A conveyancer completes a purchase by running the search, reading what it says and writing the letter as one piece of work. The three steps stay together when they sit in the same place.
Tax-aware portfolio transitions have lived with the opposite arrangement. The cost basis sits with the custodian. The optimiser sits with the platform. The client's tolerance for realised gain sits in the adviser's notes. The scenario gets rebuilt by hand before the meeting. On 27 October 2026, 55ip, the wholly owned J.P. Morgan Asset Management subsidiary, said it would bring its ActiveTax Technology into Claude for Financial Advisors through a connector, Securities.io reported. Anthropic released that suite on 14 September 2026. The point is that the tax calculation, the client constraint and the write-up can occupy the same session. The tax engine performs the calculation. The model holds the conversation.
The connector puts the tax engine inside the drafting session
An application is a destination. A connector is a route into a destination the adviser was already going to visit.
Claude for Financial Advisors is a set of connectors and workflow skills. Securities.io reported that the connectors reach systems advisers already use, among them Charles Schwab, BlackRock's Advisor Center, Addepar, Envestnet, FactSet, Orion, Vanguard and SS&C Black Diamond. The skills turn that access into pre-meeting preparation, an estate and tax brief, portfolio rebalance review, post-meeting notes, prospect intake, an alternatives brief, and a compliance check that screens client-facing language against the SEC Marketing Rule. The 55ip connector adds tax-transition analysis across models, separately managed accounts, direct indexing and unified managed accounts.
Securities.io reported that 55ip had grown assets under supervision to more than $150 billion at 30 September 2026, within a J.P. Morgan Asset Management business that reported $4.8 trillion in assets under management. The release also notes that JPMorganChase is an investor in Anthropic.
The architecture matters more than the announcement. 55ip keeps the tax logic. Claude orchestrates the question, calls the engine, and writes the answer in the shape the adviser needs. That is a claim about architecture. It says nothing about the intelligence of the model. A general-purpose model asked to compute a capital gain across three tax lots will produce a plausible number. A tax engine asked the same question will produce a number that can be reconciled. Once the engine is callable from the workflow, the second number arrives at the same speed as the first. That is the difference between a chat tool and a strategic engine: the chat tool answers a question; the engine runs a client-specific optimisation and hands back a decision the adviser can defend.
Why tax is the hard test for a language model
Language models are good at the sentence around the number. They are poor at the number itself when the number depends on account registration, holding period, lot selection, loss carryforwards and a client's tolerance for a particular level of realised gain. Tax-aware transition is a constrained optimisation problem dressed as a conversation. The model can hold the conversation. Something else has to hold the constraint.
Securities.io reported that advisers using the suite will be able to generate personalised transition scenarios showing the effect of different realised capital gain levels, in-workflow. That sentence is the product argument. The scenario is built from the client's own holdings and the client's own preference, and it is produced where the advice is being drafted.
The table below compares four ways a firm can put tax-aware transition analysis in front of an adviser. They are not equivalent products. They fail in different places.
| Approach | Where the tax calculation runs | What moves between systems | What the client receives | Who can reproduce the number | What breaks at scale |
|---|---|---|---|---|---|
| Standalone tax-transition software | Inside the specialist platform | Holdings file in, scenario file out | A scenario pack, usually after the meeting is rescheduled | The specialist who ran it | The handoff; scenarios go stale between runs |
| Custodian or platform native tool | Inside the custodian's own engine | Positions and cost basis stay in the custodian | A report the custodian formats | The custodian's operations team | Cross-custodian clients; the tool sees one book |
| Model portfolio plus adviser spreadsheet | In the adviser's file | Everything, manually | A spreadsheet the adviser explains | The adviser, if the formulas survived | Version control and key-person risk |
| Connector inside an LLM workflow | In the specialist engine, called from the session | A query and a result, not a file | A scenario drafted in the same session as the advice | The engine, with the session log as evidence | The connector's permissions and the firm's review discipline |
The fourth row puts the calculation inside the drafting environment. It also has the least mature operating history. The trade is real: speed and context against a dependency on permissions, logging and a review step that has to be designed.
Five checks before an adviser runs a live scenario
A connector is not a control. These checks have to be settled before the first live scenario leaves the building.
-
Establish which engine owns the number. Ask the vendor to show the calculation path from cost basis to realised gain to the proposed trade list. If the answer is a model prompt, stop. The value of the 55ip arrangement is that the tax logic is deterministic and separable from the language layer.
-
Test the cost basis handoff against a known account. Take a client with multiple lots, at least one recent transfer in kind and one position with a loss. Reconcile the engine's opening position to the custodian's record before running any scenario. Most failures in tax-aware transition begin with a basis that was stale. They do not begin with an optimiser that was wrong.
-
Set the realised-gain boundary in advance. The client's preference is an input. The firm decides the range it will show. Agree with the investment committee what range of realised gain the firm will show, how it will describe the tax cost, and which accounts are out of scope. A scenario that flatters the tax outcome is a compliance event waiting for a signature.
-
Decide what the adviser signs. Securities.io reported that Anthropic's design keeps the adviser in control, with approval required on critical tasks. The firm's own policy has to say what that approval covers. The signature is on the recommendation. It is not on the model's summary. Record which human reviewed which scenario.
-
Log the session as you would log a meeting. Claude for Financial Advisors is available on Enterprise plans. Securities.io reported that Anthropic recommends those plans for registered investment advisers because they include the audit logs that support recordkeeping. Turn the logging on before the pilot. A scenario that cannot be reconstructed six months later is not evidence.
Where the alpha comes from
BCG's 2026 Global Wealth Report chapter on AI and the future economics of wealth management records that a single product announcement by a small US technology firm wiped more than $140 billion off the market value of several publicly traded wealth managers. Investors read the announcement as a structural signal. BCG's reading is blunt: AI is a structural story for the industry. Two years ago, large language models hallucinated often enough that firms kept them away from client-facing work. Today they draft plans, generate portfolio rationales, automate compliance documentation and execute complex workflows with limited human intervention.
BCG sets out two scenarios. In the displacement scenario, agents handle portfolio construction, financial planning, tax optimisation and client communication at scale. Fees compress. Advantage shifts toward firms with the largest client volumes. In the disruption scenario, AI-first wealth managers expand capacity across the value chain and reshape the economics of advice while keeping a human core. BCG treats disruption as the more likely outcome.
BCG also points to a constraint that matters for tax work. Research on social cognition suggests individuals can maintain roughly 150 meaningful relationships at one time. AI changes that constraint by automating monitoring, servicing and large parts of client engagement. The productive path is to deploy released capacity toward higher-value service and clients who were previously uneconomical to serve. Firms that use AI only to cut cost will hit a ceiling.
Tax-aware transition sits directly on that path. It is a high-value conversation that many firms struggle to deliver at scale because the analysis is slow and the client's tax position is specific. A connector that produces a defensible scenario inside the drafting session turns a specialist task into a repeatable service.
UK Finance reported on Capgemini's 2026 World Wealth Report and found that fewer than one in five clients, 17%, reported receiving a seamless, personalised advisory experience. The same report found that relationship managers dedicate 41% of their time to operational tasks. That is the gap the connector addresses. The adviser gets a scenario built around the client's own tax position, and the operational preparation moves into the tool.
Deloitte's UK perspective on AI in insurance and wealth management describes the market reaction to recent AI launches. Several UK wealth managers saw sharp share-price declines as investors weighed the potential for AI-driven disintermediation. Deloitte's conclusion is that efficiency alone will not be enough. The winners will reinvent distribution, advice and operational performance around these technologies.
What the announcement does not settle
The connector makes the tax engine available inside Claude. It does not turn tax advice into an autonomous process. Securities.io reported that investment recommendations, client communications, compliance determinations and other regulated activities remain subject to human review and approval.
The product also has a boundary. 55ip expects to help advisers prepare for client conversations where tax considerations may be relevant. J.P. Morgan Asset Management will deliver more capabilities through the connector in the future. The first release is a scenario tool for the adviser's workflow. It is not a replacement for the investment committee, the compliance function or the adviser's judgement.
For firms evaluating the connector, the strategic question is simple. Will the firm use the time saved to serve more clients with the same level of care, or will it use the time saved to cut cost and leave the client experience unchanged? BCG's work suggests the first path compounds. The second path reaches a ceiling.
Sources
- 55ip Brings ActiveTax Technology to Claude for Financial Advisors
- AI and the Future Economics of Wealth Management | BCG
- Why wealth managers need more than AI to stay relevant | UK Finance
- AI's Wake-Up Call: Reinventing Insurance and Wealth Management Business Models | Deloitte UK
Where each one earns its place
Standalone tax-transition software
Firms needing specialist platforms that produce separate scenario packs.
Custodian or platform native tool
Clients with assets held in a single custodian book.
Model portfolio plus spreadsheet
Advisers who prefer manual control over formulas in their own files.
Connector inside an LLM workflow
Advisers seeking speed and context by drafting advice and scenarios in one session.
Source: 55Ip Brings ActiveTax Technology to Claude for Financial Advisors, Securities.io