There is something grounding — even comforting — about leafing through a proper reference book. Even though our world has shifted toward PDFs, data terminals, and model outputs, that slower, deliberate process is a reminder that fixed income investing has always been a details-driven discipline.

That contrast struck me again when I opened the latest Financial Analysts Journal (sent by e-mail from the CFA Institute) and saw an academic paper by Fabozzi and co-authors: “The Disappearing Edge: AI, Machine Learning and the Future of the Discretionary Portfolio Manager.”

It examines the claim that the traditional PM “edge” is eroding as machine learning becomes more powerful and data more abundant. Much of the paper argues that the industry is being reshaped around automation: faster data processing, model-driven workflows, and human oversight rather than human decision-making.

There is truth in that — but I would frame it differently. The edge isn’t disappearing. It’s shifting.

Where AI helps – and where it doesn’t (yet)

There’s no question that AI and large language models are transforming the efficiency of investment management. Tasks that used to take analysts hours — data extraction, document scanning, pattern recognition, even drafting commentary — can now be generated in seconds.

But the current generation of AI is still prone to errors that would be unacceptable in a portfolio construction context. Anyone who relies on it heavily will recognise the issue:

• simple market levels that are wrong,

• macro data mis-stated,

• fabricated citations,

• and a lack of precision around time series.

For bottom-up credit work, the gap is even wider. AI can help organise earnings transcripts, scrape covenants, flag some structured anomalies and inconsistencies, and organise information — but judgment still matters. Assessing management tone, credibility, capital allocation behaviour, competitive response, or how a company will behave under stress remains a deeply human exercise.

As Fabozzi notes, the PM’s role is moving toward curation and interpretation rather than raw information gathering. I agree — but it’s important to recognise that interpretation is where skill actually resides.

How the PM role is evolving, not disappearing

If anything, AI enhances the parts of the job that should be enhanced:

1. Broader coverage, with fewer bottlenecks
   AI will eventually allow PMs to cover far more securities without relying on large analyst teams. It won’t remove the need for judgment — it simply removes friction.

2. Better scenario analysis and risk budgeting
   Complex multi-asset scenarios that used to take a morning can now be generated in minutes, allowing more time to think about what actually matters rather than assembling the inputs.

3. More consistent process discipline
   Systematised elements (momentum screens, flow models, macro positioning frameworks) create a stable backbone. The PM then focuses on the layers where experience really adds value: regime shifts, reversals, dislocations, extreme positioning.

4. Efficiency in communication — but the client relationship remains human
   AI can streamline the administrative side of communication — gathering background data, summarising long documents, organising charts — but that is very different from replacing the communication itself.

   In practice, the conversations that matter in asset management are still human ones. Explaining the investment process to new investors, discussing positioning on a monthly call, presenting at a roadshow, recording a podcast, or writing detailed insight pieces: these are fundamental parts of the portfolio manager’s role, and clients expect to hear directly from the person making the decisions.

   Insightful reporting cannot be automated. AI may help assemble information more efficiently, but the judgement, narrative and accountability behind those reports remain squarely with the portfolio manager. That is part of the value clients rightly expect — and part of the discipline of the job.

Where human discretion remains essential

Even as AI accelerates everything around the PM, certain parts of the craft remain stubbornly human:

• recognising regime change when traditional relationships break down;

• understanding market psychology, and when positioning has become one-sided;

• knowing when a model is “right” but the trade is wrong because liquidity, narrative, politics or policy are shifting beneath the surface;

• interpreting bottom-up company nuance (especially in credit);

• managing real-world risk when volatility spikes and liquidity evaporates — something only experience teaches.

These are not functions that can be automated. They are what separates a good PM from a merely systematic one.

A more balanced future

The foundations of fixed income, built through decades of Fabozzi’s work, haven’t changed. Fixed income has always been about understanding instruments, understanding valuation, and understanding what you are being paid for the risks you choose to take. Even in a world reshaped by AI, that still holds true.

AI can speed up workflows and expand coverage, but it still misses many of the anomalies that experience and judgement can identify.

It cannot yet tell you accurately:

• when a bond is genuinely mispriced,

• when a spread compensates you for the right risks,

• or when a credit story is shifting in ways the data hasn’t yet captured.

These remain matters of judgement, experience and interpretation — the core of active value investing.

If anything, AI makes those skills more important. When everyone has access to the same data, the same models and the same processing power, the differentiator becomes how you read the nuances the models can’t see.

That has always been true — and it remains true in the age of AI.

Craig Veysey, CFA
Chief Investment Officer