All insights

Simplification is not subtraction: where complexity goes when the operating model sharpens

Cutting products, outsourcing work and deploying AI do not remove complexity. They relocate it — and every piece needs an address.

By Nick Ross7 min read
Editorial illustration for: Simplification is not subtraction: where complexity goes when the operating model sharpens
See also: Target State – Issue 7

The product review committee spent an hour on the nine funds it had agreed to retire, and ninety minutes on the one fund nobody wanted to touch. A legacy vehicle, a few hundred million, twelve share classes across three jurisdictions, and a data feed into client reporting that predates the current platform. Everyone in the room knew the fund had stopped earning its place years ago. What they also knew, without saying it, was that retiring it would mean unpicking a feed nobody understood, re-papering clients nobody wanted to disturb, and explaining to the CIO why a strategy his predecessor launched was being closed. So the fund survived the meeting, as it had survived every meeting, not because it earned money but because the complexity of removing it outweighed the complexity of keeping it.

That meeting is the sharpening problem in miniature. And this week, the industry told us it has entered a sharpening phase in earnest.

On Tuesday, Northern Trust published Driving Growth in Asset Management 2026, its biennial survey of 300 CEOs, CIOs, operations directors and similar leaders across North America, EMEA and APAC. Three numbers describe the moment. The share of firms prioritising product reduction rose from 5% to 28% in two years. The share planning to achieve their priorities by outsourcing non-core activities more than doubled, from 18% to 39%. And every single respondent reports deploying AI in some form.

What it is

The sharpening is a real strategy now, and it deserves to be treated as one thing rather than three. But every one of its three moves obeys the same hidden rule: simplification is not subtraction. Complexity does not vanish when you cut, outsource or automate. It relocates, and it always takes an address. Firms that do not know the address will meet the complexity again, later, usually as an incident.

The rule has three corollaries.

First, products carry tails. A fund is never just a fund. It is a bundle of processes, systems, data feeds, tax wrappers, jurisdictions and regulatory obligations. When you retire a product without mapping that tail, the work does not disappear. It lands on the surviving funds and the teams supporting them, quietly thickening their workloads and their exception queues. The product shelf is an operating model in disguise, and rationalising it without a process view is surgery performed in the dark.

Second, outsourcing moves complexity to the boundary. Hand a process to a provider and the work still exists, just on the other side of a contract, and the evidence says the boundary is where firms are weakest. The FCA's financial crime review found that around 40% of asset managers outsource parts of their CDD and EDD, but only 36% of those could evidence full oversight of the provider's onboarding, and some could not describe the process at all. Outsourcing more, the direction this week's survey points, makes boundary discipline more valuable, not less.

Third, AI moves complexity into the data layer. Every respondent now deploys AI in some form, while nearly half name consolidating data from multiple sources as their biggest data challenge. That pairing is the trap. An agent does not simplify the model it sits on; it inherits it, at machine speed. Deploy AI against a fragmented data estate and the complexity migrates into the model's answers, where it is hardest to see and easiest to defend.

Why it matters now

Three forces are converging, which is why the sharpening is a 2026 story rather than a perennial one.

The managers are doing it deliberately. Product expansion as a priority fell from 60% to 47%, and Angelo Calvitto, Northern Trust's head of Asia Pacific, compresses the intent into one line: "Managers are doing more with fewer, more strategic partners. The goal is not only lower cost, but a more straightforward operating model with stronger control, better quality and the scale to support future growth."

The vendors are monetising it. Automic launched a bundled fund administration and custody platform with Northern Trust underneath, sold as "a complete and connected operating model." Fund+ launched a regulated fund infrastructure platform in ADGM. Anthropic shipped a complete workflow layer for the financial advisor role. A simplified firm is increasingly a firm assembled from products, each of which owns one layer of what used to be the operating model. That makes the integrator, the manager's own operating model function, the role that cannot be outsourced.

And the regulators are asking to see the boundaries. The AIFM regime consultation will demand evidence of how firms actually operate, not what they own. The SDR entity-level reports land on 2 December for managers above £5bn. When a supervisor asks a firm to evidence its operating model, the firm whose simplification was designed, mapped and owned answers in a week. The firm whose simplification was a series of procurement events discovers how much of the model lives in people's heads.

What firms should do

Map before you cut

Before rationalising the range, map each product to the process estate it drags: its feeds, its wrappers, its filings, its owners. A retirement decision should show its full footprint before it is made. The nine-fund cull is easy. It is the tenth fund, the one with the tail, that decides whether the programme was real.

Give complexity an address

For every outsourcing decision, write down where the complexity goes: which provider owns the process, which manager owns the oversight, and what evidence proves the oversight works. If you cannot write that paragraph, the complexity has no address, and it will find one itself, in an incident queue or a regulator's letter.

Put the data layer before the AI layer

Northern Trust's own survey pairs universal adoption with the confession that nearly half the industry cannot consolidate its data. Nick Gilbert, its head of Asset Servicing, EMEA, says it outright: AI's value "will depend on the quality, governance and accessibility of the data beneath it." Build the golden source, then deploy the agents. The agents inherit whatever they sit on.

The opmodal perspective

The Architecture Canvas methodology captures an operating model across its process, system, people, data and control layers, with owners, sign-off and re-certification attached to each. That is what the sharpening clause turns out to be: a demand that simplification be executed as re-architecture rather than subtraction. Every cut, every outsourcing, every agent has a footprint across those layers, and the firms that can see the footprint before they move are the ones that get to keep the savings instead of paying for them twice.

The meeting I described at the top ends differently in those firms. The tenth fund goes on the board paper with its full footprint attached: the processes it touches, the feeds it owns, the obligations it carries, and the people who would be freed. The argument is no longer about a fund. It is about the model, and what the firm wants it to be. Simplification, done properly, is not about removing things. It is about knowing where everything lives, and choosing.