The Correspondence Drafter was one of the department’s first uses of generative AI. An internal proof-of-concept demonstrated that generative AI could significantly reduce the time required to draft responses. The next challenge was to turn that early prototype into a secure, production-ready service that aligned with departmental controls and could be trusted by civil servants in their day-to-day work.

Challenge

Drafting responses to correspondence was a manual and time-intensive process. Team members needed to locate the correct briefing packs, interpret guidance and produce responses from scratch before reviewing and sending them. While this ensured accuracy, it limited the team’s capacity and made it difficult to respond quickly as volumes grew.

The Department for Education wanted to understand whether AI could assist with drafting without undermining governance, data protection or the role of human judgement. Any solution needed to be secure, auditable and easy to evolve over time, while remaining firmly under the control of the correspondence team.

Approach

Made Tech joined a blended, multidisciplinary team working across the department to support this work. Our role was to provide technical architecture expertise, helping shape a solution that could move safely from prototype to private beta within the department’s Microsoft Azure environment.

Working closely with DfE developers, infrastructure specialists, user researchers and Microsoft stakeholders, we helped design a production-ready architecture that met departmental standards and security requirements. This included supporting internal assurance processes, advising on best practice for cloud-native AI services, and ensuring the design could withstand scrutiny from senior stakeholders and security teams.

As one of the department’s early AI initiatives, the solution was designed with a strong emphasis on responsible use. AI supports the drafting of initial responses, but civil servants remain fully accountable for reviewing, editing and approving all correspondence before it is sent.

We also helped ensure the platform could be extended sustainably. New briefing packs can be added through a simple, structured process using Azure-managed services, allowing the correspondence team to expand the tool to cover additional subjects over time without significant redevelopment or ongoing support.

Result

With the Correspondence Drafter in place, the drafting process has been significantly streamlined. Upon receiving correspondence, the drafter selects the relevant AI assistant and enters the enquiry into the tool. A first draft response is generated using approved briefing content, which is then quality-assured by the correspondence team before being sent.

In testing, this approach has significantly reduced the time taken to produce a first draft for supported correspondence types. The tool is currently in private beta on Microsoft Azure and has received positive user feedback.

Impact

By supporting the Department for Education to introduce one of its first production AI services, Made Tech helped demonstrate how AI can be adopted pragmatically within government. The Correspondence Drafter shows how carefully designed AI tools can improve productivity while preserving trust, assurance and accountability, enabling civil servants to focus more of their time on higher-value work.

The post AI-powered efficiency at the Department for Education appeared first on Made Tech.

Awaab’s Law introduced strict new requirements for how social landlords respond to hazards like damp and mould. Among these is the obligation to provide tenants with a written summary of investigation findings within three days.

This creates entirely new work. Providers were already investigating hazards and carrying out repairs, but producing a written, tenant-friendly summary within a fixed timeframe and delivering it reliably is new work. And it’s work worth doing – tenants deserve clear communication about what was found and what happens next. The challenge isn’t the requirement itself; it’s finding a way to deliver it consistently at scale.

Different providers have approached this in different ways, each with limitations:

Paper summaries left on-site: Some surveyors write a summary by hand and leave it with the tenant during the property visit. While quick, this approach lacks an audit trail – there’s no proof the summary was delivered. Handwritten notes can be lost or illegible, and tenants who don’t read English or have accessibility needs may not be able to understand them.

Back-office team writes it up: Others have surveyors send their notes to a back office team, who then draft a tenant-friendly summary and deliver it by email or post. For a provider handling ~100 damp and mould cases a month, that’s typically around 10 minutes per summary – or ~17 hours of back-office time every month spent drafting and sending summaries by hand (before any chasing, formatting, or rework)

Approach

When Made Tech built Hazard Case Management for Awaab’s Law, AI-powered summary generation was included from the start – not bolted on as an afterthought. The approach: use AI to draft tenant-friendly summaries automatically, with human oversight before delivery.

AI-generated draft

When an investigation is completed, the system uses a large language model to read the survey data and generate a clear, tenant-friendly summary. This isn’t data being pasted into ChatGPT – the LLM runs on Made Tech’s own UK-based infrastructure, and tenant data is never used to train models. This eliminates the manual drafting step while ensuring consistent quality.

Human review and editing

Once the AI draft is generated, the case is automatically flagged as ‘summary ready for review’, making it easy to see which cases need attention. Staff review and can edit the content if needed, maintaining quality control while avoiding the time-consuming blank-page problem.

Audit trail included

Every action in the platform is audited: when the surveyor’s report is added, when the summary is generated, when it’s reviewed and published, and when tenant notifications are sent. This gives them a complete timeline and clear evidence of compliance if challenged.

Digital delivery with accessibility built in

Whether the tenant is notified via letter, email or SMS, they’re given a unique URL to view their investigation summary online. Because the page uses the GOV.UK Design System, it works well with browser translation tools, so tenants can read their summary in their browser’s default language.

As a modern local authority, we’re committed to driving innovation that strengthens our communities and delivers long‑term value. This initiative represents exactly the kind of forward‑thinking partnership that helps us create resilient places, unlock new opportunities, and ensure that every resident benefits from a more connected, sustainable future” – Leona Menville, Strategic Director Housing, Environment & Communities

Results

A week of admin time saved every month

727 summaries generated in the first three months – an estimated 121 hours of manual work avoided. That’s equivalent to nearly a full week of admin capacity recovered every month.

Consistent, compliant documentation

Every summary follows the same structure and tone, reducing the risk of inconsistent communication or missing information.

Better communication for every tenant

Summaries are delivered digitally and are readable on any device. GOV.UK Design System compatibility means tenants can instantly translate the page into their preferred language – no letter to wait for.

Awaab’s Law created new compliance obligations that many providers are still figuring out how to meet at scale. By combining AI-powered drafting with digital delivery, LBBD has turned a potential admin burden into a streamlined process – saving time, improving accessibility, and building a clear audit trail.

Learn more about Hazard Case Management

The post How Barking & Dagenham saves a week of admin every month with AI-powered Awaab’s Law summaries appeared first on Made Tech.

We seek opportunities to partner with organisations delivering truly impactful work. Skills for Care is the leading source of adult social care workforce intelligence in England. They work tirelessly to provide the government with up-to-date metrics on the state of England’s adult social care sector. This includes how many care workers there are and how they’re distributed throughout the country. A task that’s made all the harder when you realise there is no legal requirement for any privately owned care home to report on these statistics.

As you might imagine, this is a mammoth task. When we first landed, the analytics group was performing the heroic feat of collating, cleaning and analysing years of data. From tens of thousands of facilities through a semi-manual and time-consuming process. The process was taking them months to complete and they knew they weren’t using their data to its full potential. They knew technologies existed that could help, but implementing them required skills and experience they didn’t (yet) have. We were eager to help them with that.

Our approach

To deliver the best value, we set out to build a solution that was both easy to maintain and powerful enough to handle their vast data collection. We also needed to make sure this was something they could continue developing once we’d left. Making upskilling the Skills for Care analysts from near-zero coding experience vital to success. Once these things were in place, it was clear that the deep sector knowledge the analysts held could be put to work – improving their reporting. 

Our data engineers wasted no time building a data pipeline. One that would automatically store, version, clean, join and transform – making sure all data was ready for analysis. The Skills for Care analysts leapt at the opportunity to develop new skills. They quickly became valuable partners in the development process by identifying opportunities for improvement and contributing significantly to the codebase.  

We demonstrated the importance of test-driven development, the power of CI/CD and the necessity of generating effective, usable documentation throughout.

By the end of the project, the team had turned a task that usually took about a month to complete into an automated process that runs every 2 weeks and takes about 20 minutes.

While this was happening, our data scientist was interrogating their data. Uncovering trends and developing new processes designed to extract as much insight as possible. They held regular showcase sessions. Presenting new analyses to shine fresh light on the trends that Skills for Care knew to be true but had never had the data to demonstrate. By working closely with their analysts we developed new machine-learning models that enable much more granular reporting. 

To make sure this bump in accuracy remains (knowing the efficacy of all models decays over time), we built monitoring steps into the pipelines. As soon as model accuracy metrics fall outside of a set range, analysts are notified and re-training jobs are triggered. Deploying a model and simply assuming it will continue to be as performant as the day you deployed it is never enough. 

Results we achieved together

The impact of this work cannot be understated. Through this partnership, Skills for Care has gained a wealth of new development and data science technical ability. They now possess a suite of bespoke, fully integrated AI models that deliver data-driven insights all the way to the highest levels of government. This transformation has allowed them to rethink their data collection process. And has resulted in a massive increase in data quality – compounding their improved reporting capability.

Skills for Care now perform complex analysis using version-controlled Python scripts and notebooks.

This allows them to access a whole world of open-source analysis tools and frees up their time to perform other aspects of their job.

Most importantly, this work will enable Skills for Care to produce their most accurate reports to date – starting from November this year. These enhanced reports will empower Ministers and MPs to deliver data-driven insights and decisions about the care sector. Leveraging modern data principles and AI, Skills for Care has significantly improved the quality and impact of their reporting.

Data and AI

Use data-driven insight to design and deliver better services to users and unlock the power of AI and Machine Learning.

Health and life sciences

As patient numbers rise, digital technology can help you lighten the load.

Speak with an expert

The post Improving data reporting in adult social care appeared first on Made Tech.

Forecasting weather requires deep scientific knowledge and vast amounts of computing power. World-leading experts, the Met Office are constantly improving their forecasting process – to help us stay safe and thrive. But how can they measure the success of a new weather model? It’s one thing to know that your model is more scientifically accurate. But when it comes to data-driven insights on how users respond to and understand that information – it’s a whole different ball game.

By using data and AI to automate the sentiment analysis of vast amounts of user feedback, we allowed the Met Office to extract patterns and trends with an efficiency that was previously impossible.

We knew we could then combine this with traditional methods of user research that provide in-depth, qualitative insights into user behaviours.

Our data science team helped the Met Office interrogate hundreds of thousands of individual pieces of user feedback. These came from from millions of daily users of the Met Office public website and mobile apps. Our challenge lay in extracting a needle of insight from this haystack of information. Together we set out to demonstrate if a new weather model was more trusted than previous iterations.

Our approach

When we joined, a sole team member was reading hundreds of comments every day on top of their day job. They were trying to extract insight from multiple spreadsheets and data sources. Our data scientists quickly recognised this tiresome task was a perfect opportunity for AI.

To begin, there were conversations around which technology to use. Should we jump on the large language model (LLM) bandwagon, or use a simpler, more tailored solution?

Despite the hype in the industry, we knew an LLM was not appropriate for this use case. Due to the complexity of these systems, their unpredictable behaviour and the large cost associated with the compute. We recommended in this early stage to use well-established natural language processing (NLP).

Following the advice of our Made Tech colleagues who are experts in this field, we opted for a tried and tested NLP model. As this would be best to categorise the thousands of items of accuracy feedback we receive. The efficiency of this system and the amount of time it has saved extracting insight from our feedback has been invaluable.

Fleur Layzell
User Journey Manager

To get started, we broke this task into 3 problems:

• how do we feed this insight back to stakeholders?
• how do we collect and process so much data?
• what insight can we extract from user comments?

In order to process the hundreds of thousands of pieces of user feedback, we built an automated analytics pipeline. This pipeline had to be easy to maintain and robust enough to support computationally heavy machine learning (ML) tasks. It also had to safely strip any personal information that users might have submitted. 

Building this platform centralised a single source of truth for all user feedback. We were then able to begin sifting through comments and developing analytics processes. It was in this space that the majority of data cleaning, enriching and preparing processes were carried out.

This central analytics environment instantly began closing the feedback loop between stakeholders and users, empowering the Met Office to begin asking for bespoke pieces of analysis and digging into feedback following feature releases. We presented their user feedback in a completely new light. We helped visualise user distributions, target specific use cases and report on different weather scenarios as they happened. Something that had not been possible before. 

We brought in junior developers to the team to provide valuable experience on at greenfield project. By aligning to the Met Office’s strong culture of excellent people, culture and developing capabilities, the team were able to provide a supportive space to learn, upskill and share knowledge across the team and wider Consumer Digital Programme.

Through their hard work and dedication, our colleagues at Made Tech have created a robust data pipeline. And in doing so, a more efficient way for the Met Office to not only monitor accuracy and cost reporting, but open the door to other areas where accuracy and potentially more general feedback can be monitored. This increasing our validity as an organisation where user-centred experience is at the heart of everything we do.

Fleur Layzell
User Journey Manager

Results: Data-driven insights at the Met Office

Improving the response to user feedback

Thanks to the development of the feedback analytics platform, the Met Office can now actively monitor user feedback and measure the perceived success of new weather models in real time. Stakeholders and product owners can track user sentiment throughout the rollout and precisely quantify the impact of any changes they make to the service.

They’re able to monitor the impact of different weather systems, check model performance across different regions and environments and rapidly respond and iterate on issues as they arise – reducing the feedback-to-development loop from weeks to days. All focused on providing the best service possible for users.

User feedback analytics platform

The new analytics platform empowers the Met Office to conduct more types of data analysis. We’re now able to use powerful NLP techniques to categorise comments, allowing a shift from anecdotal to actual evidence. We can visualise and monitor feedback as it lands, and slice it however needed. For example, if the weather science team needs to know how users in Region X responded to forecast Y,  this is now achievable with the click of just a few buttons.

By presenting this analysis in responsive, live dashboards, Made Tech has enabled an interaction with Met Office user data like never before. 

Moving forward

AI technology continues to progress at a lightning pace. Smaller, open-source generative AI solutions are now far more available, with performance and reliability nipping at the heels of the larger closed-source models. The next steps for the Met Office will be to further enhance this analysis platform by generating automated, summary reports, condensing a never-ending river of user feedback into easy-to-understand, direct points of action. 

This is an excellent example of delivering value to our public sector clients. By working in partnership with us, we have led the Met Office to the adoption of new technologies that harness the power of their own data.

Data and AI

Use data-driven insight to design and deliver better services to users and unlock the power of AI and Machine Learning.

Central government

As society becomes more digitally driven, government needs to meet citizen’s raised expectations by applying the culture, processes and technologies of the internet era.

Speak with an expert

The post From feedback to forecast at the Met Office appeared first on Made Tech.

Since 2006 Skills for Care has been collecting and analysing data on the thousands of adult social care workforce across England. Providing data to the Department of Health and Social Care helps better inform the decisions made by government when it comes to the care sector.

Many of their data analysis processes were manual, taking considerable time and not maximising data-driven insights. We worked with the team to transform their data processes to better provide intelligence quickly and accurately.

Our approach

We implemented the Amazon Web Services (AWS) data ecosystem. This helped us modernise Skills for Care’s data architecture and processing. The goal was to streamline processes, automate routine analysis and free up valuable time.

Working with the analytics team we helped put in place best practices in data engineering. We also worked with them to apply machine learning to improve their workforce estimates.

The next stage was to help Skills for Care identify what was needed to continue to develop this new data pipeline long after we left. They’re now fully prepared to continue to build their pipeline as new data becomes available.

The data engineering prototype developed in this project is the first step toward a new way of working. It will help us to report more frequently and accurately to support faster and better decision-making.

Will Fenton
Analytics Manager

Results we achieved together

Crucially the team can now import data from many data sources automatically. This has helped enrich their data, fill data gaps and review at a more granular level – improving data quality. Now that there’s no need to manually process data that team are saving time and effort. Analysis that used to take 2 days can be done in 15 minutes and the datasets available for review are also extended from one to 10 years. Data visualisations can now also better inform analysis, making it much easier to see trends over time.

For Skills for Care it’s now increasingly quick and easy to test new ideas. They are better prepared to identify and fix issues earlier when data changes outside of the organisation happen.

Related services

Data and AI

Use data-driven insight to design and deliver better services to users and unlock the power of AI and Machine Learning.

Digi­tal serv­ice delivery

We support you at every point in the delivery lifecycle, right from defining the problem right through to the live services that solve them.

Speak with an expert

The post Data-driven insights within adult social care appeared first on Made Tech.