Weekly

1. MT Newswires enters Microsoft Copilot: financial news follows MCP into office software

Integration

MCP-based access, enabled without code in Copilot and Copilot Studio

Use cases

Query company news, earnings, macro and sector developments in Word, Excel, Outlook and Teams, with source links

Value proposition

Structured, tagged original content with traceable, verifiable answers for regulated users

Distribution

Its MCP server was already available through Anthropic Claude and Perplexity

The news agency makes its distribution logic clear: clients do not want another platform to open; answers need to appear where they already work. The value of the news source stays the same, while its shelf moves.

This is an official MT Newswires release, including comments from a Microsoft product executive. No usage figures were identified.

Weston Walker, head of global sales and partnerships at MT Newswires, says in the release that clients want answers in their existing work environment rather than another platform to consult.

Further reading: source ↗

2. Yingmi launches private-fund MCP tools for qualified investors using general-purpose agents

Institutional offering

Yingmi Private Funds 1000 offers institutional share classes, with over 500 private-fund managers and over 1,000 products

Retail offering

Yingmi Private Funds Select 100 sits within a buy-side advisory framework of planning before allocation

Private-fund MCP

Product search, performance queries and matching become standardized tools. Verified qualified and professional investors can request an API key and call them through general-purpose agents

Infrastructure

Digital support throughout pre-investment, investment and post-investment processes

This is an early example among Chinese fund distributors of turning private-fund data into MCP tools. The access design matters: permissions depend on investor qualification, while the data endpoint sits inside a general-purpose agent.

Securities Daily reports on the company announcement. The coverage figures are company-reported.

Further reading: source ↗

3. Taktile Insights lets business teams question their decision data in natural language

Capabilities

Natural-language questions and instant charts within the decision platform; analyze approval rates, claims and other outcomes down to individual decisions

Customer account

A MagaluPay analyst says a policy change took three to four weeks in the old engine, versus 40 minutes to validate and deploy in Taktile

Customers and funding

Customers include Allianz, Mercury, Monzo and Faire. Total funding is $184 million; its $110 million Series C was led by Goldman Sachs Growth Equity

Its positioning is straightforward: as AI speeds up decisions, the next differentiator is how quickly organizations learn from them. Taktile's answer is to put analysis into the hands of business teams accountable for outcomes, beyond the BI team.

This is a vendor release. The efficiency figures are the customer's own account within it.

Further reading: source ↗

4. fileAI signs with AWS to turn financial documents into auditable agent workflows

Partnership

A strategic collaboration with AWS, built on Amazon Bedrock and delivered through AWS Marketplace to banks, insurers and financial institutions in Asia and the US

Products

fileForge turns enterprise documents into governed process intelligence; fileScout plans model routing before extraction to reduce token use; fileForensics feeds fraud signals into the processing pipeline

Customers and investors

Customers include MS&AD, Toshiba, PwC and KPMG. The company reports processing over one billion documents and receiving investment from SMBC Asia Rising Fund and Singtel Innov8 in August

Another company is making unstructured-data governance a selling point. Before financial agents can work, someone must ensure their input documents are clean, traceable and auditable. The Singapore-based company serves markets across Asia and the US.

This is a vendor release; document volumes and customer references are vendor-reported.

Further reading: source ↗

5. Tandems launches Hamilton, an adviser assistant it says can handle 80% of daily tasks

Scope

Financial planning and portfolio rebalancing, client reviews, meeting preparation, account opening, asset transfers and paperwork; the vendor claims coverage of around 80% of advisers' daily tasks

Architecture

Task-based routing across Claude, ChatGPT, Copilot, Gemini and customers' own models

Integrations

Pre-integrated with custodians including Schwab, Fidelity and Pershing, plus FIS, Fiserv and Salesforce; the company says deployment can take weeks

Scale

The platform supports $80 billion in assets, 2.7 million accounts and over 5,000 advisers

Controls

Real-time shutdown of individual skills, models or all AI activity; conversations adjusted to US state AI laws; audit logs of all AI interactions

Hamilton distinguishes itself from note-taking assistants by emphasizing multi-step execution across legacy systems. This is a direct move by a wealth technology provider to become the adviser's primary interface.

This is a vendor release. Claims including the 80% figure have not been independently verified.

Further reading: source ↗

6. Two RIA paths: $20 billion Modera builds Mira; LPL offers Zocks to 32,000 advisers

Modera Mira

The RIA, with $20 billion under management and over 6,300 clients, built a platform combining wealth technology data and document-system metadata; used internally since July

Acquisitions

Newly acquired firms can merge data into the shared platform, supporting its acquisition-led expansion

Zocks × LPL

Conversation capture and data synchronization tools are available to over 32,000 LPL advisers and around 1,100 institutions, integrated with WealthVision planning and CRM

Both seek adviser efficiency through different routes. Firms with scale build to retain control of their data assets; platforms procure established tools to embed in existing workflows. Both approaches advanced this week.

This is PLANADVISER's weekly product roundup, based on company disclosures.

Further reading: source ↗

7. Deutsche Bank says some corporate loan decisions now take minutes; ADA oversees its agents

Reported results

Small-business loan decisions cut from weeks to minutes, and multinational lending from months to weeks; the project went live this summer

Platform

ADA, named after Ada Lovelace, develops, monitors and controls AI agents across the bank

Investment

Over €2.5 billion in additional technology investment planned for 2026–2028

Next steps

Incoming management-board technology executive Deverdun says preparations are underway for the bank's agents to interact directly with clients' agents; cyber-risk management is his top priority

CEO Christian Sewing has set out an ambition to make Deutsche Bank a genuinely AI-driven bank. Beyond faster approvals, the revealing detail is its preparation for direct interaction between the bank's agents and its clients' agents.

This is a Reuters interview with the incoming technology executive, republished by Euronext. Performance figures are the bank's own account.

Further reading: source ↗

8. ICBA and OpenAI ally to bring AI cybersecurity tools to community banks

Alliance

The Independent Community Bankers of America and OpenAI formed a strategic alliance to help smaller community banks access AI-powered cybersecurity capabilities

Audience

Both banks ready to experiment with frontier AI and banks still developing an AI strategy

Context

Community banks are increasingly active in advanced-technology policy, but access to AI security tools remains a weakness

Large banks build platforms; smaller banks work through associations. Community banks do not have an extra €2.5 billion budget. An association negotiating bundled access with a model provider offers a practical route for thousands of smaller US banks to obtain AI security capabilities.

This is an Inside AI Policy report behind a paywall. The supplied draft notes that its publicly available portion was checked. No specific product launch date was identified.

Further reading: source ↗

9. Lemonade Agentic: an insurance interface built for customers’ personal AI agents

Positioning

The company describes the world's first fully agentic insurance experience, built for customers' personal AI agents on the Lemonade OS stack

Agent actions

16 operations including quotes, purchasing, renewals, coverage changes, adding vehicles or pets, filing claims, uploading claim photos, and changing addresses or payment methods

Status

Not yet live; a beta waitlist is open. Its slogan is “Have your agents talk to our agents”

While many companies discuss how to block outside agents, Lemonade is building them a dedicated channel to work inside its systems. The bet is that agents will reshape acquisition costs and customer experience.

This official product page describes a vision, with no live adoption data. It follows the direction of Kin's WebMCP-enabled ChatGPT quoting, but goes further: Lemonade aims to extend access from quotes to purchasing and claims.

Further reading: source ↗

10. Chengfang forum: major institutions converge on AI assistance with final human review

Academician Zheng Weimin

Tokens are becoming a new form of oil. Mooncake's disaggregated inference architecture supports heterogeneous hardware and cross-region coordination to reduce cluster inference costs

ICBC vice president Zhao Guide

AI is moving from generation to action, shifting the focus to realized value; ICBC's digital-intelligence model covers its main businesses

CITIC Securities CIO Yu Xinli

He calls this the first year of financial agents, with capital-markets AI entering deployment at scale, while retaining final human review

CPIC Group CIO Wang Lei

Presented CPIC's 633 AI+ strategy alongside the construction of an AI security governance system

Agricultural Bank of China hosted the Chengfang Fintech Forum during the annual Financial Street Forum. Institutions used different language but put final human review at the forefront. Stronger claims about deployment at scale came with clearer statements of boundaries.

This is Science and Technology Daily's conference report; statements reflect remarks at the event.

Further reading: source ↗

11. IDC: financial AI moves toward agent factories; city commercial banks join Innovation 50

Investment mix

IDC observes that financial-sector IT investment grew by less than 10% in 2025, while generative-AI investment grew by close to 50%

Selection

IDC's China Financial AI Innovation 50 drew from over 130 applicant institutions; selected banks included ICBC, China Minsheng Bank, Bank of Hangzhou and Bank of Ningbo

Bank of Ningbo

An AI coaching platform uses multiple agents for knowledge Q&A, skills training, automated scoring and job-competency assessment

Bank of Hangzhou

A retail customer platform links customer insights, product analysis and business strategies through multiple agents, covering wealth, consumer credit and credit cards

Foundation

Ant Digital Technologies' financial model and Agentar development platform provide engineering support

IDC China research manager Si Erxun argues that banks are moving from competition in manpower toward accumulating capabilities as assets. The next measure is how much reusable, collaborative and continuously improvable capability each use case leaves behind, beyond the number deployed.

This is JRJ's report on the IDC summit. Investment growth uses IDC's figures; case details come from company and IDC presentations.

Further reading: source ↗

12. Shanghai Document 22 encourages models and general-purpose agents in risk control

Document

Shanghai Financial Document [2026] No. 22, Measures for High-Quality Digital Finance Development in Shanghai Banking and Insurance, implementing Financial Office Document [2025] No. 95

AI use cases

Intelligent marketing, credit, claims assessment, risk control and internal management; coordinated domain-model development and enterprise AI platforms

Frontier exploration

Applications and safety validation of industry models and general-purpose agents in financial risk control

Smaller institutions

Combine rented and purchased computing capacity to avoid duplication; acquire mature technology from large banks commercially and share jointly built digital infrastructure

Follow-up

Regulators will explore a digital-finance evaluation system to monitor development and outcomes

The local regulatory document explicitly includes general-purpose agents in risk-control use cases and gives smaller institutions a practical cost route: rent capacity from peers and buy mature solutions from larger banks rather than building everything themselves.

The linked source is the official document.

Further reading: source ↗

13. FCA outlines good and poor AI practice and a trusted framework for agentic finance

Approach

Use existing frameworks rather than a separate AI regime; practice examples do not create new rules

Three areas

Consumer Duty: define the consumer problem before building tools. Governance: executives must understand AI risk in their remit. Operational resilience: include AI dependencies in stress tests and exit plans

Agentic finance

Work with industry, other regulators and frontier labs on a trusted framework covering agent identity, authorization, accountability and execution

AI providers

Provide clear, proportionate and consistent disclaimers to help consumers understand the limits of model-generated financial information

The poor-practice list is particularly useful: launching before defining the consumer problem, handing all AI governance to technology teams, and accepting supplier assurances without verification. Each points to a weakness supervisors can examine.

This item is based on Freshfields' October 22 analysis. FCA chief data officer Jessica Rusu also spoke that day about using AI in the regulator's own processes; this item focuses on expectations for institutions.

Further reading: source ↗