AI amplifies what is already there
What 15+ PE operating partners told each other across two days of the AI operating partner roundtable. The warning that kept coming back: fix the fundamentals before the model.
Where the room stood going in
Operating model
Who owns AI across the portfolio: one dedicated AI operating partner, or the work spread across the team.
- 44% dedicated AI operating partner
- 44% distributed model
Current model
How well their current approach to AI in the portfolio is working.
- 56% somewhat effective
Top priority
Where they are putting AI effort next.
- 56% doubling down on GTM and sales AI
Pre-event survey of roundtable attendees. The unfilled part of each bar is the rest of the room.
Executive summary
Where both days landed, in the order the room got there.
Fix the fundamentals first
AI amplifies the sales process, data and management habits a company already has. Where managers ignored the signals, win rates dropped.
Data and governance come before pilots
Clean data, cyber policy and use cases locked into the operating plan come first. Small, visible wins build the trust to scale.
GTM is where the effort is going
56% of the room is doubling down on GTM and sales AI, and 56% will judge year one on EBITDA impact or cost reduction.
Measure outcomes, and budget for talent
Track revenue per head, NRR, win rate and pipeline velocity. Forward-deployed AI engineers now cost $300K to $400K.
Answer the two questions the room answered
Operating partners answered these questions before the roundtable. Pick yours to see where you land, and what the room said about it.
Room answers from the pre-event survey. "Something else" covers answers the survey summary did not break out. Your answers are saved anonymously, with no name or email, so later recaps can compare more of the field.
Day 1: build the foundation before the pilots
AI champions network
One PE operating partner has built a champions network with one AI lead per portfolio company. The group tracks ROI on any initiative above $5K and maintains a shared failure catalogue. Their view is that failure should be encouraged.
Governance before pilots
For another operating partner, the first priority has been cyber policy and committee governance. AI use cases come later. He sees governance as the unglamorous foundation that makes pilots actually stick.
Context engine demo
Dan Bernoske walked through a live coaching tool that pulls from call transcripts, CRM data and a capability rubric. It surfaces rep-level gaps and manager talking points. His estimate: a team can build something like this in 2 to 4 weeks.
KPIs: revenue per head plus NRR
Dan Bernoske introduced two emerging metrics that tell him AI is working. Revenue per head is the productivity signal, and NRR confirms service quality hasn’t slipped in the process.
The talent premium is real
An executive recruiter in the room noted that forward-deployed AI engineers are now commanding $300K to $400K, up roughly 50% in three years. Accelerated vesting with no cliff is increasingly what top AI hires expect.
Shadow IT and vendor lock-in
A serial CEO and operating partner argued for an LLM abstraction layer, so portfolios are not beholden to any single vendor. He flagged token-based pricing as a real and underestimated cost exposure.
Day 2: AI does not fix broken fundamentals
Signals nobody acted on
One operating partner deployed Gong with a full sales methodology, and win rates dropped. Managers were not acting on the signals. The takeaway: AI amplifies what is already there but does not fix what is not.
The three-legged stool
Dan Bernoske framed the architecture for GTM AI: a context engine, a forward-deployed AI engineer and a domain expert. All three are required. Remove any one leg and the whole system fails.
Everyone must be AI-enabled
One operating partner runs a team of three AI GTM operating partners who embed directly in portfolio companies. In his view, one specialist is not enough. Every GTM function needs to be on a path to AI-native.
Data quality is the real bottleneck
A former PE operating executive argued for setting use cases, outcomes and ROI targets first. Lock them into the annual operating plan before touching any model. He and others pointed to Snowflake and Databricks as the standard starting point for data infrastructure.
Outcome metrics only
An operating partner at a large PE firm pushed the group to track only outcome metrics: win rate, pipeline velocity, time to value. In that view, activity metrics are distractions, and everything needs to tie back to the P&L.
The lower market needs its own playbook
A lower middle market operating partner noted that some of his portfolio companies have no CRM yet. In his experience, cost-side AI wins came first because they are more tangible. Founder trust has to be built before any larger transformation takes hold.
You're just strapping a rocket onto something that's broken. It's not going to get you there.
Every go-to-market function needs to be AI-enabled and on a path to becoming an AI-native version of that job description.
AI is great, but if you don't execute on the fundamentals, it doesn't help.
What both days kept coming back to
Apply to the waiting list for the next roundtable
Small rooms, operating partners only, Chatham House rules. We will email you when the next date is set.
Thank you. You are on the waiting list, and we will email you when the next date is set.
Shared under Chatham House rules: ideas are credited to roles, and attendee names are withheld.