Jethwa studied engineering science at the University of Oxford, where he focused on what were then considered niche areas, such as robotics, artificial intelligence (AI) and computer vision. He subsequently honed this interest in a doctorate at MIT, researching how images could be used to create models of city spaces. The opportunity to work for OS two decades later felt a bit like coming home.
“Part of that loop had been reconnected,” he says. “I’d been in the world of creating city-scale models. And then, before I knew it, I was helping to map the whole nation at the OS.”
Jethwa joined OS in late 2023. He was previously CTO at construction software specialist Causeway Technologies, a role he’d held since August 2022, following the company’s acquisition of asset management software firm Yotta. Jethwa had worked at Yotta for 20 years, leading product and technology. As OS CTO, he’s putting his data experience into practice.
“The explosion in AI means it’s in every organisation’s business strategy,” he says. “AI and computer vision are a core part of the research we do here, such as automatically extracting features from aerial imagery and at street level.”
For the majority of his 18 months at OS, Jethwa says he’s been more focused on the “chief” part of his CTO role than “technology” because he’s helping to drive wide-scale organisational change. That focus is understandable for an individual who has worked in entrepreneurial businesses and joined an organisation that’s 234 years old.
“Having come from a product-led business to an organisation that was more focused on delivery, we’ve had to change certain things, and that’s been reflected in our strategy,” he says.
Creating a value chain
Jethwa says his focus on organisational change has helped break down OS into a value chain of key service pillars that deliver benefits to customers.
The organisation has five pillars that cover the following areas: positioning – meaning the capability around the UK to make accurate measurements; data sourcing – aggregating and collecting information; refinery – adding value to data and merging insights; distribution – pushing data feeds to clients; and application – delivering insight to customers.
We’ve built up this value chain from sourcing data to delivering benefits to the customer. Customers don’t want lightbulbs, they want light. At the OS, customers don’t want data, they want insights. Having that model helps define our approach Manish Jethwa, Ordnance Survey
“We’ve built up this value chain from sourcing data to delivering benefits to the customer,” he says, before using an analogy: “Customers don’t want lightbulbs, they want light. At the OS, customers don’t want data, they want insights. Having that model helps define our approach.”
Jethwa says the result is lots of conversations about the value chain. “I get wheeled out to talk about it,” he says. “But it’s a nice way to visualise the work and relationships of an organisation of 1,400 people. We now have a high-level structure to explain how we operate.”
Jethwa says he can’t take sole credit for the organisational change programme. A drive for transformation pre-dated his arrival. His emphasis was on defining clear business services. As a result of this work, from the start of April, OS has a service model for operation across the five key pillars.
“We have service leads across those areas, and those services are broken down into further components,” he says. “The nice thing about that approach is that, within those components, we can manage our technology stack using the same framework.”
Focusing on product delivery
Jethwa says the new organisational structure has enabled a shift from project-based ways of working. Rather than focusing on smaller initiatives that fix short-term issues, the IT team looks to meet objectives as part of an enduring programme of work.
“Staff put forward a programme proposal that will then get accepted,” he says. “You have a budget, you formulate a team, you execute against that. But to deliver a long-term vision, you typically have enduring teams that work on your technology products, so they understand the customer, they know how to deliver results, and they also manage systems.”
Jethwa says the result of this shift from projects to products is a fundamental change in how OS operates. “These enduring teams have a specific, defined vision about what they need to deliver, and how the results will be passed on to internal or external customers,” he says. “They have the opportunity and power to work out the best way of delivering products.”
Jethwa says the teams always focus on an element of transformation. He refers to the development of products within the sourcing pillar. The professionals consider how to collect data and the optimal way of fulfilling that process, whether it’s through drone capture or street-level surveys.
“The approach to sourcing shifts continuously depending on the landscape and the types of features we want to deliver,” he says.
“The challenge for us is that our data is not delivered in one specific year, and then we’re done. Once we’ve delivered that data, we have to provide consistency and deliver updates henceforth until we decide to retire a product, which may take at least 10 years. It could be longer than that.”
This requirement for continuous delivery creates a new set of challenges, such as dealing with the roll-out of emerging technologies and extracting information and features from source imagery. Jethwa says AI can help automate these processes. However, care must be taken to ensure that AI-enabled technologies don’t make assumptions about images.
“Yes, AI can boost productivity and efficiency,” he says. “But an automated method for analysing images can be fooled. So, we must consider how to manage uncertainty as we deliver insights to customers. That area presents a fresh challenge for us.”
Embracing technological innovation
Another issue is technical specifications. Jethwa says OS wants to democratise data access for customers. While the premise is simple, the process must clear some obstacles, notably the fact that most data comes with detailed technical specifications.
“The explosion in AI means it’s in every organisation’s business strategy. AI and computer vision are a core part of the research we do here [at OS], such as automatically extracting features from aerial imagery and at street level”
Manish Jethwa, Ordnance Survey
“A good example might be a customer who says, ‘I want to know good areas to go out and exercise’,” he says. “Now, our data, from a technical standpoint, might not reference the word exercise anywhere, but it will include things like sports stadiums, fields and skateboard parks.”
Jethwa says the traditional approach to building a connection between data and technical specifications would be to use natural language processing. Building that mechanism wouldn’t be straightforward. The good news now is that large language models (LLMs) are perfect for creating a relationship between data and specifications.
“If you fine-tune a large language model agent, and point it to our data and documentation, you can ask a question, such as, ‘Tell me about places to exercise’, and have that question directly translated into an API [application programming interface] request that will say, ‘Here are some sports facilities in the buildings data and here are some recreational areas in the green spaces data’,” he says.
Jethwa says OS is tapping into the full breadth of AI technologies. The organisation builds many of its processes on mainstream models, undertaking fine-tuning based on internal documentation. However, one significant issue is that many high-profile LLMs are trained on commercially available data.
To overcome this challenge, OS taps into the high-precision UK geography data its teams have collected over decades of work. In this core area, where the organisation extracts geographical information, such as roof materials or biodiversity features, the technology team is building foundational models from the ground up.
“The UK landscape is very different from other geographies,” he says. “We know that training models on our data will give a richer and more reliable output than trying to rely on models that have been trained on imagery taken directly from the web.”
Leading the way
Jethwa says senior executives at OS regularly discuss how people might interact with geospatial data during the next decade. One thing is already clear – the relationship has already changed and will continue to transform.
OS has traditionally been known for producing paper-based maps that help people navigate their environment. These maps are designed to make it easy for people to filter the data they require. Jethwa gives the example of someone on a hike who can see contour lines on a map and understand the elevation of their walk.
If you look ahead five or 10 years, the relationship has to be one in which people have much more of a conversation with the map and the interface Manish Jethwa, Ordnance Survey
However, the traditional relationship is changing. Rather than focusing on filtering, a large proportion of the organisation’s work is now centred on analysis. OS is developing AI-enabled processes that customers use to draw insights from data. Jethwa expects the nature of this interaction to intensify.
“If you look ahead five or 10 years, the relationship has to be one in which people have much more of a conversation with the map and the interface,” he says. “This relationship will be one where customers ask a question and get a response back from the map, but they’ll also get questions from the map.”
Jethwa says emerging technology is the key to unlocking this capability. Alongside explorations into AI, OS uses the Snowflake Marketplace to share open data with public sector organisations and explore new commercial avenues. The priority now is ensuring OS is ready to embrace the next wave of data-led advances.
“We need to ensure we’ve enabled all the mechanisms internally. So, how easy is it to ask the questions and receive answers via APIs? Are we providing the data across the right sources, whether it’s Snowflake, AWS or elsewhere? And are we providing hooks into the right outlets to ensure that, as models are built, insight is accessible?” he says.
“We need to lead the way because we don’t want to be late into the game. If we’re late, there’ll be other providers of data that might be less authoritative, but because they’re easier to access, they’ll win out. We need to be at the forefront.”
Jonathan is a tech enthusiast and the mind behind Tech AI Verse. With a passion for artificial intelligence, consumer tech, and emerging innovations, he deliver clear, insightful content to keep readers informed. From cutting-edge gadgets to AI advancements and cryptocurrency trends, Jonathan breaks down complex topics to make technology accessible to all.
