This session covers where the industry stands, where Federated Hermes is positioned, and the decisions that will determine whether AI becomes a structural edge or a sunk cost.
Foundation models and agentic workflows are on every roadmap across the industry. The gap between ambition and production-grade deployment is where the real strategic question sits — and closing it requires a clear-eyed view of what AI actually is, how capital is flowing, and where the genuine risk lies.
The term "AI" is used to describe fundamentally different technologies with different maturity profiles, risk characteristics, and investment timelines. Clarity on which member of the family is under discussion is the first discipline a board should apply.
Quietly generating real ROI in portfolio construction, risk modelling, earnings analysis, and credit scoring for years. Proven, interpretable, reliable — and underappreciated relative to newer technologies that attract more attention. The unsung hero keeping the household running.
The ChatGPT moment shifted public and institutional attention decisively. Creative, fast, and increasingly capable — but still prone to hallucination, accuracy issues, and inconsistent execution. All potential, inconsistent delivery. Infrastructure buildout is now ($500B+ in 2026); enterprise monetisation follows from 2026–27.
Just learning to walk. Agentic systems that can plan, reason, use tools, and execute multi-step workflows are showing real flashes of capability — but governance frameworks, security architecture, and reliability standards are still being built. Rapid growth is expected; patience is required.
Discussed constantly; does not yet exist. Artificial General Intelligence capable of human-level reasoning across arbitrary domains remains speculative. It causes more board debates than any technology that is actually in production — and should be weighted accordingly in investment and strategy discussions.
"The operating model changes required to embed AI into business processes are the real differentiator — not the technology investment itself. Firms that treat AI as a technology project rather than an operating model transformation will not capture the value."
Alpha FMC Global Outlook 2026
AI investment flows through sequential waves, each with different beneficiaries, timelines, and risk profiles. Timing across these waves is as important as whether AI succeeds at all — each wave can disappoint independently even if the broader thesis holds.
Nvidia (GPUs — 5.5% of MSCI World), Broadcom (networking), Micron and storage players, hyperscalers building capacity. Capital required: $500B+ in 2026 alone. GPU utilisation currently near 100%.
Efficiency gains (e.g. DeepSeek-style model compression) could reduce compute requirements sharply. Accelerating GPU generation cycles may strand investment in current-generation hardware.
Snowflake, Databricks (data architecture), Palo Alto, CrowdStrike (cybersecurity), hyperscalers again (infrastructure-as-a-service). Infrastructure is ready; enterprises are moving from pilots to deployment. 3–5 year runway.
Efficiency innovations can reprice compute demand overnight. Cloud backlog growth vs. revenue growth is the key metric. Each new model generation introduces repricing risk for the previous wave.
Anthropic, OpenAI (frontier LLMs), Microsoft Copilot across Office and Azure, ServiceNow, Salesforce (AI in workflows), enterprise platforms with proprietary data moats. AI revenue as a percentage of total platform revenue is the metric to watch.
Point solutions face rapid disruption from native AI tools built into existing platforms. This is where stock-specific research matters most — the spread between winners and losers widens sharply.
Current AI productivity contribution: 10–20bps annually. At full scale: 50–150bps annually. Multi-trillion dollar economic impact. Winners are companies that successfully integrate AI into core operations; losers are those that do not. This is where the gap becomes permanent.
Most difficult wave to quantify and easiest to get timing wrong. Market pricing today assumes success across all four waves simultaneously — history suggests reality is considerably messier.
Proof-of-concept work is broadly complete. The firms pulling ahead are those translating pilots into production-grade systems with governance, reliability, and measurable outcomes. The majority have not made that transition yet.
74%of firms are struggling to scale AI beyond pilots
The real strategic question is which components to select and how to assemble them. Cloud infrastructure, data platforms, and foundation models are procured. Competitive advantage comes from proprietary data quality and the depth of use cases built on top — not from model selection.
Multi-agent systems that plan, execute, and route across tools are transitioning from research to production. Investment management CTOs are actively building Model Context Protocol (MCP) capability as named infrastructure in 2026 — not a future consideration.
63% of firms cite data management and quality as the top pain point limiting AI value delivery. Without governed, semantically consistent data, agent outputs cannot be trusted or scaled. MDM is a strategic sequencer, not a background IT project.
63%cite data management as the top AI constraint
AI accelerates attack sophistication faster than most defence budgets can respond. The Allianz Risk Barometer ranks cyber incidents as the number-one business risk for the fifth consecutive year, with AI ranked second and rising. Markets are not yet pricing cyber resilience as a competitive moat — but they will.
Next-best-action intelligence, automated RFP and DDQ responses, personalised content, and pre-meeting client briefs are delivering measurable efficiency gains. Distribution is where AI ROI is most attributable and adoption is accelerating fastest across the peer group.
Employees use AI tools informally in day-to-day work, outside any firm-approved framework. High productivity upside; significant governance and data security risk. Almost universal across the industry whether firms acknowledge it or not — and a signal to accelerate the official programme rather than restrict access.
Secure, firm-approved AI available for drafting, search, and retrieval-augmented generation on internal documents. Most large asset managers are at or approaching this stage. M365 Copilot and similar enterprise rollouts represent the typical entry point. FHI's firm-wide Copilot deployment sits here.
Deployment of native AI capabilities embedded in existing SaaS platforms — Salesforce Einstein, Seismic AI, SimCorp Copilot. Fast to activate, limited to platform boundaries, and increasingly becoming table stakes. Useful, but insufficient for structural differentiation on its own.
Custom AI systems accessing multiple databases and applications — proprietary agents on firm data, multi-LLM orchestration, agentic workflows. This is where structural competitive advantage is built. Athena, Iris, and Pacer sit in this category. Most firms have not arrived here yet.
AI investment is concentrated in areas with clear attribution to revenue, efficiency, or client experience. The following landscape covers where the industry is deploying — and where Federated Hermes has material opportunity.
AI-driven identification and prioritisation of high-probability prospects by channel, product, and market conditions — replacing manual territory research with continuous signal processing across third-party and CRM data.
Intelligent daily signals surfacing which advisors, consultants, or clients to contact, what to say, and which content is most relevant — built on CRM, flow, and market data. The Iris programme is FHI's production implementation.
AI-generated first-draft responses grounded in the approved data room, compliance-checked before human review. Reduces response cycle time from weeks to hours. Already in production across several large peers.
Automated briefing documents assembling client AUM, flows, service history, portfolio health, and market context ahead of every meeting — removing manual preparation time entirely for the coverage team.
AI-assisted reading, classification, and draft responses for common service queries — routed by type, escalated by complexity, with human review on outbound. Reduces operational bottleneck in client services.
Market intelligence combining third-party flow data, Broadridge, and CRM to identify share-of-wallet opportunity, competitive displacement, and channel sizing — feeding coverage prioritisation decisions.
AI-assembled pitch decks from approved content blocks, tailored to the prospect's current holdings, mandate type, and stated priorities — without manual production time on the distribution team's side.
Machine learning models identifying at-risk client relationships based on engagement, flow trends, service interactions, and market conditions — enabling proactive outreach before mandates are at risk of redemption.
AI combining earnings transcripts, analyst reports, filings, and alternative data into decision-ready summaries — first-draft memos produced in seconds, human-reviewed and approved. Works; does not replace judgement.
AI processing earnings transcripts to extract financial signals, management tone shifts, competitive mentions, and guidance changes — catching signals before consensus catches up. Proven alpha in practice.
Continuous AI monitoring of exposures, benchmark drift, factor tilts, and risk limits — surfacing alerts in natural language at the point of decision. The Pacer programme is FHI's implementation of this capability.
AI normalising and synthesising ESG signals from multiple external and proprietary sources into consistent scoring frameworks. A structural advantage for FHI given the depth of proprietary stewardship and engagement data.
Satellite imagery for retail traffic and construction; web scraping for pricing, reviews, and job postings; foot traffic and credit card data. AI brings alternative data into the research process at a speed and scale impossible manually.
Multi-factor bond similarity, correlation matrix cleaning (noise suppression via random matrix methods), tail risk scenario modelling, and Monte Carlo optimisation. Traditional ML doing reliable, unglamorous work that quantitative teams depend on daily.
Agentic AI identifying, classifying, and resolving routine settlement breaks using pre-approved playbooks — escalating only genuinely ambiguous cases to operations teams. Demonstrably reduces headcount requirements at scale.
AI extracting and populating subscription documents, performing AML and KYC pre-checks, and tracking onboarding status — compressing a multi-week process into days across both retail and institutional channels.
Continuous AI monitoring of trades, communications, and reporting obligations against current regulatory requirements — flagging exceptions with evidence for human review rather than requiring manual surveillance sweeps.
AI review and maintenance of distribution agreements across counterparties — automated alerts for renewal dates, cap breaches, and compliance obligations within complex cross-border agreement structures.
AI-assisted production of regulatory filings and investor reports — first-draft output grounded in approved data sources, reducing manual production time and error rates in high-stakes deliverables with hard deadlines.
AI extracting key terms, obligations, and risk flags from complex legal documents — providing structured summaries to legal and compliance teams without replacing their review and ultimate judgement.
AI producing first-draft thought leadership, fund commentary, and campaign materials from approved briefs — reducing production time substantially while maintaining editorial standards and regulatory compliance requirements.
AI analysing multi-channel engagement and campaign performance to identify what is driving leads, conversions, and advisor activation — replacing manual monthly reporting cycles with continuous intelligence.
AI adapting website and email content to the interests, segment, and behaviour of individual visitors and contacts — moving from broadcast to personalised engagement at a scale no manual editorial team can achieve.
AI monitoring market commentary, competitor communications, regulatory announcements, and media coverage — providing real-time intelligence on positioning gaps and messaging opportunities across distribution channels.
AI-assisted processing of expense allocation, variance analysis, and management reporting — reducing the manual overhead of finance operations and improving the speed and accuracy of information reaching senior decision-makers.
AI supporting recruitment screening, skills mapping, and internal mobility — creating more consistent and data-informed people decisions, and identifying capability gaps before they become operational constraints.
AI reviewing vendor contracts, tracking obligations, and flagging renewal and renegotiation opportunities — applied across a vendor estate that typically runs to hundreds of agreements in large asset managers.
AI enabling employees to search across internal documents, policies, research, and institutional knowledge using natural language — reducing time spent locating information and improving the use of existing intellectual capital.
| Lifecycle Stage | Highest-Focus AI Applications | Industry Priority | FHI Opportunity |
|---|---|---|---|
| Create Awareness | Content generation, campaign automation, digital personalisation, social intelligence | Medium | Content velocity and campaign ROI at lower cost-per-lead |
| Develop Pipeline | Lead scoring, prospect clustering, market sizing, territory and coverage optimisation | High | Advisor and institutional prospecting at scale with data-driven prioritisation |
| Convert Opportunities | RFP and DDQ automation, pitch personalisation, pre-meeting briefs, compliance review | High | Reduce response cost and cycle time; improve win rates across institutional channels |
| Onboard Clients | Subscription document processing, KYC automation, onboarding tracking, account funding support | Medium | Faster time-to-funded across both retail and institutional book |
| Service Accounts | Query response, relationship health monitoring, meeting summarisation, next-best-action | High | Scale service quality without proportional headcount growth; proactive retention |
| Report & Review | Commentary generation, regulatory reporting, attribution narratives, ESG disclosure drafting | High | Reduce production cost and cycle time; improve consistency at volume |
Federated Hermes is in production — not in planning. The strategic foundations are in place: executive sponsorship, a modern data platform, a dedicated team, and an articulated framework covering people, platforms, and proprietary data.
"The industry has shifted. Foundation models, embedded copilots, and agentic workflows are on every roadmap — but most asset managers have not translated proof-of-concept into production-grade systems. Federated Hermes is an exception: we are in production."
Frank Amato, Enterprise AI & Data Science · July 2026
M365 Copilot Chat available firm-wide, with role-based licensing for specialist copilots where the case is clear. Custom models and agents accessible through Athena, our internal AI marketplace, extend capability into specific workflows without requiring technical expertise from end users.
Activating the native AI capabilities built into existing partner platforms — while selectively building custom AI models and agents alongside them where the value case is clear and where SaaS capability falls short of what the business genuinely requires.
Proprietary models and agents built on Federated Hermes data to improve investment insight, client intelligence, and operational decision-making. This is where structural competitive differentiation is created — not in foundation model selection, but in the quality of data and depth of use cases built on top of it.
Our internal marketplace of AI tools — assistants, workflows, agents, and models — built for teams across the firm. Athena is the distribution mechanism for AI capability: a governed, searchable library of tools that teams can deploy without engineering dependency on a central team.
Data-driven, agentic AI insights that drive next-best-action for sales and relationship teams. Iris brings together CRM, flows, client data, and market signals to surface prioritised daily actions across the distribution organisation. Lead: Jonah Woods and Hanne Richardson.
An AI-powered portfolio manager workspace that answers questions across the full investment lifecycle — combining portfolio data, market intelligence, and research synthesis in a single interface. Lead: Dan Clymer and Kaavya Subramanian.
Beyond the operational use cases most firms are now pursuing, a distinct set of more structurally disruptive opportunities is emerging. These represent the next generation of AI application — and early movers are already building.
The next frontier in client servicing is AI agents that operate on behalf of the asset manager — answering investor questions, providing portfolio updates, and surfacing relevant content without requiring human intermediation for every interaction. This is not a chatbot: it is a persistent, knowledgeable agent grounded in fund data, portfolio holdings, and client relationship history.
For Federated Hermes, the Responsible Investment story is a natural anchor. Investors increasingly want direct, informed access to stewardship and ESG positions — not just annual written reports.
AgenticClient ExperienceDifferentiatingT. Rowe Price Labs represents an approach to AI innovation worth studying carefully: a dedicated internal function with a mandate to rapidly prototype, test, and productionise AI use cases — distinct from both the core technology function and individual business lines.
The model works because it provides speed without the governance overhead that slows enterprise IT, while maintaining accountability structures that prevent uncontrolled proliferation. Several other large asset managers are building similar functions. The question for FHI is whether Athena evolves into this model or whether a separate innovation mandate is needed.
Operating ModelInnovationThe traditional asset manager website as a primary distribution channel is being disrupted. Content is increasingly consumed through embedded AI tools, adviser platforms with integrated intelligence layers, and direct data feeds — not web visits. AI agents, not human browsers, are becoming the primary consumer of investment content.
This means structured data, machine-readable content formats, and API-accessible fund information become distribution assets in their own right. FHI's digital strategy needs to account for this transition explicitly.
DistributionDigitalForward-LookingFederated Hermes holds a distinct position in ESG and responsible investment. AI creates an opportunity to extend that position meaningfully: synthesising stewardship engagement data, proxy voting records, and responsible investment signals into client-facing intelligence products that competitors cannot easily replicate — because doing so requires both the data and the institutional credibility to make it trusted.
This is a genuine first-mover opportunity in an area where FHI has a structural head start that no technology investment alone can replicate.
ESGStewardshipProprietary DataThe Iris programme is the foundation for a more ambitious vision: an AI coverage model where agents handle the majority of adviser and intermediary engagement through digital channels — surfacing product insights, answering questions, scheduling human touchpoints — while the human distribution team focuses on complex and high-value relationships where personal engagement is irreplaceable.
This is not replacing relationship managers. It is extending their effective reach across a coverage universe that no human team can serve personally at the required depth.
DistributionAgenticScaleInstitutional clients and sophisticated retail investors increasingly expect on-demand access to portfolio data, attribution, and commentary — not quarterly PDF reports arriving by post. AI makes continuous, personalised reporting technically and economically feasible for the first time: automated commentary from live data, natural language portfolio queries, and reporting that would previously have required a dedicated operations team.
Client ReportingOperationsClient ExperienceBeyond individual use cases, AI is enabling new ways of structuring the work itself — changes that compound across every function over time and represent the operating model transformation that separates early movers from followers.
Rather than managing AI as a collection of individual projects, leading firms are establishing dedicated AI product teams with continuous ownership of use cases — building, measuring, improving, and deprecating AI products on a rolling cycle. This is the model Athena is structured around, and it is the right delivery architecture for sustained innovation.
The greatest risk in AI deployment is bolting AI onto existing processes and measuring the time saved. The firms generating structural advantage are redesigning work around AI — asking what the right division of labour between human and machine is for each task, rather than how to automate the existing process unchanged.
BlackRock's approach to AI deployment within Aladdin is instructive: rather than building one large AI system, approximately 50 to 60 specialist teams each own a specific agent or plugin. Each agent does one thing well, is maintained by the team that understands the domain, and integrates into a common orchestration layer.
AI is eliminating the cost constraints that historically limited content production in asset management. The strategic shift is from managing content as a scarce resource to managing it as a network — where taxonomy, distribution, and personalisation determine value more than production volume alone.
Most AI programmes fail not because the technology does not work, but because the surrounding conditions are not in place. The considerations below are the practical determinants of whether AI investment delivers sustained returns at Federated Hermes.
Traditional IT budgeting was built around licence costs and headcount. AI introduces a fundamentally different cost model — one that scales with usage, not users.
How AI capability is built and governed is as important as what is built. The delivery model determines pace, quality, and the firm's ability to maintain and evolve what it deploys.
The decisions made in the next twelve months on data infrastructure and orchestration will constrain or enable everything that follows. The wrong architecture is expensive and slow to unwind.
The Carnegie Mellon AI literacy programme is the right foundation — but technical fluency alone does not drive adoption or genuine cultural change across the firm.
The firms making fastest progress are not trying to solve AI broadly. They select specific use cases decisively and build them well before moving to the next one.
Strategy
Which use cases will we treat as production priorities — with dedicated teams, clear success metrics, and board-level accountability — rather than as pilots running in parallel with everything else?
Data
Is the data underpinning our AI programmes governed, trusted, and maintained as a product — or are we building on foundations that will require costly remediation once systems are in production?
People
Are we building genuine AI capability in our people, or providing access to AI tools and calling it a programme? The difference between the two is the difference between adoption and shelfware.
Governance
How are we measuring what AI is doing in production — not just at deployment — and what is our mechanism for identifying quality degradation before it reaches clients or creates regulatory exposure?