We chaired our latest Research Leadership Network roundtable this week, where we were joined by leaders of research teams from global consultancies, law firms, professional services groups, and global software giants.
It was very clear to all that research teams are being asked to keep pace with AI change that the majority in the room feel is not sufficiently coordinated, verified, or owned. Three things stood out.
The first is coordination, or the lack of it. There’s a real whack-a-mole problem right now: unchecked AI projects popping up across firms, with different innovation groups launching new products, platforms and tools that directly affect how research is carried out, often with little coordination between them. Sometimes it’s not even clear which group a research team should be aligning with, and often it’s obvious early on that what’s being built just isn’t fit for purpose.
The second is provenance and accountability. There’s a tangible issue with people across the business using inappropriate tools: LLMs pulling in third party data with no clear provenance, when the research team already has subscriptions to premium sources that would answer the question properly. Research teams know which sources are reliable; other areas of the business often don’t. There’s an accountability gap in some places, and a significant void in training on how AI actually works and how to embed critical thinking into the process rather than just trusting the output.
The third is that research is still fighting for its seat at the table. A few people in the room talked about AI being pushed on their teams in ways that feel performative, more about being seen to be doing something than actually doing something useful. Similarly, some firms are now using tools to make AI outputs seem more human, which, sat with for a second, is a very strange place for this to have ended up. Meanwhile many groups within firms want to own AI, and plenty want to run the transformation, but very few actually have the domain knowledge to do it properly in a research context.
Fundamentally, a lot of these governance issues are ones organisations will inevitably run into, and when they do, it will make them realise the real value of the institutional knowledge sitting in their research teams. Quoting from sources without proper checks on provenance or quality, building solutions that don’t actually address real research workflows, and the wider sprawl of uncoordinated initiatives will, at some point, be measured against token cost and human capital for return on investment as well as create issues in quality of work. The group’s prediction was that we’ll reach a stabilisation, perhaps not now, but within two to three years.
Picking up on some of the key topics and sentiment from this week’s discussion, we’re going to undertake further research across our wider Research Leadership Network. We want to understand how offshore research functions are actually structured, the seniority of insight teams, the skill sets firms are hiring for, and what’s being done to embed governance and quality assurance into research outputs, such as checking against verified sources. We’ll also be benchmarking infrastructure, to understand how firms make decisions on AI. Do you have an AI steering committee or champions? Who owns it? Is ownership split across different parts of the business? Where does research actually sit in that picture?
We’ll be asking about AI tokenomics and budgeting too, establishing if firms and stakeholders are really measuring value and how they’re assessing return on investment. We’ll also be looking at the direction of travel on salaries.
There’s so much noise out there right now. Stakeholders often come to our members with stories about what strategies other firms are adopting, and much of this is half-truths, filtered, or in some cases, pure rumour. Things are moving fast enough that a proper, honest state of play would be extremely beneficial, particularly in this highly transitional period for a lot of firms.
If you lead a research function and would like to be part of this research, please do get in touch.