and girls cases a day.
Bobbi was built and tested in 12 weeks in November 2025 using Agentforce's low-code Agent Builder. Salesforce reports that it handles more than 200 conversations a day, resolves 75% independently and has cut cost per interaction by 97%, with a CSAT score of 4.6 and 93% positive feedback. "Every single day, we've seen Bobbi correctly identify at least one high-risk situation and get it routed to someone who's trained to help," said Tom Boyd, Digital Product Manager. The forces expect a 900% return on investment and 19,000 staff hours reclaimed, and plan to let citizens submit reports directly in chat.
Details
- City
- Kidlington
- Organization