Jethwa, a graduate of the University of Oxford in engineering science, focused on niche areas such as robotics and artificial intelligence (AI). He then refined this interest with a doctorate from MIT, where he researched how images could create models of urban spaces. It was a little like coming home when he got the chance to work at OS 20 years later.
He says that “part of the loop had been reconnected.” “I had been working on city-scale models. Jethwa, who joined OS in 2023, was already working on mapping the entire nation.
Jethwa had been CTO at construction software specialist Causeway Technologies since August 2022, following the company’s acquisition of asset management software firm Yotta. He was CTO of construction software specialist Causeway Technologies. He held this position since August 2022 after the company acquired asset management software firm Yotta. Jethwa was the product and technology leader at Yotta. He had been there for 20 years. He puts his data expertise into practice as OS CTO.
He says that AI is now a part of every company’s business strategy. “AI and computer-vision are core parts of the research that we do here. For example, automatically extracting features from street level and aerial imagery.” This is not surprising for someone who has been in entrepreneurial businesses before joining an organisation with a 234-year history.
He says, “We’ve had to make some changes to our strategy, having moved from a business that was product-driven to one that was more focused in delivery.”
Creating value chain
Jethwa said his focus on organisational changes has helped break OS down into a chain of key service elements that deliver benefits to the customers.
There are five pillars in the organisation that cover: positioning, which is the ability to accurately measure the UK; data sourcing, which is the aggregation and collection of information; refinery, which adds value to the data and merges insights; distribution, or pushing data feeds to customers; and application.
This value chain has been built from sourcing the data to delivering the benefits to customers. Customers don’t care about lightbulbs; they are interested in light. Customers at the OS don’t want data; they want insights. This model helps us define our approach.
Manish Jthwa, Ordnance Survey.
He says: “We have built up this value-chain from sourcing data to providing benefits to customers. Customers at OS don’t want data; they want insights. This model helps us define our approach.”
Jethwa said the result was lots of conversations about value chain. “I’m wheeled out and asked to speak about it,” says Jethwa. “But it’s an easy way to visualize the work and relationships within a 1,400-person organisation. We now have a high level structure to explain how our organisation operates.”
Jethwa claims he cannot take the sole credit for the organization change programme. His arrival was preceded by a drive for transformation. His focus was on defining clear, business-oriented services. OS now has a service-based model for its operation that spans the five key pillars.
He says that “we have service leads in those areas and these services are broken down further into components.” “The nice thing about this approach is that we can manage our entire technology stack within these components using the same framework.”
Focus on product delivery
Jethwa explains that the new organisational structures have enabled a move away from project-based working. Instead of focusing on small initiatives that solve short-term problems, the IT team aims to achieve objectives as part an enduring program of work.
He says, “Staff submitted a program proposal that was accepted.” “You set a budget and form a team. Then you execute.” To deliver a long term vision, you usually have enduring teams who work on your technology product. They understand the customer, know how to deliver the results, and also manage systems.
Jethwa said the result of this change from projects to products was a fundamental shift in how OS functions. “These enduring team have a defined vision of what they must deliver and how results will be communicated to internal or external clients,” he says. “They have the power and opportunity to determine the best way to deliver products.”
Jethwa said the teams always focus an element of transformation. He is referring to the development and creation of products within the sourcing sector. The professionals are evaluating the best way to collect data, whether that is through drone capture or street level surveys. He says that the approach to sourcing changes constantly depending on the landscape, and the features we want to deliver.
The challenge for us is the fact that our data will not be delivered in a specific year and then we are done. We have to deliver updates and maintain consistency until we decide to retire the product, which could take at least 10 year. It could take longer.”
The requirement for continuous delivery presents a new set challenges, including dealing with emerging technologies and extracting features and information from source imagery. Jethwa believes AI can automate these processes. AI-enabled technology must not make assumptions about images.
He says that AI can increase productivity and efficiency. “But an automated method of analysing images can easily be fooled. We must therefore consider how we can manage uncertainty when we provide insights to our customers. This area presents us with a new challenge.”
Adopting technological innovation
Another concern is technical specifications. Jethwa said OS wants to democratise the data access for customers. The premise is simple but the process has to overcome some obstacles. For example, most data come with detailed technical specifications.
The explosion of AI is a factor in every business strategy. AI and computer vision is a core part in the research we do [at OS]such as automatically extracting street features and aerial imagery.
Manish Jethwa from Ordnance Survey.
A good example would be a customer asking, “I want to know where I can go to exercise.” “Our data, from a technological standpoint, might not refer to the word exercise anywhere. But it will include things such as sports stadiums and skateboard park.”
Jethwa said the traditional approach to building the connection between data specifications and technical specifications would use natural language processing. It would be difficult to build that mechanism. Large language models (LLMs), however, are ideal for creating a relationship with data and specifications.
If you fine-tune an agent for a large language and point it at our data, you can ask questions like, “Tell me where to exercise” and have the question translated directly into an API [application programming interface] Request that will say: ‘Here are sports facilities in the building data and here are recreational areas in green spaces data,’ he says.
Jethwa claims that OS is utilizing the full range of AI technologies. The organisation bases many of its processes off mainstream models and fine-tunes them based on documentation. One significant issue is that high-profile LLMs receive training using commercially available data.
In order to overcome this challenge, OS uses the high-precision UK geographic data that its teams have gathered over decades of work. The technology team is building models from scratch in this core area where the organisation extracts geographic information, such roof materials or biodiversity features.
He says that the UK landscape is different from other geographies. “We know that relying on models trained on images taken directly from the internet will not give us a reliable and richer output.”
Leading the way
Jethwa said senior executives at OS discuss regularly how people may interact with geospatial information during the next decade. One thing is clear: the relationship has changed and will continue changing.
OS is known for its paper-based maps which help people navigate through their environment. These maps are designed so that people can filter the data they need. Jethwa uses the example of a hiker who can understand the elevation on a map by looking at contour lines.
Looking ahead five or ten years, people will have a much more conversational relationship with the map and interface
Manish Jethwa from Ordnance Survey.
But the traditional relationship is changing. Instead of focusing on filtering the organisation’s focus is now on analysis. OS is developing AI processes that customers can use to gain insights from data. Jethwa believes that the nature of this relationship will intensify.
If you look five or ten years ahead, the relationship will be one where people have a much more conversation with the map and interface,” he says. “This relationship will allow customers to ask questions and receive a response from the map. They’ll also be able to ask questions of the map.” OS also uses Snowflake Marketplace for sharing open data with public sector organizations and exploring new commercial avenues. OS must be ready to embrace data-driven advances.
We need to make sure we’ve enabled the mechanisms internally. How easy is it to receive answers and ask questions via APIs? Are we supplying the data from the right sources – whether Snowflake, AWS, or elsewhere? “Are we providing hooks to the right outlets so that insight can be accessed as models are being built?” he asks.
We don’t want be late to the game, so we need to set the pace. If we are late, other data providers will have more authoritative data, but they’ll also be easier to access. We need to be in the forefront.”
Photo: (c)Crown Copyright 2025 Ordnance Survey. Media 045/25 (19659046)
