LinkedIn
ForceFront Homes is the Agentforce solution that won the London TDX 2025 hackathon - and a trip to Dreamforce. It imagines a social-housing provider where tenants report repairs and ask questions through a customer site and messaging, and AI agents do the heavy lifting behind the scenes.
What it does:
- Tenants report repairs and raise questions through a self-service site and live messaging
- Agents triage each Case, summarise it and update its key fields
- Contractor visits are booked against the right fault type
- Possible safeguarding concerns are flagged for a person to review
- Employee agents help staff answer questions, grounded in the housing provider's own data
Built by team ForceFront, a two-person team with me doing the build, it has nine agents, 21 agent actions and nine prompt templates, and was designed to speed up service and improve safety while adapting to the workflows of any industry. Salesforce Ben later featured the approach in "Strategy to Solution: How to Win an Agentforce Hackathon".
How I built it
A housing provider's customer service team hears things like "the boiler's broken again", "I can't afford this month's rent" and "there's water coming through the ceiling", by email, phone, web form and messaging. Every one of those used to be typed into a Case by hand.
The Case Processing Agent reads the free-text description of a new Case and fills in the structured fields: priority, subject, origin, status, reason and sentiment. Each field has a fixed set of values and a safe default when the description does not make the answer clear, so the agent never leaves a Case half-classified. The subject is kept under 100 characters, and the sentiment is a short phrase describing how the tenant feels.
When a description suggests a safeguarding concern, such as a tenant who sounds distressed or children left without heating, a linked Safeguarding record is created and the right team is alerted. The agent flags; a person reviews.
Repairs are booked as Contractor Visits against the right fault type, and employee agents answer staff questions grounded in the housing provider's own data in Data Cloud. Tenants reach all of this through an Experience Cloud site and live messaging.
The approach
The important point is the division of labour. The agent handles the language; flows write the records. Keeping the flows simple and letting the agent interpret the text made the solution quick to build and easy to adapt to other industries.
What I would tell another team
- Give every field the agent sets a fixed list of values and a safe default
- Let the agent handle language and let Flow write the records
- Route anything sensitive, like safeguarding, to a person
I design, build and review agents like this as an Agentforce consultant in the UK.