- Tools are the cherry on top. DDQs still need human judgment, and no tool helps until you know where your data comes from and who owns it.
- A single source of truth has to be provable. If you can't trace a figure back to its source and approval, trust breaks and every request starts from scratch.
- Put approved content where people already work. CVC's Claude project of data room documents gives RMs instant answers and cuts the basic questions reaching IR.
- More AI output isn't more value. Use AI to sharpen briefing notes and responses, not to create volume nobody reads.
- Buy the core, decide the rest case by case. CRMs and investor portals are still best bought off the shelf, while DDQ tools sit on the fence.
- Spend the time AI saves on LPs and people. LPs will want deeper deal conversations, and junior staff still need to learn by reviewing AI output properly.
Fundraising is harder, LP expectations keep rising, and most Investor Relations (IR) teams are meeting both without adding headcount. That was the starting point for "Investor Relations at Scale: Solving for Efficiency," the panel we hosted with NextGen IR last week for around 40 IR professionals.
Tom Hughes, our Head of Commercial, moderated. He was joined by Dinoya Karunainathan, Vice President, Client and Product Solutions at CVC; Adam Aziz, Head of Project Management and Investor Servicing at Ambienta; and Christopher Vilhelmsen, Managing Consultant at Holland Mountain.
"AI is playing a huge part every day now," Tom said in his opening, "both on investor expectations, but also internal demands on teams as well."
DDQs Still Take Hours, Even with AI
Tom opened by asking what takes up time that LPs and partners don't see. Dinoya asked the room who works on DDQs, and almost everyone raised a hand.
Three years ago, CVC onboarded an RFP automation tool whose generative AI features were cutting edge at the time. The market has since moved on. "We're still spending so much time matching answers to questions," Dinoya said. Analysts are "sometimes spending five, six, seven hours fixing the formatting of the exports." CVC is now testing a Claude skill to fix the formatting.
AI hasn't made DDQs quick, either. On co-investment, for example, the right response depends on how much co-investment that LP has seen and how much it matters to them. "All of that is quite nuanced and requires some element of human judgment."
No tool is "a silver bullet," Chris said. "Applying a nice tool off the shelf or using Claude skills is the cherry on top, really. It's getting the foundation right: Where's that data coming from, who owns it, and how does it get there?"
A Single Source of Truth Has to Be Provable
Adam has seen data governance from both sides, after 13 years at UBS and a recent move to Ambienta. "In a larger firm, you tend to have more structure around these sorts of things, which can help," he said. The trade-off is speed: "Sometimes a slightly more homemade approach means you can react quicker to certain queries." The pressure point is an LP asking for data cut a new way.
Chris pointed out that IR data comes from portfolio monitoring, ESG, fund accounting, and value creation. IR teams should push for a single source of truth across all of it, and trust is the test. "A source of truth is only the truth if you can prove where it comes from," he said. "When it goes wrong, even just a little bit, you've lost that trust, and immediately the cycle is broken."
Put Approved Content Where Sales Teams Can Find It
CVC's IR team produces a lot of content, but it's only useful if relationship managers (RMs) know it exists. Instead of relying on SharePoint and email, the team put approved documents into a Claude project that RMs can query on the road.
That covers the basic questions. "Have you had questions like 'What's the term of this fund?'" Dinoya asked. "From someone who's worked at the firm for 10 years."
"We go to bed easy at night knowing they're using approved content, rather than a DDQ they found on the drive from 10 years ago," she said. Her team gets more time for bespoke analysis. Messaging also stays consistent, because everything in the project has already been through legal and compliance.
More AI Output Isn't Always More Value
Asked for quick wins, Chris suggested AI-generated LP briefing notes that pull together commitments, transfers, and recent questions. The hard part is taking ideas like this "from that pilot phase into something that looks like a fully-fledged, integrated system."
Adam added a warning. A colleague used AI to prepare a briefing note for a US meeting. "This thing had headshots, it was 10 pages long, it had the whole history, chapter and verse. I'm not sure anyone read it," he said. "So, the danger with AI is that, if you're not careful, you're just creating volume, and that's not what people need."
Across more than 200 GPs we work with, Tom said, firms are agreeing on approved templates and processes, so teams stop repeating the same work.
AI Investment Starts with the Deal Team
At both CVC and Ambienta, AI investment starts with the deal team. Ambienta has hired a data engineer to lead a firm-wide buy versus build review, including its portfolio companies. Adam is pushing for IR to be part of it, but "historically, the IR team has not been where people think, 'How do we structure this data to make it easy for IR?'"
CVC's house view is that "AI is a friend, not a foe," Dinoya said. Its first projects helped deal teams stress-test investment committee memos. Then the firm noticed the IR team kept running out of Claude credits, so it came to them next. CVC is now replacing Excel DDQ trackers with a hub where AI routes each incoming request to the right strategy lead, who assigns the work using a capacity dashboard.
Buy the Core Platforms, Decide the Rest Case by Case
Chris was clear on buy versus build. "Core platforms are still very much buy off the shelf," he said. Security and regulatory demands mean "getting a CRM or an investor portal off the shelf just makes sense."
Beyond the core, each tool is its own decision: build some, buy others. "DDQ is probably on the fence," Chris said.
LPs Are Using AI Too
Adam hasn't seen a big shift in what LPs expect yet. One LP even gave Ambienta access to a platform that used AI to complete the DDQ on the firm's behalf. "They have the same issues on their side," he said. The bigger opportunity is what IR teams do with the time they save: more proactive content, events, and investor education.
Dinoya expects AI to change what LPs want from meetings. If they can run desktop due diligence through ChatGPT or Claude, they'll want the nuance only deal teams can give: why a deal was done, how it's performing, and how it might exit.
"Everything we used to throw at LPs, they can now chuck into ChatGPT and Claude and churn out really good questions, and then we have to answer them," she said. Deal teams need training to answer them well.
The Next Generation of IR Still Needs to Be in the Detail
Tom's final question was the one on everyone's mind: What does this mean for people starting out in IR?
Adam called for balance. Junior team members still need the basics, but they also need to critically analyze AI output. "AI is how you make those individuals more powerful," he said. Dinoya pointed to her time with senior stakeholders at CVC, who share the firm's vision, potential M&A activity, and growth plans. "No AI can replace that," she said.
Chris questioned how much human review happens today. "If we're honest, how often are those human reviews actually doing anything, rather than ticking a box and moving it along?" His answer is to break AI workflows into segments that each need a real review. That way, junior staff still learn the data and the process.
Questions from the Floor
- What goes into the Claude project? Nothing new, Dinoya said. The team uses documents already in the data room that LPs can access, and replaces them every quarter. Claude projects limit how much content they hold, so the set is curated to get RMs "80% of the way there." Five or six months in, she gets far fewer basic questions.
- How do you govern AI answers? Adam is looking at a DDQ tool that labels each answer as referenced, inferred, or calculated. Pre-approved content can go straight through, and anything fabricated needs human review.
- How do you reuse client-specific language? CVC flags tailored answers in its RFP library, so a warning appears if someone reuses them. Every response also gets a four-eyes review. That's part of why CVC hasn't moved its DDQs entirely into Claude.
- What about agentic AI? Adam would like to apply it to quarterly reporting, a repeatable process that takes a lot of time.
Where We Go from Here
Across our clients, Tom said, LP-facing chatbots and agentic tools sometimes work well. Other times, LPs ignore them and send an email anyway. "We're still trialing different elements of the investor experience," he said.
The panel's message was consistent. Get the data, ownership, and approved content right first. Then use AI to free up time for the work LPs value most.
See how Atominvest's Investor Relations platform gives your team a single source of truth for fundraising, onboarding, and LP reporting, so you can meet rising LP expectations without adding headcount. Book a demo.
Frequently Asked Questions
Every answer still needs human judgment to stay current and relevant to the LP receiving it, and older RFP tools can leave analysts spending five to seven hours fixing export formatting.
It has to be provable, because trust breaks as soon as one figure can't be traced back to its source and approval.
CVC puts approved data room documents into a Claude project that relationship managers can query on the road, and replaces the files every quarter.
Buy core platforms such as a CRM or investor portal off the shelf, then decide case by case for everything else, with DDQ tools currently on the fence.
As LPs run desktop due diligence through AI tools, the panel expects sharper questions and meetings that focus on the deal's nuance only investment teams can give.
.png)
