Key Takeaways

  • Goldman Sachs Asset Management deploys a network of 110 operating partners to execute operational improvements and talent upgrades across middle-market companies.
  • Michael Bruun requires portfolio companies to narrow all AI rollouts into two direct financial buckets: scaling top-line revenue or expanding EBITDA margins.
  • Goldman established an internal AI University to educate portfolio CEOs, backed by the enterprise purchasing power of 12,000 internal engineers to secure priority access to model providers.
  • Cross-portfolio rollouts transfer tools across sectors, such as adapting an automated appointment reminder system from retail optics into financial services and cybersecurity companies.
  • The firm runs this transition across buyouts through the Goldman Sachs Portfolio AI Adoption Blueprint.

The Goldman Sachs Portfolio AI Adoption Blueprint

  • Step 1: Top-Down Mandate and CEO Alignment: Establish from the executive level that AI is a strategic must. Require CEOs to complete targeted AI education (such as Goldman's AI University) so they can effectively lead transformation discussions with boards and employees.
  • Step 2: Executive Committee Readiness Audit: Evaluate whether the executive committee (CTO, CRO/Head of Sales, COO) embraces AI tooling or clings to pre-2022 operating methods. Replace or realign leaders who resist the transition.
  • Step 3: Outcome-Focused Use Case Selection: Constrain initial deployments to a finite set of measurable outcomes targeting either scaling revenue (client service, RevOps, sales relevance, scheduling optimization) or margin improvement (repetitive task automation, software engineering acceleration).
  • Step 4: Controlled Sandbox Experimentation: Establish secure, compliant environments that allow ground-level employees to experiment with AI workflows ('letting a thousand flowers bloom') to discover emergent operational efficiencies.
  • Step 5: Centralized Vendor Leverage and Cross-Portfolio Scaling: Leverage enterprise platform scale to secure priority access and commercial terms with leading frontier model vendors, then rapidly cross-pollinate proven AI playbooks across disparate portfolio companies.

When This Works (and When It Doesn't)

This blueprint applies when integrating generative AI and automated systems into middle-market and upper-middle-market portfolio companies across diverse industries. It works best when a sponsor holds clear governance control and can mandate executive education, reallocate technical budgets, and re-engineer executive teams. The model thrives in service-heavy, distribution, and software businesses where automated workflows remove manual data entry or speed up customer response times.

It fails when applied to smaller lower-middle-market companies that lack the data infrastructure required to feed commercial models. If a company operates on fragmented, on-premise legacy databases, running sandbox experiments only creates isolated shadow IT projects. Without clean customer records and centralized sales data, top-down mandates stall at the management layer, wasting capital on tools that frontline staff abandon.

Why It Matters

With buyout multiples elevated and higher base rates eliminating cheap leverage, private equity sponsors can no longer manufacture returns through debt paydown and multiple expansion alone. Value creation has shifted to real operational compounding. Sponsors who treat AI as an open-ended R&D exercise will burn cash on shelfware, while firms that enforce executive accountability and direct use cases toward revenue or EBITDA will widen the performance gap across fund vintages.