The data roles you need underpinning AI

'Move with the times: Adapt or die’

Daniel Neaves, senior recruitment consultant at Harvey Nash, explores how the long-standing challenge of getting value from data is impacting the rollout of modern AI systems.

As businesses across sectors race to implement AI solutions, it’s widely recognised that good quality data and sound data processes are fundamental to success: data is the foundation on which AI is built.

However, getting data right has long been a challenging area for organisations. A recurring finding in the Nash Squared/Harvey Nash Digital Leadership Report (DLR) over the years is that digital leaders struggle to get real value or business-useful insights from their data.

In the 2025 DLR, the challenge is being exacerbated by skills shortages, with nearly four in ten respondents reporting a shortage of data/analytics talent, the second worst area behind AI itself.

In this article, I will therefore offer some practical insights on the data-related roles and structures that companies need to enable successful AI.

Getting the foundations in place

The data/AI challenge starts with getting the basics of data right. After all, AI is itself essentially a form of data modelling, so the best practices for ‘conventional’ data apply here too.

A great overview of good data practice is set out in this Computing piece by our Chief Operating Officer Jason Pyle. He identifies a handful of core principles: ensuring that the data strategy and business strategy are aligned; creating the right technical environment such as a ‘lakehouse’ approach; establishing a company-wide data culture in which staff are educated and upskilled on how to treat, curate, maintain and store data; and having an effective data talent recruitment/retention strategy.

Key data roles for AI

Traditional data roles remain key – such as data architects to build the data infrastructure and frameworks, and data engineers to cleanse, optimise and maintain data pipelines.

These are foundational – but the really critical roles from an AI perspective are data scientists and ML ops engineers.

Data scientists assess the data the business already has and create ways to automate it in key areas. Experience of agentic AI is becoming increasingly important here. Data scientists utilise AI and automation to streamline and speed up key data flows, which increases efficiency, capacity and productivity in the business. They build AI applications that will generate business or process useful insights from the data – helping to solve that persistent challenge reported by digital leaders in our DLR.

ML ops engineers work hand-in-hand with data scientists to build large language models (LLMs) that ingest and process all of the relevant data – whether that’s text, images, audio, structured or unstructured data sets – and make it accessible to the business. Increasingly, they are using retrieval-augmented generation (RAG) techniques as an accuracy enhancer to prevent hallucinations.

AI engineers are also starting to become more common – similar to data scientists but entirely focused on AI, including AI strategy and AI networks. These are very specialist roles and we often find that companies turn to consultancies to access them rather than employing them in-house.

Then there is the question of executive ownership and accountability for data and AI. There are many different models in the market – it really does depend on the specifics of the organisation and what is already in place.

In some businesses, there is a Chief Data Officer (CDO) who will naturally have primary responsibility. Some organisations may have a slightly less senior Head of Data instead. In other businesses, there may not be a CDO or Head of Data, with the CIO or CTO taking in responsibility for data/AI as part of their brief. We are also seeing more Heads of AI being appointed.

The key question is not what job titles and specifications are in play, but that there should be clarity over roles, responsibilities and accountability. This is what will drive progress and success.

Get on the journey

Data has long been a challenging area for most businesses, but the stakes are even higher now given AI's strategic priority. It really is a case of moving with the times: adapt or die!

This means rigorously reviewing your data roles and processes against your objectives for AI and identifying gaps and weaknesses where capabilities need to be strengthened or existing roles changed.

To a large extent, the fundamentals remain the same, with data architects, data engineers, data scientists and ML engineers needing to closely collaborate and coordinate, under clear leadership at the executive level. Once the structures are in place for data to flow through and feed your LLM, it is also critical to keep humans in the loop and continually assess the outputs, considering what needs to be refined. Just like data itself, AI is a journey that requires constant iterating and optimisation. Perhaps the biggest single piece of advice is therefore quite simply: Get started on the journey now!