Two: Only keeping AI at the fund level
Roemer says the AI story goes beyond funds to portfolio companies. It becomes marketing if a GP mandates AI adoption across their portfolio but less than half of those companies are implementing it for specific results.
The domino effect leads to risking credibility of the firm, a vulnerable position in the LP-GP relationship and eventually eroding trust between parties.
“LPs are getting better at asking follow-up questions and the answers are starting to matter in fundraising conversations,” Roemer says. “LP relationships are long and the community is small. A GP that overclaims on AI in 2024 and can’t point to results in 2026 has a problem that goes beyond the AI story.”
Open the gate
Every AI tool is built off a certain database, but how GPs implement and present it can be the key to preventing AI washing. In conversation with The Drawdown, OpenGate’s Mohin shares how he approaches governance of the firm’s AI deployments.
When Mohin presents to LPs, the test he applies is whether it is quantifiable and defensible. This can include showing a specific workflow, measurable output or KPI. Mohin says, “if you cannot answer these questions, you are not ready to present it to an LP.”
He continues: “The near-term use cases may not be glamorous, they are repetitive, auditable and boring but important. Examples include document review and synthesis, quality control checks, reconciliation and research. This is high-volume, low-creativity work that consumes human computing power. This is where the ROI is measurable: error rates and hours recovered.”
The specificity and metrics Mohin describes add credibility. Roemer has a similar perspective because LPs can evaluate which workflows changed, which metrics moved over what time period and in which portfolio companies. “They cannot evaluate a slide that says AI is a priority.”
However, the specificity depends on what stage the GP is in addressing their operational issues. Roemer further explains that if a GP’s first approach is to solve the operational problems before diving into the technology, the framing shifts from “we are deploying AI across the portfolio” to “we identified that portfolio companies were spending significant finance team capacity on manual close processes, and here is what we did about it and what it produced.”
“The GPs building real trust with LPs right now are the ones acknowledging that this is hard, that adoption is the real challenge and that they are investing in the organisational capability to make it stick, not just the technology,” adds Roemer. “That framing tends to be more credible than perfection because it matches what LPs are hearing everywhere else.”
A final word of caution for GPs: it can be difficult to assess the long-term costs of using AI tools, given how heavily subsidised AI tokens are right now. Rushing to commit to AI implementation and overstating use cases is not only a potential reputational risk with LPs, it could also become an unexpected financial burden.
As the potential of AI implementation bubbles up, tread carefully to avoid slipping on its soap suds.