How to Measure Your AI Resilience: Inside Centellic's Content Audit
For CEOs, MDs, content and editorial leaders in specialized subscription, membership and events businesses
If you lead a specialized information, subscription or membership business, you already know the feeling Clare Bolton described to Renewd members: no matter how much time you spend thinking about AI or testing the tools, there is always something new. A new competitive threat, a new tool, another thing your clients are using. It is easy to end up running around with your hair on fire, fixing one thing over here and another over there, with no real sense of whether any of it is having an impact.
In this Renewd case study session, Clare Bolton, Chief Content Innovation Officer at Centellic (formerly Law Business Research, and home to Law.com and Lexology), shared how her team stopped reacting and built a structured way to measure exactly where AI posed a real risk, where it created an opportunity, and where they could afford to do nothing.
🎥 Watch Now - AI Resilience Audit: A Renewd Case Study with Clare Bolton, Centellic
Continue the conversation at Renewd
If these challenges resonate, sign up for the Renewd Community Digest to receive regular insights and resources for media and events leaders.
What is an AI resilience audit?
An AI resilience audit is a structured review of your entire content or product portfolio to work out how exposed each part is to AI disruption. Rather than treating AI as one overwhelming threat, you grade each content set on its own merits, plot it by value and risk, and then decide what to grow, what to protect, what to transition, and what to stop.
The goal is to replace a vague, anxious sense of threat with a clear, shared picture of where the real risks and opportunities sit, so decisions get made on evidence rather than instinct.
Why Centellic ran one
Centellic is a global intelligence and information services business serving legal, risk, compliance and IP communities, formed in 2025 from the merger of Law Business Research and ALM, and rebranded to Centellic in May 2026. With around 950 employees and 10 platforms, its core mission is built on data, intelligence and analysis, so AI disruption was not abstract. It was aimed straight at the value proposition.
One of its legal research products had launched about three years before ChatGPT, occupying a clear middle ground: better than a general search engine, but cheaper than the premium incumbents. ChatGPT now sits squarely in that same middle ground, competing directly for the same use case. The team could see the threat coming.
The bigger problem, though, was not any single threat like this one. It was that there was no structured way to assess risk across the whole portfolio or to shape a response, which left the team at risk of reacting to everything and changing nothing that mattered. As Clare put it, picking an approach and committing to it was valuable in its own right, because it gave the team a shared way to measure risk and act on it.
The six AI resilience factors Centellic used
Centellic took every content set across every platform, put them into a grid, and graded each one against six factors, split into internal factors (what the business puts in) and external factors (what clients get out).
Internal factors
-
Proprietary data - is the data genuinely yours, or is it replicable and available elsewhere? Exclusive data that others do not have is one of the strongest defences.
-
Network effects - some content becomes more valuable as the audience or connected data grows. Centellic's Lexology model is a good example: law firms pay to have their content republished to an engaged audience, and in return get data on who read it for their own business development. Scale is exactly what makes it valuable.
-
Content differentiation - exclusive insight or content that needs specialist expertise, deep contextual knowledge and real human relationships. The kind of work that comes from an analyst who has spent years in a niche and knows everyone in it.
External factors
-
Value chain position and risk levels - where the content sits in a client's decision. If a decision is mission critical, directly affecting revenue, regulatory compliance, or "keeping a CEO out of jail", the user is far less likely to settle for an unverified AI answer.
-
Workflow integration depth - data that is deeply embedded in a client's workflow, for example flowing by API into their compliance processes, is much harder to rip out and replace. Lower substitution risk.
-
Competitive landscape - how replicable the content is by AI or low-cost alternatives. The real risk is where AI has become good enough, and B2B buyers will accept lower quality for a substantial cost saving.
Each content set was also weighed against engagement data and a considered view of its value as part of its wider subscription, since individual content sets rarely carry their own revenue line.
The test that changed the conversation
One step made the whole audit credible internally. Clare asked her journalists to take a piece they were proud of from the last three to six months and get as close as they could to recreating it in ChatGPT by refining the prompts.
This mattered because it moved the assessment away from gut feel and preconceived bias. People who had assumed AI could not come close to their work were made to sit down and test it, and for some the result was an uncomfortable realisation of how near it got. The outcome varied by type of content: more niche areas held up better, while more generalist output was easier to replicate. That became a category the team named "BAU news" - work that, at its worst, is commoditised and replicable.
As Clare put it, nobody subscribes to Lexology or Law.com to get content they could get anywhere. Testing it honestly, rather than debating it, was what generated genuine buy-in and decisive action.
The output: a value versus AI risk matrix
Every content set was plotted on a classic two-by-two, value on one axis and AI risk on the other. That produced four quadrants, each with a clear instruction:
-
Low risk, low value - carry on, do not spend energy on it.
-
Low risk, high value - grow it. This is where to expand and invest.
-
High value, high risk - take action. Either milk it, accepting it may eventually drop off a cliff, or transition it by actively reducing its AI risk and moving it left.
-
High risk, low value - stop or automate it, and significantly reduce the time spent here.
From this, Centellic set five concrete actions per platform to complete over the following six to nine months, ready for the next iteration of the audit:
-
Deepen proprietary data in tools built on public data, through richer tagging and analysis.
-
Deepen co-published content to make it more technical and complete, closing jurisdictional gaps.
-
Enrich and expand network links between data sets to compound their value.
-
Amalgamate high-value proprietary assets with linked use cases behind an AI interface.
-
Stop or automate low-value work. Centellic cut around 40% of its business-as-usual news and automated much of the rest, running the occasional necessary story through a prompt-based first draft with a quick human check.
What Centellic learned
Not all content is equally threatened by AI, so precision targeting matters more than sweeping change. Proprietary data remains the strongest source of defensibility, and generic content is already commoditised, which means you can act on it decisively.
Beyond the content decisions, the audit delivered several things Clare had not fully expected:
-
A concrete action plan in the midst of chaos, and a shared, tested sense of direction the team could get behind.
-
The value of focus - knowing what not to invest in proved as important as knowing what to scale.
-
Space for the right conversations, even the difficult ones about the competitiveness of AI against the team's own content.
-
A credible way to talk to investors - a private equity board responds well to a resilience score expressed in a matrix.
-
A shift in talent strategy, moving investment out of traditional editorial functions and into data acquisition.
There was a people dimension too. Like most organisations rolling out AI, Centellic saw the full range of reactions you would expect, from early builders who took the tools and ran with them, to a large and willing middle, to understandable caution about what it all means for how people work. Clare was candid that this is a genuine change-management journey, not a switch you flip. What the audit did was give everyone a shared, evidence-based way to have the conversation, which moved many people forward even when the discussion was a difficult one. Her view is that the audit is not an endpoint. It is another step in a longer shift towards working with AI rather than against it, something the whole team is learning together.
How the audit changed the way Centellic sells
The audit was not designed to change go-to-market, but the commercial teams found it genuinely useful. It gave them language and specific mechanisms to counter AI-related objections, for example when a client argued they could do something in ChatGPT and should therefore pay less. Some of those objections were real and some were negotiating tactics, and the audit helped teams respond intelligently to both.
It also supported a wider shift from a product-led to a client-led approach, giving account teams a way to show how Centellic is investing across the whole portfolio, not just within single products.
The takeaway for Renewd members
You do not need to fix everything at once, and you cannot. The value is in a structured method that tells you where AI is a real threat, where it is an opportunity, and where you can afford to do nothing. Pick an approach, test your own content honestly against the tools, and let the evidence drive the decisions.
Frequently asked questions
What is an AI resilience audit?
A structured review of an organisation's entire content or product portfolio to assess how exposed each part is to AI disruption. Centellic graded every content set against six factors and plotted each on a value versus AI risk matrix to decide what to grow, transition or stop.
What are the six AI resilience factors?
Three internal: proprietary data, network effects, and content differentiation. Three external: value chain position and risk levels, workflow integration depth, and competitive landscape.
How do you test whether content is resilient to AI?
Centellic asked its journalists to try recreating their own recent work in ChatGPT by refining prompts, to see how close AI could get. This replaced gut feel with real evidence and drove team buy-in.
What did Centellic do as a result of the audit?
It set five actions per platform, including deepening proprietary data, enriching network links, and automating low-value work. It cut around 40% of business-as-usual news, and shifted talent investment from traditional editorial roles into data acquisition.
This post is based on the Renewd case study session "How to measure your AI resilience," featuring Clare Bolton, Chief Content Innovation Officer at Centellic.
🎥 Watch Now - AI Resilience Audit: A Renewd Case Study with Clare Bolton, Centellic
Continue the conversation at Renewd
If these challenges resonate, sign up for the Renewd Community Digest to receive regular insights and resources for media and events leaders, or check out our free and Member-Exclusive events here.
Or join Renewd as a Premium Member and have these conversations in person, with peers who know exactly what it's like to sit in your seat.
Premium Membership North America Premium Membership UK/Europe
