Sometimes there are inflection points in technological advancement that stop you in your tracks. Experiences that transcend the hitherto norms; that widen the aperture of your eyes.
Like accessing the internet for the first time way back when (1997) with Excite.com. Or using MySpace to create your first ‘social’ network in 2003. Then discovering in 2005 that Google Maps had mapped the entirety of Earth for you to explore.
These paradigms: internet, networking, and location based services, have been the drivers for virtually all innovation that has since followed. Indeed, their capabilities have become so ubiquitous that they’re now fundamental and embedded into virtually every home and business on earth. If we think of search, cloud, app stores, digital banking, digital health, and services like Uber, Deliveroo, Just Eat, Amazon they’ve all been made possible by these initial transcendental services.
But what’s next? What is the thing that will become so ubiquitous as to be embedded into everything we do.
The answer is AI; specifically Generative AI.
The distinction is necessary because, since I wrote about the art of the possible with AI in 2021, a revolution has occurred in the power and outcomes of large language models (LLMs) and generative pre-trained transformers (GPTs).
GPTs and, in particular, OpenAI’s ChatGPT has launched this technology through the stratosphere of the geekerati into the orbit of the masses. Indeed, with its exponentially advancing capabilities tools like ChatGPT have, in a little over two years, become the lay person’s definition of ‘AI’.
Plainly, Generative AI creates content from data and, as a result of being ‘trained’ on huge data sets, are capable of generating text, code, images, audio, and video from relatively simple prompts.
But it’s much more because, as adorable as puppies playing in the snow are, why would something that is relatively gimmicky make for a new all-encompassing category?
First, we need to understand the capabilities of Generative AI in more structured terms:
These are my hot take on a top 8 use cases for SaaS product companies but the use cases are vast and ChatGPT created the following word cloud conveying over 50 different use case paradigms:
The sheer scale of use cases means the ubiquity of Generative AI is inevitable and a vast wave of embedded services will soon be here. This Autumn (2024) Generative AI capabilities are being added into the operating system of iOS. There will be 2.2 billion users natively accessing generative AI seamlessly as a part of Siri:
Re-writing emails, creating reports and stories also becomes a breeze with the ability to generate prose and accompanying images becoming native to all Macs:
These use cases, mirrored in the Microsoft and Google ecosystems, are just the tip of the iceberg, as the demand for applications in the public and private sectors far exceeds consumer use cases.
To add some gravity to that statement see the following:
Life Sciences — Identifying Cancer Faster / Designing novel drugs
Finance — ChatBots that deliver UX and value
Education — Personalised education with synthetic tutoring
Marketing — Automating Marketing and Brand Voice adherence
But what does this all mean for consumers of public services and for the public sector?
The opportunity within the public sector for Generative AI is vast and recognised by the advent of a new-ish AI for Government Incubator and the Office for Artificial Intelligence becoming a core element of the new Department for Science, Innovation, and Technology.
Indeed, the Tony Blair Institute for Global Change estimates a cool £40 BILLION could be saved each year leveraging the AI capabilities and infrastructure that exist today. So, with exponentially improving AI capabilities, these savings are going to balloon.
The Alan Turing Institute also estimates that, in central government alone, there are some 143 MILLION transactions across 400 services that could be fully automated. Astonishingly saving 1 minute per transaction would equate to ~1,200 years of effort. If we add the administrative side of the NHS and the 800 services of Local Government, one can start to comprehend where the £40 Billion from earlier can materialise.
But how might these savings be realised in practical terms?
Within the public sector, the first and last consideration of any technology application is data security with data privacy and accuracy close followers. So, when seeking to apply this technology, it is of vital importance that it is being done so within guardrails of trusted frameworks and standards.
Fortuitously, the UK has been leading the way in this field in terms of both standardisation and frameworks for responsible AI. With responsible AI governance implemented, one can start to consider examples of how the different generative AI use cases could be adopted in local government.
In this article we will give the first of these examples with more to follow.
Example — Intelligent Agents
It’s fair to say chatbots have not been the average citizen’s favourite way to interact with an organisation. But this is mainly because they were dumb FAQ-based tools with little to no ability to diverge from an extremely narrow happy path.
Essentially, chatbots prior to 2023 were a set of pre-programmed workflows driven by a set of clicks or loosely established intents with little to no truly dynamic behaviours. GPTs, Retrieval Augmented Generation LLMs, and advanced natural language programming are poised to completely transform this.
How so?
In Local Government there are 317 councils delivering the 800 distinct services with each council receiving thousands/millions of online interactions, telephone calls and emails every year. Imagine, if every interaction/transaction could be fielded end-to-end by an intelligent agent.
To get the biggest bang for our buck we’re going to start with the high volume transaction areas of Council Tax, Waste, Housing, and Benefits. The data related to the services is all held in (albeit a great many) systems and access to the data is largely possible just without a helluva lot of standardisation. This is because there are a core set of systems supporting these service areas built by a variety of software suppliers with little to no collaboration. But, that doesn’t make it impossible.
In fact, virtually all of these systems have all the necessary APIs to respond to a huge proportion of the calls and emails being received. Additionally, while getting access to the data might be a wee bit tricky, the laws and policies that underpin services are all documented.
By using a RAG LLM, it is possible to augment the capabilities of a standard LLM/GPT to provide our intelligent agent with comprehensive knowledge on best practices, consistent approaches, and how to field any question a citizen might ask. Additionally, the combination of the Advanced RAG LLM and the authentication and data connectors will assure consistent and real-time quality results.
This Intelligent Agent can jump from handling a complex welfare query to council tax recovery and repair queries with ease. Unlike the current, and highly disjointed, experiences that both customer services staff and citizens face where even things like finding out how much you owe in rent and council tax aren’t trivial.
Intelligent Agents in practice
If we look at a large unitary council like Leeds City Council, its most recent customer service report highlights that, with its ~190 staff, it handles ~1 million calls and 150,000 emails across five high level areas:
Council Tax and Benefits
Housing and Planning
Transactional Services (waste, registrars, school admissions, highways, pest control, elections, blue badge)
Care and Safeguarding
Out of Hours (emergency/crisis events like emergency housing repairs)
We can see our four areas of Council Tax, Benefits, Housing, Waste are found within this master set. If the Intelligent Agent that is capable of dealing with even 30% of the total calls and emails, one would be looking at savings of ~£2m every single year without even considering the value add/other strategic value that could be delivered and savings in training/productivity.
Something not touched upon yet is the concept of failure demand. Within the aforementioned report one sees a metric called ‘Right First Time’:
We can extrapolate that the median value for this is 74%. Meaning that 26% of calls were not resolved on the call. Another way to see this is that if all calls were answered first time Leeds Council would receive 26% less calls and could save the council somewhere in the region of £900,000 per annum.
Leeds City Council are not atypical. It is estimated that failure demand is anywhere between 20% and a mind-blowing 80% across the public sector. Customer eXperience Management (CXM) apps like GovMetric’s CX currently enable organisations to understand their failure demand and better inform strategic decision-making on system and process improvements. Generative AI is used to automatically establish trends and make recommendations to reduce this.
For example, in data seen from GovMetric’s CX platform, feedback from the most widely used online council tax and housing transactional services consistently highlight that dissatisfaction comes as a result of:
“Not being able to do the task I wanted to do”
“Not being able to find the information I wanted”
“The user experience making it difficult to complete the task I wanted to do”
One will notice in the architecture diagram above there is an element titled ‘Quality Assurance’. The purpose of this is to appraise customer satisfaction through quantitative and qualitative feedback. Then using Generative AI learn from this and automatically adapt and improve the Intelligent Agent’s capabilities to ‘adapt and learn on the job’.
What this example conveys is the incredible value that could be achieved using Generative AI within customer services in the public sector. It enables a giant leap forward in terms of capability and, importantly, provides alignment with the ever increasing expectations of citizens driven by the transcendence of these capabilities across all their consumer experiences.
Creation of an Intelligent Agent will be seen as challenging, and by some that the odds are stacked against the public sector.
I believe that the remarkable and far reaching capabilities of Generative AI will enable councils to transcend the barriers faced and deliver truly unparalleled, value driving citizen services.
P.S. next time we will look at how Generative AI will enable the replacement of online forms and digital platforms for the ultimate user experience and massive public sector efficiency savings.
Customer engagement solutions built for the public sector
Harness the power of the customer voice to design services that meet the needs of your citizens and communities.
Comentarios