AI in Investment Management: From Pilots to Practical, Firm-Wide Use
Based on a November 2025 survey run with Corporate Counsel, three leaders-asset management partner Melissa C. Bender, private equity partner Howard S. Glazer, and Director of Technology Innovation Sergey Polak-shared how firms are moving from experiments to enterprise adoption. The headline: adoption is rising, impact is uneven, and the winners are treating AI as both a technology and a management discipline.
Below are the most useful takeaways for management teams responsible for P&L, risk, and execution.
Where managers are actually using AI
- Portfolio management and optimization
- Investment due diligence and underwriting
- Ongoing investment monitoring and analysis
- Legal, compliance, and risk functions
- Operations and contract workflows (NDAs, transfers, term review)
- Benchmarking disclosures and market terms across peers
Early success on the investment side is now pulling AI across legal, finance, IR, and operations.
Why adoption is accelerating
- Operational efficiency leads by a wide margin
- Strengthening compliance and risk management
- Competitive parity-keeping pace with peers
Despite headlines, broad headcount reduction is not the primary driver in this sector-yet. Leaders are targeting time-consuming, repeatable tasks to free teams for higher-value work.
Governance that actually works
Most respondents have policies and processes in place (or in flight). Effective setups share a pattern: a cross-functional group (business, legal, compliance, IT) vets tools, sets use standards, and updates policies as models and regulations shift.
Focus areas include data privacy, confidentiality, acceptable use, and vendor controls. An industry reference some teams use: the NIST AI Risk Management Framework.
Tool selection: security first, use case second
Security, confidentiality, and tool sprawl are the top concerns-not budget. The market is noisy, and "buy the shiny thing" is an expensive habit.
- Vet vendors for data handling, access controls, retention, and contractual protections
- Prioritize 2-3 high-value use cases over broad, unfocused rollouts
- Plan integration with legacy systems early; don't bolt it on at the end
- Invest in change management, not just licenses-AI demands new habits
Barriers leaders should plan for
- Security and confidentiality risks (especially with cloud-based tools)
- Tool sprawl and overlapping features
- Integration with existing systems and data sources
- Underestimated change management and training needs
- Non-deterministic outputs that require review and spot checks
- Budgetary constraints exist, but they're not the primary blocker
How to measure ROI without fooling yourself
Many firms haven't measured results yet, or it's too early. That's normal. Define what success means before the pilot begins.
- Accuracy: What level is "good enough" for the task? Define review protocols.
- Efficiency: Time and cost vs. the old baseline. Track pre/post numbers.
- Scope: New analyses you can run that were previously cost-prohibitive.
- Risk reduction: Fewer misses in compliance or contract terms.
- Vendor adaptability: Speed of iteration to your use cases.
Used poorly, AI can slow teams down. Research from Harvard has shown mixed outcomes when tasks aren't well matched to the tools-set guardrails and train for the work you actually do. A helpful overview: How to Use AI at Work (Harvard Business School).
Case study: a 1-hour benchmark for digital asset custody agreements
Melissa described building a Hebbia matrix to analyze and compare digital asset custody terms across dozens of providers. When a client needed a quick turnaround, the team ingested the agreement, extracted the key clauses, and benchmarked them against prior deals within an hour.
A senior lawyer then reviewed and negotiated based on those findings. The bonus: junior team members learned faster, because building the matrix forced them to understand indemnities, liability caps, and the terms that actually matter.
What's next: deals, agentic tools, and infrastructure
Adoption is rising across the industry, but bottom-line impact lags as teams learn, refine prompts, and adjust workflows. Interest in agentic AI is real, but most enterprise use remains early and focused on narrow, supervised tasks.
Capital is flowing into data infrastructure in a big way. As noted in the session, economist Jason Furman attributed a large share of U.S. GDP growth in early 2025 to AI data center investment. Partnerships between big tech and private capital are accelerating, and add-on acquisitions to buy AI capabilities are common.
A pragmatic rollout plan for management teams
- Pick 2-3 use cases with clear business value and measurable outcomes
- Stand up a cross-functional AI council (business, legal, compliance, IT)
- Run vendor diligence: security, data handling, IP, model updates, support
- Pilot with an opt-in team; document the workflow and review steps
- Set success metrics before kickoff; report results monthly
- Train on prompts and task design; share wins and prompts internally
- Integrate with systems and data; reduce copy/paste work
- Iterate quarterly-tools change fast, and so should your playbook
Resources
Level up your team
If your organization needs practical, role-based upskilling, see:
Your membership also unlocks:
