Northern Trains are a prominent Train Operating Company (TOC) in the United Kingdom who carry approximately 100 million passengers each year, serving a network of 15 million people. They operate an extensive network of regional train services across the North of England, connecting major urban centres and facilitating commuter and passenger movement, , serving both urban and rural communities. Their Sheffield customer experience centre, the central hub for customer queries and concerns, consistently fields a high volume of customer queries and provides essential support to passengers.

Challenge

To improve customer experience and wait times, Northern Trains was looking to create a self-service path for customers to address questions for which answers can be found on Northern’s website, as well as automating common processes where possible.

The requirement to Alscient consisted of the following:

• Create a chatbot that the public can interact with and answer their Northern Tains related questions where possible and transfer to agent where necessary.

Integrate the bot with existing complex Salesforce architecture and upon transfer to an agent create a contact and case, all using omnichannel.

What We Did

For a risk-adverse approach, predictive AI was used (rather than generative AI) which uses pre-set dialogs, giving the product owner complete control of the outputs of the bot. Alscient started by gathering a substantial amount of UK train knowledge to produce a set of intents to represent the majority of queries that Northern Trains’ customers may have. We then created an initial set of utterances for each intent, in line with Northern Train’s official tone of voice and specific best practices set by the team expert. Agentforce was used to help generate a deeper set of intent matches to account for text variations.

Custom logic was created to gather information from the bot user and surface it to the omnichannel agent before, and at the moment of, transfer to live chat. Lastly, Northern Train’s branding was implemented into the UI of the Salesforce Einstein Bot before it was handed over to the beta testers. From this round of testing, Alscient were able to further train the Einstein bot with real world interactions and make adjustments to the intents to align further with how Northern Train’s customers wanted to use the bot.

Industry metrics suggest that the solution will deliver 20–40% contact deflection, significantly reducing inbound calls and live chat volumes. This leads to operational cost savings of 15–30% through lower agent demand and improved scalability during peak travel periods. Average response times are also expected to drop from minutes to seconds, improving customer satisfaction and driving NPS increases. Additional benefits include 24/7 availability, improved first-contact resolution (up 10–25%), more consistent information delivery, reduced queue times, and better agent focus on complex, high-value customer interactions.

Result

The first release of the Einstein bot was delivered with an F1 (prediction accuracy across the entire dataset) score of 0.88 (which is above industry averages). This was then extended to include bot replies grounded in Salesforce Knowledge (using Data Cloud and Agentforce) for a better user experience. As part of a larger, overarching plan to revolutionise Northern’s digital capabilities, Alscient successfully delivered this solution from the ground up in 8 weeks using Agentforce for AI-Assisted build support. The internal team have seen a 25% reduction in general enquiry calls due to the initiative with the solution breaking even in terms of cost/savings inside the first year.