Building your own AI Solutions

In my previous blog, we went through what the AI solution in M365 called Copilot is, and how we are seeing Local Authorities make use of the technology.

Copilot is a Software as a Service offering – that is to say, a Service which Microsoft have written, will maintain, and offer customers as part of a Per User Per Month cost offering to make use of.

This type of offering however isn’t the end to the AI story with Microsoft – far from it.? Microsoft also make available all the building blocks you might need as a Council to start building out your own AI solutions – at which point, the limitation becomes what you can imagine.? Let me cover off some of the technology that allows you to build your own solutions within Azure, using the Azure AI Services.

For Local Government, there are a few standout services with the Azure AI Services offering to call out:

The Azure Open AI Service is essentially one place where you can securely connect to, and make use of OpenAIs Large Language Models (LLMs) GPT 3.5 and GPT4.? This is fantastic for building out rich chat bots, both in Customer Services and internal bots for your own custom applications, or providing a natural language starting point to a process or request.

Azure AI Search provides rich indexing and querying of data, and semantic ranking of results.? This can offer an exceptional search engine across internal and external data, and act as one of the key building blocks to creating a solution to resolve FOI queries with minimal human interaction when coupled with the Open AI Service.

Bot Service is a solution which allows you to create chat bots, which can leverage LLMs from Open AI to answer any type of question, be that from residents or internal staff.

Immersive Reader is a tool which allows you to implement accessibility features to make websites and apps you provide more accessible.? It uses reading comprehension techniques to help make content accessible for new readers, language learners and individuals with learning differences.? It can also help display pictures for common words, help read content aloud, translate content in real time and split words in to syllables.? This solution can help with creating easy read documents for residents with learning difficulties, and to translate content for residents who don’t speak English as their first language, or at all.

Azure AI Language is a service which uses natural language understanding to extract information from unstructured text.? It can help identify key phrases in text, find Personally Identifiable Information (PII), summarise text, find linked entities, or apply document classification. ?There is also a Text Analytics module for Health, which enables you to extract healthcare specific information from clinical notes and reports, which is important in a Social Care context.?

There are a whole range of use cases that this enables in Local Government, from rich search across documents to identify information relating to a person or place for case management, locating data for FOI requests, summarising health and social care data into a care record, and information redaction and summarisation for publication of documents.

Speech Service is a service that recognises speech, synthesises speech, can enable real time Language translation, transcribe conversations and integrate speech into bot experiences.? This technology is one we are seeing being used already in contact centre scenarios to record conversations, summarise them, and create case notes.? We are seeing interest in how this could be used in social care and face to face scenarios, and it is already being used by partners to help contact centre deflection, integrated into a bot service – more on that later!

Azure AI Translator is a solution which allows you to translate text and documents using modern neural machine translation technology.? We are already seeing local authorities use this to translate documents to store in their document repositories, and also to translate letters being sent to residents.

Azure AI Vision is a technology that can analyse images and videos and apply algorithms to help provide different outputs.? For example, it can provide Optical character recognition which could be used for scanning documents or extracting information like number plates or sign details from street scene images.? If can provide image tagging and caption capabilities if using images as part of a process for example for fly tipping or graffiti reporting.? It can provide Spatial Analysis which could help with scenarios like footfall analysis or over crowding scenarios.

Azure AI Video Indexer allows you to analyse video and audio captured.? You can apply deep search to find specific moments, perfect for allowing residents to search cabinet meeting footage for discussion related to an issue that concerns them.? You can make videos accessible, and translate into different languages, fantastic for any informational videos a council might publish.

Custom Vision allows you to apply your own learning model to images and videos.? This is important for Place services offered by local authorities.? For example, this could be used to identify pot holes, damaged signs or street furniture, graffiti or fly tipping if applied to footage captured by vehicles.? It can help with traffic monitoring, or collision identification when applied to road footage.? We have examples of it being used to help with identification of RAAC, or other home based issues such as black mould or damp.

?

The great thing about all these services is that they can be leveraged, or pulled together to create richer solutions.? For example, a partner of ours called ICS.AI are leveraging multiple elements here to create not only a rich chat bot for contact centre deflection, but one that can speak multiple languages, be exposed to IVR to help deflect cases when people call in, and we can then apply Speech Services if the call is passed to an agent to transcribe the call, and summarise into case notes.

In Local Government, we are seeing a wide range of projects being stood up already to make use of these capabilities.? Examples we are already seeing include:

·?????? Derby City deploying an AI powered phone assistant, helping to handle 62,000 calls, deflecting 43% of calls, and saving ￡200,000 with our partner ICS.AI in their initial phase. Derby are now extending the AI capabilities in more service areas after a series of workshops identified 54 different use cases. Over the next four months, phase 1 will look at Adult Social Care, Customer Services and Debt Recovery, with plans in place to roll out phases 2 and 3 across other services once it is running successfully which will lead to an expected ￡12.25m

·?????? A London borough has deployed a Document Translation service, reducing the cost of translating letters from ￡120 to just 20p per letter.

·?????? A local authority are currently building a FOI response service, which will apply rule based responses and sensitive and PII redacted responses to reduce the time and effort handling FOI requests

·?????? Several councils have built Invoice Processing systems, which can analyse a Invoice provided to the authority in social care, and extract the data needed to comply to the councils process ready for payment out.

·?????? South London Partnership worked to pull together disparate IoT solutions to be able to create rich views across the boroughs to target things like Fly Tipping, Air Quality and Air Quality in Schools to make decisions around car free zones at drop off, as well as monitoring gulley's to see blockages early and prevent road flooding and damage.

·?????? In Canada, the City of Kelwona has built a bot to support planning applications and streamline the process until a planner is needed to engage

·?????? Mexico’s Supreme Court of Justice has built a solution to provide residents access to historical documents

·?????? Government of Nunavut in Canada have built out a service to help promote and support Indigenous languages through translation of documents, but also creating educational content for school

·?????? The Municipality of Alkmaar in Holland is building a Digital Twin of the region and looking at improving traffic flow and traffic light performance

?These examples are just beginning to scratch the surface of what could be achieved in Local Government, and we are seeing a rapid rise of use cases building that can help improve services, and reduce cost.? From speeding up recommendations in Social Care to identifying Fraud in the system, from Resident facing chat bots, to using bots and process automation tools to help speed up internal processes we are seeing a range of use cases being explored.?

The possibilities aren’t just related to people and processes, but also place.? From traffic management, air quality, footfall monitoring, city planning, waterway management, broadband and connectivity planning, there is a range of possibilities emerging to help with city planning, transport planning, and addressing the housing crisis.

The key here for Local Authorities isn’t to boil the ocean, and try to everything in one go, but rather to engage with employees, and look for tasks which AI could be applied to in order to speed up the process, drive out cost, or provide a better outcome, and to look to start small, and find successes quickly or fail fast.

Microsoft is here to help navigate from ideation to realisation, with a rich and vibrant partner community who can support you in standing up solutions at pace, and providing expertise to help you on the journey. If you are interested in learning more, please reach out.