Three Agents in Three Months: Farm Focus Sows Agentforce At Speed

Companies are understandably taking a measured approach to deploying agents, a new technology with uncertain risks.

And then there's Farm Focus. It rolled out two agents into production within three months and has scoped a third.

Even more remarkable – the agents were built on the Agentforce platform which Farm Focus deployed while moving to Salesforce in the same quarter.

"It was an intense project," said Blair Rooney, Farm Focus' head of growth. "AI is coming, whether you like it or not, and businesses need to adapt," he said.

Farm Focus may have an advantage in technology adoption given that it's a software company in New Zealand filled with developers. It has sold financial applications to farmers for 45 years.

The deployment shows how quickly a willing organisation can move.

‍

Service first

Farm Focus chose the service agent first because that was what Salesforce had just released. "The first agent that Salesforce released was the service agent, so we sort of just dived into that," Rooney said. "That was one use case that we saw straight away that would work."

The company runs a customer success team that handles 12,000 to 13,000 chats a year. Rooney had two goals: cut the cost of that function, and get customers answers faster. Not fewer contacts. "The value that the customer would get by getting the answer quicker was actually better for them," he said. After three months, the agent was handling three quarters of inquiries on its own.

The team's workload did not fall. "Inevitably, our team actually take more calls," Rooney said.

Keeping the agent accurate requires active management. The team monitors chats for quality, runs test scripts through Salesforce's testing centre before changes go live, and keeps the agent from straying into financial advice. Salesforce's own tooling uses sentiment signals from customer responses to flag likely wrong answers. Accuracy runs at about 80%. "I would say about 80% accurate," Rooney said.

Some failures were specific. The agent sometimes invented interface guidance from help articles it had never seen in action. "It doesn't know what the product looks like, but it'll say just click that button," Rooney said.

The second agent works in marketing, building customer segments. A marketer types a request in plain language – dairy farmers in New South Wales who have not logged in for 30 days – and the agent returns the segment.

"That used to take us three weeks and potentially $10,000 of people's time," Rooney said. "Now my marketing team can do that in about 30 seconds."

Rooney put a figure on the wider saving. He said the move to Salesforce cut costs by "around $1.1 million per year", which he attributed to dropping other tools and outside help.

What still has to hold

The company is building a third agent on AWS Bedrock, which it hopes to use with financial data. Despite success with the first two agents, Rooney is wary of letting an agent act on its own. "This whole autonomous agents thing is a bit scary for people, especially with what we do, because it's financial," he said. The team has built guardrails to stop the agent giving financial advice.

Blair Rooney was head of growth at Farm Focus at the time of this interview. He is now managing partner at Enable. The full interview is available on the Scale100 YouTube channel.