BOTTOM LINE UPFRONT
We just got back from SuperReturn 2026. Five things stayed with us: the metrics are changing, the operating models are not, expertise is becoming the real moat, the mood is better than the math, and the exit is no longer the strategy. Everyone in private equity is doing AI. Almost nobody is doing it in a way that will show up in the multiple.
Berlin 2026 was the year the private markets AI conversation grew up. The question of whether AI matters has been settled. It does. What hasn’t been settled is who is building something durable with it. The distance between firms with a real answer and those still running pilots is starting to show.
That theme ran through nearly every conversation across the week. But it shared the room with something else worth noting: the mood in Berlin was noticeably more optimistic than it had been in years, even though the underlying numbers haven’t caught up. Distributions remain depressed, hold periods have stretched well beyond historical norms, and liquidity is still the central frustration for LPs. It’s an interesting paradox: the sentiment has improved but the math has not.
Five things worth bringing home:
1. The era of token maxing is over
Last year, the implicit KPI for AI in private markets was usage. At one session, 98% of PE managers reported mandated AI adoption. Pilots, Copilots, Internal dashboards: The conversation was still largely about whether firms were doing AI.
Berlin 2026 felt meaningfully different. The question has shifted from “are we using it?” to “what has it delivered?” What revenue lift was created, what cost was removed, and what EBITDA moved. Token consumption is still discussed, but increasingly through the lens of cost, governance, and return on investment. What was once celebrated as a sign of progress is now being asked to justify itself.
One example we referenced throughout the week: using AI to analyze large volumes of unstructured customer email data to surface churn risk and sentiment, feeding those insights directly into commercial workflows and ultimately driving an $18 million revenue improvement. That is what a real AI story looks like now, not because AI existed, but because measurable business value existed.
2. The electrification lesson nobody learned
The best analogy we heard all week came from the history of electrification. When electricity first arrived, factories rushed to adopt it. Steam engines were swapped for electric motors. And productivity barely moved. Companies treated electricity as a drop-in replacement rather than a reason to redesign. Same workflows, same layouts, same operating model…just better technology. The productivity gains came years later, when factories rebuilt themselves around what electricity made possible.
AI is following a remarkably similar path. Most organizations are still asking how to insert AI into existing processes. The more interesting question, and the one where real value lies, is what the process would look like if it were designed from scratch today, knowing AI exists. The firms creating the most value are redesigning how work flows through the organization, where decisions get made, where expertise sits, and how judgment gets applied. That is harder than buying software and it is also where the distance between firms will grow.
3. The exit is no longer the strategy
When exit multiples compress and hold periods extend, portfolio company performance becomes the primary lever. The industry is now averaging closer to seven years per hold versus the traditional three to five, and the conversation has clearly shifted from how to buy well to how to create measurable value during ownership. Pricing, commercial acceleration, procurement, data infrastructure, workflow redesign. These are no longer back-office considerations, they’re the whole game.
AI is accelerating the divide between firms that have built repeatable value creation capabilities and those still relying on financial engineering to do the heavy lifting. And it is changing how investors think about terminal value in the process. Questions that once lived in technology diligence (how vulnerable is this business model to disruption, how quickly can management adapt) are now shaping investment committee discussions and underwriting assumptions before capital is committed.
4. The mood is better than the math
One of the more striking contradictions in Berlin: sentiment felt noticeably more optimistic than 2023 or 2024, with more energy, more confidence, and less existential anxiety in the room. The underlying numbers tell a different story. Bain data cited throughout the week showed buyout distributions running at roughly 14% of NAV against a historical average closer to 25%. The industry is reportedly sitting on approximately 33,000 unsold portfolio companies, and more than half of LPs say the distribution slowdown is constraining their ability to make new commitments. This is now roughly the fifth consecutive year of below-normal distributions, and firms are feeling it. Confidence has returned faster than the cycle has resolved, and that tension ran through nearly every conversation.
5. AI is making expertise matter more, not less
Perhaps the most counterintuitive theme from the week: AI may increase the returns to domain expertise, not reduce them. There was broad agreement that foundation models will continue to improve rapidly and that access to powerful AI will eventually look more like a utility than a differentiator. What will not become a commodity is the proprietary data inside every organization, and the expertise needed to ask the right questions of it.
Several speakers made the point that AI is acting as a stress test for organizational data. Firms that invested in data quality and governance years ago are extracting value faster. Those that did not are finding that poor data remains a hard constraint regardless of how powerful the underlying model becomes. One GP demonstrated a system functioning as a shadow investment committee member, drawing on decades of internal papers and portfolio history to pressure-test theses and surface assumptions that had historically proved wrong. The value of that kind of capability is non-linear: one better decision, one avoided mistake, can have a disproportionate effect on overall returns.
Running underneath all of this was an underappreciated point about the private markets ecosystem itself. The asset class is now roughly $15 trillion and widely expected to double by 2030, yet much of the underlying infrastructure still feels built for a different era. As private markets continue moving toward wealth and retail channels, the pressure for cleaner data and more digital infrastructure will only grow. It feels a bit like ETFs around 2003: still early, somewhat clunky, but with enormous long-term potential if the infrastructure catches up.
The reckoning is already sorting firms into two groups. Those doing the hard organizational work of redesigning how decisions get made, building genuine data advantages, and treating operations as a source of competitive advantage. And those still running pilots. That window is closing, but for the firms and leaders in that room, the opportunity has never been bigger.