Target State – Issue 006

Welcome back. This week, the machines got quieter and louder at the same time. Quieter, because the industry's big news was not a deal or a deadline but a set of quiet product launches: Brown Brothers Harriman spun out a technology business to sell the data transformation work every operations team still does by hand, US Bank productised the private equity waterfall, and Broadridge put institutional workflows on tokenised rails. Louder, because the regulators started measuring the AI that is beginning to run inside those machines. The FCA published its first market-wide analysis of the UK alternatives sector built from its own reporting data, and its first multi-firm review of how financial firms use frontier AI. The SEC rewrote the transfer agent rulebook for the first time in four decades and handed private funds nine more months on Form PF. None of it made front pages. All of it changes what your operating model owns.

Half of post-trade uses generative AI. The priority is not more AI, it is fewer errors. ValueExchange's latest research, reported by Global Custodian, finds roughly half of the post-trade industry is now using generative AI in some form, but agentic AI adoption lags materially behind. The finding that matters: the priority firms name when asked what they want from the technology is not more automation but improved accuracy and fewer operational errors. Two years into the AI wave, the industry's stated measure of success has shifted from "did we deploy it" to "did it stop breaking things." That is a maturing market, and a warning to any vendor whose pitch is still about the model.

The FCA's frontier AI review lands on one word: harness. On 2 September the FCA published its multi-firm review of how financial services firms use, test and prepare for frontier AI models with cybersecurity capabilities. The core finding deserves to be pinned to the wall: the utility of a frontier AI model depends heavily on "the governance, tooling, controls, operating context, and human expertise surrounding it" – what the FCA calls the model's harness. The review's practical message is that frontier AI is increasingly "a test of firms' existing organisational, cyber, and operational resilience, rather than a new standalone technology tool." It also finds frontier AI is accelerating vulnerability discovery faster than firms' validation, prioritisation and remediation processes can absorb – and warns that conventional severity scores may miss combinations of lower-rated weaknesses that can be chained together. The harness is the bottleneck. More on that in the Canvas Corner.

The private equity waterfall becomes a product. On 8 September US Bank launched a private markets waterfall engine that automates carried interest, performance fee and distribution calculations for private equity and private credit funds. Why it matters: the waterfall is the most manually intensive calculation in alternatives administration, historically a spreadsheet ritual with the highest error cost per cell in fund accounting. A custodian productising it is the clearest signal yet that private markets operations are moving from bespoke spreadsheets to industrialised engines – and that the administrators intend to own that layer.

Something structural is happening, and it is worth naming: the middle layer of asset management – the data transformation and connectivity work between systems – is being turned into a product that custodians and vendors sell back to the firms that used to do it themselves.

BBH spins out Braid. On 9 September Brown Brothers Harriman launched Braid LLC, a new technology affiliate "focused on solving enterprise data transformation for financial services." The platform is agentic AI-native: business users describe the transformation they need in natural language, AI agents build it, and the result runs in a "controlled and deterministic environment" with a centralised oversight dashboard for breaks and exceptions. The numbers from BBH's 2025 pilot are the pitch: efficiency gains as high as 98% across three use cases – converting data into standard formats, automating trade instruction ingestion more efficiently than traditional extract, transform, load (ETL) tools, and unifying data from more than 150 external providers. The telling detail is who it is for: "business analysts and operations teams, not just data engineers." BBH, a custodian, is selling the connective tissue that used to live inside its clients' operating models.

Broadridge puts institutional workflows on tokenised rails. Also on 9 September, Broadridge launched its platform for tokenised markets, combining multi-chain infrastructure, tokenisation and institutional workflows across traditional and digital asset markets. Tokenisation coverage so far has focused on issuance. Broadridge's move is about the operational plumbing between the two worlds: the workflows that connect digital assets to the custody, settlement and reporting rails institutions already run.

The AI layer's new owners include the firms it will rewire. Mistral's €3 billion Series D, announced on 9 September, is the largest equity raise ever completed by a European company and values the French AI firm at €21 billion. Among the participants: asset managers Eurazeo, Carmignac and BlackRock, alongside Advent, Bpifrance and DST Global. The firms whose operating models AI will most change are now on its cap table. Expect the vendor-client line to blur further as managers buy equity in the technology that will run their middle office.

Architecture Canvas lens: when a custodian spins out a data transformation business and a bank productises the waterfall, the industry is redrawing the boundaries of the operating model from the outside. Every capability that becomes a product is a capability your firm no longer needs to own – and a decision you no longer get to defer. The question for COOs is not whether to use these products. It is which layers of the model you intend to keep, and why.

Northern Trust wins the data conversation at Warwickshire. On 7 September Northern Trust was appointed asset servicer to the £3.6 billion Warwickshire Pension Fund, with a mandate covering global custody, valuation reporting, capital call execution and performance measurement. Asset Servicing Times notes the pitch: bringing "custody data, reporting, and performance information together." The LGPS story is quietly becoming one of the most instructive operating model trends in UK asset servicing: funds under pooling pressure are consolidating providers not around custody price but around the ability to deliver one integrated picture of data, reporting and performance. Custody is the commodity. The integrated picture is the product.

fundcraft's digital fund administration scales on expansion revenue. On 9 September fundcraft secured €12 million in growth financing co-led by RAC and CCAP, bringing total capital raised to €40 million for its digital fund operations platform for alternative investment funds. Chief executive Julian Mayer's statistic is the one that matters: "nearly one-third of new fund mandates committed to fundcraft in the first six months of 2026 came from existing clients." Digital-first administrators are now competing on the same metric the incumbents defend with: existing relationships expanding, not just new logos.

The FCA just measured the alternatives market for the first time. On 3 September the FCA published its first market-wide analysis of the UK alternative investment fund sector, built from regulatory reporting data rather than surveys. The headline numbers: £1.8 trillion of AIFs managed in the UK in 2025, and private credit more than doubling between 2021 and 2025 to roughly £335 billion. The structural findings are the operational ones: leverage and liquidity risks are concentrated in particular strategies and funds rather than spread across the sector, and a small subset of firms manages very substantial pools of alternative assets. The FCA is explicit that the research is "a pre-reform baseline" against which its proposed modernisation of the AIFM regime – and the FRAME reporting reforms – will be judged. For operations teams, the read-through is direct: the reporting you build now will be measured, and the more differentiated supervisory approach the FCA is designing will land differently on different firms. Reporting architecture is no longer a compliance cost. It is how your firm will be categorised.

Liquidity management gets an operating requirement. FCA policy statement PS26/17, published 13 August, finalises new liquidity risk management requirements for authorised fund managers of UK UCITS and NURS funds: firms must maintain appropriate anti-dilution arrangements where dilution poses a material risk, review their operations at least annually, and calibrate arrangements using both explicit and implicit transaction costs. The rules take effect on 1 February 2027. Beneath the policy language sits real operating machinery: swing pricing or dilution levies that must trigger reliably, annual reviews that must be evidenced, and cost data that must be producible on demand. Dealing desks and transfer agents will carry most of it.

The SEC rewrites the transfer agent rulebook, four decades late. On 3 September the SEC proposed the first substantive overhaul of transfer agent rules since they were adopted in the late 1970s and early 1980s: updating existing rules and forms, rescinding one rule, introducing two new ones, and explicitly accommodating electronic recordkeeping and blockchain-based recording. Chairman Paul Atkins' framing is the rationale: the proposal would "reflect transfer agents' current processes and operations." Jamie Selway, director of the SEC's Division of Trading and Markets, adds the principle: "As technology changes and the competitive marketplace evolves, good government requires revisiting legacy rules and regulations." The operational consequence, if adopted, is significant: the official record of who owns what could live on public blockchains, with transfer agents operating the rails. For fund operations teams, the register of holders is the anchor of every investor servicing process. This is the rule that governs it.

Form PF gets nine more months. The SEC and CFTC have further extended the compliance date for the February 2024 Form PF amendments from 1 October 2026 to 1 July 2027, and in April proposed amendments that would significantly narrow Form PF's scope. For private fund operations teams, the message is mixed: more runway on the current build, but a moving target on the final shape. The sensible posture is to keep the data infrastructure investment – every version of Form PF needs the same underlying positions, valuations and counterparty data – and avoid over-building reporting logic against rules that may yet be trimmed.

Moves from the past two months.

This week's operating model insight: The Harness Clause.

The FCA gave the industry a word this week, and it is worth keeping. A frontier AI model, the regulator writes, is only as useful as the harness around it: the governance, tooling, controls, operating context and human expertise surrounding it. Nobody designed the harness. It accreted: the exception queues, the sign-off chains, the escalation routes, the spreadsheets that patch the gaps. The models are new. The harness is the operating model.

Read this week's news through that lens and it all points the same way. BBH productised the harness, spinning out Braid to sell governed data transformation with break and exception management built in. ValueExchange found half of post-trade already uses generative AI, and that the priority firms name is accuracy and fewer operational errors – a harness metric, not a model metric. The FCA's review warns that frontier AI is finding vulnerabilities faster than firms' validation and remediation processes can absorb. Every story is about the layers around the model, not the model.

Three things to do:

The firms that treat AI as a model procurement will inherit whatever harness they already have – which is the operating model they already have, minus the parts that work. The firms that treat the harness as the project will find that the technology finally delivers what the pilots promised. The clause was always there. The regulators just made it explicit.

This week's companion piece on opmodal's Insights page goes deeper: The harness is the operating model: why AI projects fail in the layers around the model.

Target State is published weekly by opmodal Ltd.
Subscribe free target-state.beehiiv.com
Paid tier with deep analysis, benchmark data, and templates launching November 2026.

Architecture Canvas is a methodology developed by opmodal for capturing, governing, and improving operating models in complex businesses.