The client's request sounds reasonable, she wants to remain invested in her firm's balanced model portfolio, but with a few adjustments. The client has a large holding in her former employer, so she does not want further exposure to the same sector. She would like to avoid a handful of companies for ethical reasons and may need access to capital in the next two years, and her adviser knows that selling certain legacy positions too quickly could create an unwelcome tax bill.
This is not an unusual conversation. It is increasingly the kind of service wealth managers want to provide: a portfolio built around a client's actual circumstances, rather than one defined solely by a risk label.
The difficulty begins after the meeting as the portfolio team must determine how the proposed changes affect the client's strategic allocation. The adviser must be satisfied that the recommendation remains suitable. Operations may need to translate instructions across systems. Compliance may need evidence that restrictions have been implemented properly. The tax implications may need separate review. Future rebalances must remember that this client is not quite like everyone else in the model.
To the client, this feels like personal service, inside the firm, it is an exception.
Multiply that experience across hundreds or thousands of clients and the challenge becomes clear. Wealth firms are trying to offer more bespoke model portfolios while relying on systems, processes and data structures built for a more standardised world. The result is rarely one dramatic failure. More often, it is a gradual build-up of operational strain: spreadsheets maintained by a few experienced people, manually applied restrictions, fragmented records, and portfolios whose rationale is understood but not fully captured anywhere.
This is the invisible complexity problem as many wealth firms did not consciously design the operating system they use today. It evolved. A new CRM was introduced after an acquisition. An investment platform solved an immediate need. A spreadsheet was created to manage a small number of client exclusions. A manual report filled a gap in a compliance process. Over time, products, wrappers, mandates and client preferences accumulated.
Each decision may have been perfectly sensible. Together, however, they can create an operating system that no one intended but everyone depends upon.
For a firm with a tightly controlled suite of central models and only limited variation, this can work. Model portfolios create clarity: an investment philosophy, a repeatable process, defined governance, and a clear basis on which to monitor outcomes. That standardisation is not merely efficient. It is a source of investment discipline.
But client expectations are changing. Investors want portfolios that reflect their tax position, ethical preferences, concentrated holdings, income needs, sustainability views and life-stage goals. Advisers want to meet those needs without moving every client into a fully bespoke, high-cost discretionary mandate.
The ambition is compelling - personalisation at scale.
Yet personalisation changes the nature of a model portfolio business. The more a portfolio is adapted for an individual client, the greater the risk that it ceases to operate as a model in any meaningful sense.
Each deviation may make sense on its own. One client's sector limit, another's ethical screen, a third's legacy shareholding and a fourth's tax constraint may all be appropriate. But each creates questions that must be answered consistently. Does the portfolio still meet the intended risk profile? Is the deviation temporary or permanent? What happens when the underlying model changes? Which clients should follow that change, and which require different treatment?
Without a deliberate way to manage these questions, personalisation becomes an accumulation of exceptions.
The risk is not only operational. It is a risk to investment-process discipline itself. When a firm cannot clearly identify the difference between the underlying model and each client's version of it, governance weakens. Advisers may make reasonable but inconsistent judgments. Portfolio managers may lack a complete view of client-level deviations. Compliance may struggle to reconstruct why a decision was taken, who approved it, and whether it remains appropriate.
Over time, a firm can lose the ability to distinguish between deliberate personalisation and accidental drift.
The traditional response is to add people, spreadsheets and checks. These measures can keep the system working, especially in the short term. But they do not make it coherent. They often increase dependence on tacit knowledge: the portfolio specialist who knows where the latest restriction list sits, the operations manager who understands an account's unusual history, or the adviser who remembers why a client was not rebalanced six months ago.
Heroic effort is not a scalable operating model, as such this is where agentic AI has a potentially important role. Not as an autonomous portfolio manager, and not as a substitute for advisers, investment professionals or compliance teams. Instead, it can become an intentional operating layer around the investment process.
An agentic workflow could identify portfolio drift, a breach of a client-specific restriction, a tax opportunity, a cash need, or a relevant change to an underlying model. It could gather the client mandate, portfolio holdings, model rules, tax information and prior decisions. It could assess the situation against defined policies, then prepare a recommended action and supporting evidence.
The value is not simply speed, it is visibility and consistency.
A system might identify that a client's equity exposure has risen because a legacy holding has appreciated. It could show that an immediate sale would create a material taxable gain, confirm that the client mandate allows a temporary deviation, and propose rebalancing the remainder of the portfolio while escalating the concentrated holding for adviser and tax review.
That is not autonomous investing. It is controlled automation that makes complexity visible and turns it into a structured workflow.
The human approval point is central to this model. Investment decisions are not simply optimisation problems. They involve judgment about client objectives, suitability, market uncertainty, tax consequences and fairness. They involve accountability.
An adviser may know something about the client's priorities that is not contained in structured data. A portfolio manager may conclude that a recommended deviation compromises the investment philosophy. A compliance professional may judge that the evidence is insufficient. These are not exceptions to effective automation; they are the reason for its boundaries.
The role of AI should be to identify, gather, compare, propose and document. The role of the authorised professional is to review, challenge, approve, amend or reject. Escalation should be triggered where a decision is material, uncertain, outside policy or shaped by unusual client circumstances.
Done well, this can create a stronger control environment than one dependent on inboxes, manual hand-offs and local spreadsheets. It can also improve the working lives of the people inside it. Advisers spend less time finding information and more time discussing real choices with clients. Portfolio teams spend less time resolving avoidable ambiguity and more time overseeing the investment process. Compliance teams can focus on meaningful exceptions rather than reconstructing routine decisions after the fact.
The objective is not to make bespoke portfolios look identical to standard models. It is to ensure that their differences are intentional, visible and governable.
That requires firms to define what may be personalised, what is non-negotiable, and what requires escalation. It requires a connected view of client preferences, mandates, holdings, model rules, tax circumstances and approval history. And it requires leaders to see personalisation not as a front-office feature, but as an operating-system challenge.
The firms that succeed will not necessarily be those with the most ambitious AI programme. They will be those that clarify their investment guardrails, define decision rights, connect the right data and automate where automation strengthens consistency without obscuring accountability.
To clients, the result should feel simple: a portfolio that reflects their circumstances, managed with care and explained clearly. Behind that simplicity must sit an operating system designed deliberately, rather than one assembled by accident.
Until next time.
Allan Lane