What problem does it solve?
Clients expect a periodic report that explains what happened to their money and why, not just a table of numbers. Writing that narrative is slow: portfolio managers and specialist writers draft commentary for each strategy, and advisors adapt it for individual clients, often in several languages. At Quilter, specialist writers interview a portfolio manager and then need a few days to turn that into a commentary; Neurons Lab describes a monthly investor report that took an investment firm 20 days to complete. The result can be generic text that says little about the client's own portfolio, or reports that reach clients well after the period they describe.
The risk is also real. A commentary is a client communication under conduct rules; a wrong number, an unbalanced claim about performance or a forward looking statement without the right disclaimer is a compliance issue.
How does it work?
- Take numbers from the record. Performance, attribution, holdings and transactions come from the portfolio accounting and performance systems, not from the model.
- Add the context. The house view, market commentary and the portfolio manager's notes are retrieved for the period.
- Draft within a template. The model writes the narrative sections in an approved structure, inserting figures from the data and explaining drivers in plain language, per client or per strategy.
- Localize. The approved narrative is adapted to the client's language and segment.
- Check and approve. Automated checks compare every number in the text with the source data and flag banned phrases; a reviewer approves before the report is assembled and delivered, and a log of sources and edits is kept.
- Audience
- Back office
- Autonomy
- Copilot
- Adoption
- Early adopters
- Channels
- Internal tools, Email
What is it worth?
Benchmarks are computed from the public deployments below: one data point per organization per KPI, with who made each claim.
No public deployment has disclosed a measurable outcome yet.
Value drivers: Employee productivity, Speed and cycle time, Customer experience, Compliance quality.
Indicative value
A wealth manager sending quarterly reports to 20,000 client portfolios
USD 800,000 to USD 4.8 million
Value of staff time released from commentary drafting per year
How this is calculated
Formula: portfolios * reportsPerYear * minutesSaved / 60 * hourlyCost. The low scenario uses every low input, the high scenario every high input.
| Input | Low | High | Basis |
|---|---|---|---|
| Client portfolios with a personalized commentary portfolios, portfolios | 20,000 | 20,000 | The reference firm. |
| Reports per portfolio per year reportsPerYear, reports per year | 4 | 4 | Quarterly reporting. |
| Minutes of drafting and adaptation saved per report minutesSaved, minutes per report | 10 | 30 | Editorial assumption, replace with your own. No deployment on this page publishes a measured saving per report; the only test on this page (Quilter) was a single commentary drafted in about 45 minutes of prompting and editing instead of a few days. |
| Fully loaded cost per hour of the staff who write and adapt commentary hourlyCost, USD per hour | 60 | 120 | Editorial assumption, replace with your own fully loaded cost. |
What it leaves out: Where commentary is not personalized per client today, the saving may show up as better reports rather than fewer hours. The figure leaves out the review effort, platform costs and the effect on client retention.
Who already uses it?
2 public deployments, strongest evidence first. Grades: A regulator or audit, B the organization itself, C vendor case study, D anonymous or estimate.
Morgan Stanley
United States · Wealth and asset management · 2025
BlackRock announced on 2 October 2025 that Morgan Stanley Wealth Management's Portfolio Risk Platform would be the first to implement Auto Commentary, a generative AI feature of Aladdin Wealth, with advisors in the U.S. getting access from October. The tool combines Aladdin risk analytics, the firm's Chief Investment Office outlook and the client's holdings and investment preferences to draft concise insights for the advisor, highlighting issues such as overweights or misalignment with the client's objectives or the firm's market view. Trade press describes the output as bullet point insights inside a template, not full scripts or emails, so it supports the advisor's conversation rather than producing a finished client report. No outcome figures were published.
No outcome disclosed.
Quilter
United Kingdom · Wealth and asset management · 2025
Quilter, a UK wealth manager, rolled out Microsoft 365 Copilot and names meetings and transcriptions as its biggest use case. Microsoft reports that Quilter estimates Copilot will save more than 13,000 hours per month of post call admin time; an investment manager at Quilter Cheviot builds that estimate from an assumed 45 minutes saved per client meeting across 174 investment managers doing about 100 meetings each. Both figures are projections, not measured savings, so neither is recorded as a metric. Quilter also tested turning a portfolio manager interview transcript into an investment commentary: about 15 minutes of prompting and half an hour of editing instead of a few days, which it describes as a one off test.
No outcome disclosed.
How do you implement it?
A model agnostic playbook: what to prepare, the order to build in, and what goes wrong.
Data you need
- Performance, attribution and holdings data per portfolio and period
- House view and market commentary for the period
- Approved templates, disclaimers and banned phrases per market
- Client language and segment preferences
Systems to integrate
- Portfolio accounting and performance measurement systems
- Research and CIO content
- Report assembly and document generation
- Client portal or email for delivery
Complexity: Medium
Data is the hard part: reliable performance and attribution data per portfolio, mapped to a template. The generation is well understood when numbers are inserted, not generated, and output goes through review.
- 1
Start at strategy level
Generate commentary per strategy or model portfolio first, where one reviewed text serves many clients, before personalizing per client.
- 2
Insert numbers, do not generate them
Pass figures as structured data and require the model to reference them, then compare every number in the output with the source automatically.
- 3
Agree the compliance rules up front
Encode disclaimers, fair and balanced presentation rules and banned phrases, and have compliance approve the templates.
- 4
Review by exception at scale
When personalizing per client, review all outputs at first, then move to risk based sampling once error rates are proven low, keeping full review for outliers.
- 5
Keep the decision log
Store the data, retrieved context, draft, edits and approver for every report, so any sentence can be traced later.
Guardrails
- Every figure from the system of record, checked automatically against the source
- Approved templates, disclaimers and banned phrase lists per market
- Human approval before delivery, with risk based sampling only after proven accuracy
- No forecasts or promises beyond the approved house view wording
- Log of data, context, draft and approver for every report
KPIs to instrument
- Days from period end to report delivery
- Number mismatches caught by automated checks per thousand reports
- Reviewer edit rate and rejection reasons
- Share of clients receiving personalized commentary
- Client feedback or complaints about reports
Human in the loop
Portfolio managers or specialist writers approve strategy commentary; reviewers or advisors approve personalized reports; compliance approves templates and samples output.
Common failure modes
- A wrong number reaches a client
- The model restates or rounds a figure incorrectly. Insert numbers from data and check every figure before release.
- Unbalanced performance claims
- Commentary highlights gains and glosses over losses. Encode fair and balanced rules and review for them.
- Generic text at scale
- Personalized reports all say the same thing. Require references to the client's own holdings and drivers.
- Review becomes a formality
- Reviewers approve thousands of reports without reading them. Use risk based sampling with clear accountability.
What are the risks and rules?
EU AI Act
Depends on design
Drafting client reports for human review is not listed in Annex III and is not a practice prohibited by Article 5, so the tier turns on the firm's role under Article 50. A firm that deploys a third party generator (for example a feature of its portfolio platform) for private client reports has no Article 50 duty: the Article 50(4) disclosure duty covers AI generated text published to inform the public on matters of public interest, which private client reports are not, and it lapses anyway after human review under editorial responsibility. For that firm the tier is minimal. A firm that builds the generating system or places it on the market under its own name is a provider under Article 50(2) and must mark the synthetic text in a machine readable format; drafting whole commentaries goes beyond the exemption for an assistive function for standard editing, so for that firm the tier is limited.
Rules that apply
Guidance
- ESMA public statement on the use of AI in the provision of retail investment services (European Securities and Markets Authority, Europe). MiFID II duties on accurate, fair and not misleading client information apply to AI drafted communications, and ESMA expects records of AI use.
- Article 50, transparency obligations for providers and deployers of certain AI systems (European Union, Europe). Providers must mark AI generated text in a machine readable format (paragraph 2); the disclosure duty for text published on matters of public interest does not apply after human review under editorial responsibility (paragraph 4).
- PRIN 2A.5 Consumer Duty: retail customer outcome on consumer understanding (Financial Conduct Authority, Europe). The consumer understanding outcome of the Consumer Duty applies to how performance and risks are explained in reports to retail clients.
Controls to put in place
- Automated number reconciliation between text and source data
- Compliance approved templates with version control
- Decision log per report kept for the retention period
- Risk based sampling with documented reviewer accountability
- Inventory entry for the generation system with an accountable owner
Frequently asked questions
- Who is doing this today?
- BlackRock announced in October 2025 that Morgan Stanley Wealth Management's Portfolio Risk Platform would be the first to implement Auto Commentary, a generative AI feature of Aladdin Wealth, with advisors getting access from that month. The tool drafts concise talking points for advisors from risk analytics, the Chief Investment Office outlook and the client's portfolio, not full client reports. Quilter tested turning a portfolio manager interview into an investment commentary in about 15 minutes of prompting and half an hour of editing instead of a few days, which it called a one off test.
- How do you prevent wrong numbers?
- Never let the model produce figures. Pass them in from the system of record, reference them in the draft, and reconcile every number automatically before a human reviews the text.
- Can personalized reports go out without review?
- Start with full review. Move to risk based sampling only when automated checks and error rates justify it, and keep the decision log for every report.
How to cite this page
Blits.ai AI Use Case Library, "AI generated client portfolio reports and commentary", last verified 27 September 2026, https://www.blits.ai/ai-use-cases/portfolio-reporting-and-commentary. Licensed under CC BY 4.0. Method: how we verify use cases.
Changelog
- 27 September 2026: First published