About UK Energy Networks
UK Energy Networks is the biggest electrical energy distributor within the UK. It maintains electrical energy cables and contours in London, the East, and the Southeast of England. UK Energy Networks delivers power to 19 million folks throughout its remit. It ensures {that electrical} networks are secure, safe, and dependable. It’s a key participant in serving to the UK meet Web Zero and focuses on supporting renewable power, low-carbon heating, and electrical automobile chargers.
UK Energy Networks drives innovation within the power sector by consistently partaking in technological developments. The corporate leverages its information belongings and utilises machine studying to sort out Web Zero, automate inside processes and supply dependable, cost-effective options to prospects. Buyer satisfaction lies on the coronary heart of their enterprise. The corporate cultivates the customer support tradition and ensures that each one prospects, particularly the susceptible ones, are protected.
Overview of the problem
Clients are more and more utilizing digital means to get in contact, and one of many challenges confronted by the customer support crew at UK Energy Networks is a large variety of emails coming every day (300-400 emails with a ten% improve 12 months on 12 months). These emails may be labeled into three classes. The primary class contains requests for actions (jobs), similar to getting ready quotations or disconnecting shoppers. The second class contains questions on current jobs or different inquiries concerning the enterprise. The ultimate class consists of emails of decrease precedence to the power distributor, similar to automated replies or assembly notifications. The customer support crew used to manually overview incoming emails and assign classes to them, indicating their urgency. This course of was time-consuming and susceptible to error, which may result in UK Energy Networks’ delays for some prospects and within the worst case service stage agreements (SLAs) not being met.
UK Energy Networks partnered with Databricks, Microsoft and CKDelta to create an automated answer that may pace up this course of. The impact of that collaboration was making a Buyer Digital Agent. It is a new Outlook inbox expertise, enhanced by clever options similar to e mail classification and summarisation, utilizing Giant Language Fashions (LLMs) on Azure Databricks. The answer was developed as a Proof of Idea to check the capabilities of generative AI for a customer support use case.
Creating the answer with CKDelta
CKDelta has been UK Energy Networks’ technical associate throughout a number of innovation tasks utilizing machine studying, together with Highlight, Envision and Optimise Prime. CKDelta builds data-driven AI functions and machine studying fashions, empowering prospects to attain sustainable, secure and environment friendly enterprise outcomes.
As a part of the CK Hutchison Holdings Group, CKDelta has entry to uniquely enriched and constantly refreshed information from industrial-scale sources. The AI functions underpinned by this multiple-sector information current a uncommon alternative to be taught from the previous and predict the long run with confidence.
The Buyer Digital Agent was in-built partnership with Databricks and Microsoft. Companies offered by these companions are on the core of the developed answer. Beneath the hood, the e-mail processing workflow runs on the Azure Logic Apps platform. Every incoming e mail triggers this workflow. In step one, it saves e mail our bodies, topics, and attachments in Azure Blob Storage. Then it strikes to the computational a part of the answer. That half takes place within the Databricks platform.
Databricks permits utilizing PySpark for information preprocessing simply and effectively. PySpark was a vital device to remodel UK Energy Networks’s e mail dataset. It consists of enormous volumes of textual content information which is straightforward to load into PySpark DataFrames. They will then be seen, explored, and interacted inside Databricks’ notebooks. The e-mail information was cleaned with the assistance of LLMs. The UK Energy Networks’ occasion of the GPT-3.5 Turbo mannequin was accessed immediately from Databricks utilizing Azure OpenAI API. The mannequin was used to establish components of the emails necessary from the standpoint of the customer support crew. The directions, containing UK Energy Networks’ inside information, are part of the immediate handed to the mannequin. LLMs are additionally getting used to assign e mail classes and supply summaries of the messages.
As talked about within the earlier part, three GPT-3.5 Turbo fashions are answerable for dealing with incoming emails. Every of the fashions has precisely one job to carry out, i.e., establish necessary components of the emails, assign classes, or summarise the content material. Prompts handed to the fashions embrace area information that the customer support crew at UK Energy Networks makes use of to make selections about incoming requests. Altering these prompts is straightforward and permits for fast enhancements within the accuracy of the mannequin. It additionally permits tailoring the fashions in direction of necessary enterprise metrics. MLFlow experiments had been used to check what prompts have the perfect impression on the mannequin’s efficiency. Experiment monitoring in Databricks is very efficient and with new LLMOps options launched in MLFlow 2.4. it is easy to match prompts and observe LLM-specific metrics. Databricks additionally permits to simply register LLMs and to entry them for brand new incoming emails.
Within the final step, the anticipated e mail classes and summaries are despatched again to Outlook. A class is added to the unique e mail as an e mail tag. This manner members of the customer support crew can see the main target of the e-mail, with out even having to overview its content material.
An e mail abstract is then added in a unique color firstly of the unique e mail. This makes the division between the unique content material and the LLM output clear and intuitive. Most significantly, the answer enhances the expertise of the customer support crew, with out altering the way in which they proceed. It would not require any extra work on their behalf nor modifies the strategy that has confirmed to be optimum for them.
Advantages of the answer
The Buyer Digital Agent saves time that the customer support crew can use to offer extra customised assist to their shoppers. Having out-of-the-box classes for all emails accelerates reviewing emails and minimises the chance of lacking SLAs and receiving penalties. Summaries are notably helpful within the case of lengthy chains of emails between prospects and UK Energy Networks departments. Studying a abstract can provide material specialists (SMEs) an concept concerning the nature and urgency of the request, in addition to a listing of actions already taken by UK Energy Networks. It can be over 10 instances sooner than studying a really lengthy chain of emails. The outputs offered by LLMs enhance enterprise processes in UK Energy Networks and immediately impression buyer satisfaction.
Subsequent steps for the Buyer Digital Agent
The answer has confirmed that LLMs may be efficiently used to optimise enterprise processes linked to customer support. The SMEs discovered the answer efficient and time-saving, highlighting the reliability of outputs offered by LLMs. It’ll have a big impression on the main target of their work, which is able to not require handbook classification of incoming emails. Contemplating the optimistic reception of the answer, the subsequent step will contain productionising it to make Buyer Digital Agent a everlasting a part of the method that emails undergo. Over time the answer may grow to be extra customised, addressing ongoing suggestions from the customer support crew. Due to Databricks Mannequin Serving it’s going to be straightforward to check completely different mannequin households, together with Llama and Mistral fashions, to match their outcomes with the initially picked GPT-3.5 Turbo. Mannequin Serving gives a unified interface that permits managing a number of fashions with a single API, making future experimenting with completely different LLMs extra agile. These fashions is also examined for different potential options of the answer, similar to alerting concerning the urgency of requests or offering drafts of responses.
