Image credit: https://unsplash.com/@oliver_photographer

TfL eyes chatbots to drive down contact centre demand: Beware "deviations" and "harsh" voices...

London transport organisation also "interested in any capabilities related to technology in contact centres that go beyond these five capabilities"

Transport for London (TfL) hopes chatbots can help it reduce pressure on its contact centres – and is going to market in September for technology suppliers that could help it stand up a programme to deliver them. 

TfL has started early market engagement as it seeks to understand supplier capabilities around “self-service chatbots” – a tender this week studiously avoids using the term “AI” and suggests that TfL is thinking more expansively about how it modernises its customer services. 

TfL manages a bus fleet of around 9,300 vehicles operating across 675 routes, and over 19,000 bus stops; 11 tube lines covering 402 kilometres and serving 272 stations; the driverless, computerised Docklands Light Railway; and eight piers along the Thames, among other responsibilities.

In 2022-2023 its passenger income was £4.2 billion but it is aiming to drive down costs across its estate and is targeting savings of £600 million from 2022/23 to 2025/26, its last annual report (September 2023) shows.

A March 7 notice highlights its focus on “reducing the demand coming into our contact centres; maximising the use of current and new data to produce strategic insights on TfL customers… [improving] our customer's experience as well as the experience of contact centre advisors.”

AI chatbots come with challenges… 

TfL’s digital leaders may need to be warned however: Every organisation out there freshly looking to use generative AI to deliver customer-facing chatbots is finding it a more thorny experience than they realised. 

Speaking at a financial services event this week attended by The Stack, for example, Vladislavs Mironovs (Chief Strategy & Business Development Officer at Latvia’s Citadele bankas) emphasised that he had a strong data science team and a “great database of customer decision trees” to start.

“Placing these all into genAI seems like a good starting moment, but brings a lot of issues…” he explained on March 7 at moneyLIVE.

“On-premises model deployments were spitting out lots of “deviations”; the generative AI “could not recognise” payments data and the “tone of voice was harsh” amongst other earlier challenges, he told the audience

(After a lot of wrenching and a decision to use GPT-4 Turbo that has now been improved and a beta chatbot is now live and able to handle 70% of requests, he said – a bold claim given it had only launched that same day.) 

See also: Can £160 million and the cloud fix the DWP’s dismal contact centres?


TfL, meanwhile, said it wants to hear from suppliers if they have “solutions based on the following five capabilities:

1: Self-service Chatbots: A first point of contact to resolve TfL customer issues through self-serving our customers, where possible
2: Omni-channel / New Channels: A capability offering new and diverse methods of communication for TfL customers, such as WhatsApp, as well as the ability to collect and analyse data from across multiple channels
3: Agent Assist: A capability to augment and improve the contact centre advisors' ability to respond to customers as well as improve their productivity
4: Case Management: A capability that utilises automation to triage and manage cases

TfL is, it added, “also interested in any capabilities related to technology in contact centres that go beyond these five capabilities...”

Join peers following The Stack on LinkedIn