Recordsure develops AI-powered RegTech solutions for clients in the financial services, regulated industries and public sector.
It wanted to unify two distinct tech stacks and chose to migrate wholly to Microsoft Azure to reduce costs, and to access AI innovation faster.
Recordsure’s production environment costs are now a quarter of what they were, and innovation cycles have accelerated dramatically, with new products reaching production in as little as two weeks.
We liked Microsoft’s open, commoditised approach. Frankly, we were convinced that Microsoft was going to win in the areas of AI we care about.
Kit Ruparel, CTO, Recordsure
“Recordsure was created as an AI company from the get-go. Considering we are 13 years old; we have a strong market presence and a well-established track record with clients and regulators,” says Joe Norburn, CEO at Recordsure.
The business was founded on the understanding that the quality of financial advice is directly correlated with how well the person giving it understands their client and their circumstances. Recordsure applies advanced analytics and machine learning to the recording of client interactions to deliver enduring, high-quality transcription and provide summarisation, compliance analysis and performance insights. To tackle the challenges of documentation reviews and evidence, Recordsure designed thorough automated AI processes and workflows that save clients’ review time, ensure compliance and flag process inconsistencies, with superior speed and accuracy.
This unique, rapidly deployable technology enables Recordsure clients to ensure compliance – both with relevant laws and regulations and with the processes and experiences the organisation wants to deliver to its customers. In this way, financial services firms and regulated companies can underpin and drive best practice – and deliver good customer outcomes and excellent client interaction.
When Recordsure first launched its RegTech solution over a decade ago, most financial services companies were reluctant to use cloud software and services. To meet its clients’ needs, Recordsure ran its software as an on-premises solution, with its own environments secured in a colocation data centre.
As the sector has digitalised, Recordsure expanded from recording face-to-face interactions to also deliver the same recording, compliance and analysis capabilities for video meetings and call centres. By this stage, attitudes in the industry were changing and Recordsure’s customers were beginning to consider cloud solutions.
“At that time, the machine learning capabilities we were using were largely open-source. We were using Linux and Python, so we gravitated to a large hyperscaler where we felt those technologies were at home,” says Kit Ruparel, Chief Technology Officer at Recordsure.
Initially built for conversation analytics, the next evolution for Recordsure was to apply its compliance analysis capabilities to documents and digital artefacts so that its solutions span all client interactions. This has delivered significant commercial benefits for its clients.
Kit Ruparel explains, “By the time we launched our AI-led document analytics solution seven years ago, our customers were more comfortable with cloud. And, because we would be working with documents and emails – which are Microsoft’s domain – we built on the Microsoft Azure stack.”
“At the time, we had two different engineering teams, two sets of data scientists – it wasn’t efficient internally and I couldn’t easily share resources to the application where they were needed,” recalls Kit Ruparel. “We needed to find a better, more integrated and efficient way of working.”
“We needed to bring our two technology stacks, conversation and document AI analytics, together – for a unified product architecture as well as a unified technology architecture,” recalls Joe Norburn.
“If we could analyse voice, documents and data together, it would open opportunities for much richer product capabilities,” confirms Kit Ruparel. “Plus, it made sense to consolidate from an engineering perspective.”
By reducing engineering and operational technology costs, Recordsure sought to reduce the price of its solutions and lower the cost of entry. Joe Norburn reveals, “This way, more organisations can benefit from Recordsure’s capabilities to drive compliance, best practice, and better outcomes for their customers.”
Determined to consolidate, Recordsure reviewed global cloud providers to assess which service would best meet the organisation and its customers’ needs.
Microsoft helps us to understand what emerging best practice looks like – and puts the tools in our hands to innovate.
Joe Norburn, CEO, Recordsure
Recordsure saw that by moving its entire stack into Microsoft Azure, it would be able to access the latest and greatest AI solutions while optimising management costs.
“Microsoft speech recognition and cognitive search were very much on our radar at that point,” says Kit Ruparel. “Microsoft was embracing the open-source world and was prioritising machine learning. Plus, Microsoft was taking a very open approach to AI. For example, Microsoft Azure AI Foundry brings together many commercial and open-source AI components into a single, easy to manage interface with pay-as-you-go pricing.”
Kit Ruparel confirms, “We liked Microsoft’s open, commoditised approach. Frankly, we were convinced that Microsoft was going to win in the areas of AI we care about.”
The consolidation on Azure gave Recordsure an opportunity to rethink its solution architecture which the team was determined to exploit.
“Moving to consuming cloud as a platform on a pay-as-you-go basis would enable us to get real cost savings and lower the barrier to entry for our customers,” advises Kit Ruparel.
Craig Connolly, Development & QA Manager at Recordsure, agrees, “We want to properly leverage cloud features following the design principle of ‘scale to zero’. Being on Azure makes that easy for us.”
“Azure is a turnkey solution for us – everything is in one place,” continues Craig Connolly. “Azure Kubernetes Service (AKS) has been a game changer for us. We can be much more scalable and stable. Once we made that transition from VMs to AKS, we transitioned to Azure DevOps and Azure Container Registry. Azure Key Vault security means less local configuration. Azure Blob Storage ensures we have accessible, scalable storage. And we’ve integrated single sign-on for our products with Microsoft Entra ID which has delivered fundamental benefit for our clients.”
“As well as containerisation, we’re using Azure Function Apps in certain parts of our architecture,” adds Craig Connolly. “We have a reporting function app using embedded Power BI. This means the Power BI licencing is absorbed by us so it’s a lot simpler for clients to onboard with us – they can enjoy the reporting out of the box.”
“In Azure, everything just works together,” agrees Kit Ruparel. “In the open-source world, it’s up to you to lash everything together and make it work. You need to manage versions and integrations and deprecations, which is super challenging. Instead, the seamless integration on Azure saves us oodles of time and scarce resource capability.”
Craig Connolly emphasises, “Using Azure like this means we can offer our clients a solution that is faster, more scalable, more secure and at a better price point for them which, ultimately, makes us more marketable.”
Recordsure’s first foray into native Azure AI tooling was with Azure AI Speech for transcription services and Azure AI Search for cognitive search.
“It really blew us away with the performance,” reveals Sarah Airey, Head of Data Science at Recordsure. “From that first introduction, we explored the whole suite of tools available to us on Azure.”
For Recordsure, the speed at which new technologies and tooling become available on the Azure platform is transformational.
Sarah Airey says, “When my team are excited about a new data science or Azure Machine Learning technique and want to try it out, all we need to do is look on Azure and, invariably, there are several different options available. This on-demand access to all the latest tooling makes experimentation very easy.”
“Microsoft was the first to commoditise Generative AI for secure, enterprise use,” enthuses Kit Ruparel. “We went from thinking Generative AI was going to be too expensive to integrate into our products, to our customers to being able to experiment with it and use it on a pay-as-you-go basis. Plus – and as importantly, to use it in combination with our traditional predictive AI, all in the same Entra Directory,”
Sarah Airey agrees, “We can access different Azure OpenAI and GPT models at different cost points and experiment to choose our entry point. We can test whether we need to go in at the high end to get the results we need or whether lower priced models will suffice.”
“This has really reinforced that our decision to go with Microsoft was the right one,” Kit Ruparel continues. “And that trend continues. With every new release there’s something we can use on a pay-as-you-go basis – and that has had a huge impact on the speed at which we can innovate.”
“The access to compute power that we have on Microsoft Azure is the other thing that has freed up my team to innovate,” adds Sarah Airey. “We used to have to work within the constraints of the compute power we had. With Azure, we can access the compute power and GPUs as we need them, while setting cost boundaries and budgets. It really liberates us to experiment.”
The freedom to experiment and try new models enables Recordsure to deliver new features and innovation to customers faster.
“I think the best time to delivery we have now is two weeks,” reports Sarah Airey. “Previously, we just wouldn’t have been able to turn around new products this quickly.”
“The hastened speed of innovation aids rapid client deployment – this means Recordsure’s clients can start reaping the benefits of using our AI analytics faster than ever before,” emphasises Joe Norburn.
Production environment costs 25% of what they were
“There’s a notable security angle to this as well,” states Kit Ruparel. “When different customer environments were in our data centre, we had to get customer permission to move their data out of their environment, bring it to the GPU AI training server, work on it, move it back and delete it. Now, with Azure on-demand computing, we fire up a GPU in the environment in which the customer’s data resides. Rather than bringing the data to the GPU, we bring the GPU to the data – and that’s far more secure.”
“Similarly, we can spin up secure environments for our annotation team to collect our training data. It’s fast and they’re in the same data security environment,” adds Sarah Airey. “That safe space on Azure means we can get our training data quicker and have production-ready models faster.”
Alongside the agility and security benefits, the on-demand model has delivered significant cost savings for Recordsure.
Kit Ruparel advises, “With our previous cloud provider, the production environment of a single tier one bank with relatively low usage was costing us four times the current monthly outlay on Microsoft Azure of our entire production environment serving all our customers.”
“Engineering is much more efficient too,” Kit Ruparel continues. “Instead of my team managing the scaffolding and boilerplate architecture, we can focus on innovation. Azure AI speeds innovation at an affordable price.”
In addition, working with Microsoft delivers cybersecurity and data governance benefits which create commercial advantage for Recordsure.
Joe Norburn says, “The enhanced security benefits are very helpful in our client conversations.”
“Every financial services company has now made the switch to cloud. They’ve mentally switched to say ‘yeah, Microsoft knows what it’s doing with cybersecurity!’.” explains Kit Ruparel. “And, importantly, using Microsoft enables our solutions to be much more in tenancy. The data isn’t leaving the client’s tenancy. It’s still resident in the geographic area of each client’s choice. That is key for us.”
“When you’re working with regulated industries and central or local government, it can take a while to work through the security considerations,” Kit Ruparel continues. “Working with very confidential documents in government or financial services is much easier when you can assure the client that you’re working with the same technology provider that is running their SharePoint. It simply clicks with the business users and makes their security teams happier. As they’re already running emails in Microsoft 365 and use Microsoft Azure themselves, it makes the solution selling easier, especially when we leverage Microsoft’s co-selling support.”
“The benchmark in our sector is higher than anywhere else, that’s why we need our solution to deliver the high level of accuracy required in financial services and Government,” says Joe Norburn.
“Productisation on the Microsoft stack means we are able to deliver far more consistently and with a higher quality outcome. This remains critical as we continue our expansion internationally and into adjacent markets – not least being able to work in different languages, beyond English. It gives us the ability to go broader, faster.”
“We have exceptional R&D and data science teams and punch above our weight in terms of what we do. But we can’t be across everything,” continues Joe Norburn. “Microsoft helps us to understand what emerging best practice looks like – and puts the tools in our hands to innovate.”
Using Azure like this means we can offer our clients a solution that is faster, more scalable, more secure and at a better price point for them which, ultimately, makes us more marketable.
Craig Connolly, Development & QA Manager, Recordsure
Follow Microsoft