Key Takeaways

  • Jon Webster leads artificial intelligence deployment across 2,000 investment professionals managing $580 billion at CPP Investments on behalf of 22 million Canadians.
  • Software called Memo Coach uses multi-agent review teams to red-team draft memos across risk, logic, and valuation scenarios before they ever reach the Investment Committee.
  • Applying Clayton Christensen's Law of Conservation of Attractive Profits, Webster argues that commoditizing quantitative analysis shifts alpha toward relationship trust and emotional intelligence.
  • Technology alone confers zero lasting edge because identical models are accessible to every rival pension fund and private equity sponsor.
  • The institutional workflow follows CPPIB's Institutional AI Triad: Read, Remember, Challenge.

The CPPIB's Institutional AI Triad: Read, Remember, Challenge

  • Pillar 1: Read Everything: Deploy natural language processing across all corporate filings, earnings call transcripts, and underlying secondary portfolio assets to ingest data at a volume no human team could manually cover.
  • Pillar 2: Remember Everything: Structure all historical deal decisions, governance rights, and CRM touchpoints into unified knowledge graphs and queryable databases so every conversation is backed by full institutional memory.
  • Pillar 3: Challenge Everything: Utilize multi-agent review platforms (such as Memo Coach) to stress-test investment recommendations across logic, scenario modeling, and risk factors before presentation to the Investment Committee.

When This Works (and When It Doesn't)

Used at institutional scale to augment underwriting capabilities across multi-asset investment platforms while protecting against analytical blind spots.

This structure works when an allocator runs thousands of disparate assets and faces institutional amnesia. When deal teams turn over every three years, junior staff repeat past mistakes. Pillar 2 stops firms from re-litigating terms or re-underwriting bad assumptions that previous partners already solved. Pillar 3 succeeds because deal sponsors naturally suffer from deal momentum. Software has no career risk and no incentive to push a bad acquisition forward just to deploy capital.

Where this triad breaks down is in proprietary, off-market situations where historical data does not exist. If an investment thesis relies on sudden regulatory shifts or complex founder dynamics, automated agents hallucinate precision. Relying on automated red teams can also create a false sense of security. Investment committees may assume every angle was covered simply because an algorithm cleared the memo, dulling the human skepticism required when underwriting unconventional risks.

Why It Matters

Webster points out the core economic reality facing institutional capital: “Technology never confers a lasting competitive advantage because whatever is available to you is available to everybody else.” When every private equity shop and sovereign wealth fund runs the same automated data pipelines, baseline underwriting intelligence becomes table stakes.

Webster applies Clayton Christensen's economic principle: “When one part of the value chain is attacked, then value accrues to the adjacent parts of the value chain. If IQ is under attack, value is going to accrue to EQ.” When financial modeling and document analysis take seconds instead of weeks, standard financial engineering loses its pricing power. The spread shifts to sourcing proprietary deal flow, reading founder body language, managing complex boardroom politics, and building trust during corporate carve-outs. Allocators who win will not be the ones with the fastest summary tools, but the ones who reallocate saved analyst hours into relationship capital.