Get in touch
Chris Dunmire
Partner · Content and Marketing Enablement SME (NA)
Asset management · Content operating model
Asset managers build finished documents by hand, channel by channel, quarter after quarter. Digital-first means authoring components instead: a commentary, a chart, a disclosure, each approved once and reused everywhere. Factsheets, pitchbooks, pages and emails become assemblies drawn from one approved source. This framework covers what that means in practice, the architecture and teams behind it, and how the platforms you already own evolve into it.
01 · What we mean by digital content
And the analog runs deeper than the PDF. The production method is print-era too: plan the calendar up front, standardize the templates, batch-produce the quarterly run, ship, repeat. Digital was supposed to mean the opposite. Content assembled at run time, for this client, in this market, the moment it is asked for.
A factsheet PDF, a pitchbook, a commentary, a web page. Each one is built largely by hand, approved as a whole, and frozen the day it ships. Everything valuable inside is trapped in the format: the product story, the performance narrative, the disclosure logic. So next quarter the whole thing gets rebuilt, and every channel rebuilds it again. That is a document operating model running on digital tools.
Digital content is content built as structured components. A commentary, a chart, a bio, a disclosure. Each one is authored once, approved once, and carries machine-readable meaning: what it is, which product it references, which markets it can enter, when it expires. Documents, pages and pitchbooks become assemblies of those components, built on demand for a specific client, channel and market.
The shift is from producing finished documents to governing the parts inside them. Update a component once and every assembly that uses it updates or flags for rebuild. That single property is where the speed, cost and compliance advantages come from.
What firms that get this right can expect
Assemblies take hours, not weeks. A product update propagates everywhere at once instead of triggering fifty document revisions.
Approval lives on the component and travels with it. Expired or superseded content cannot enter a new assembly. The audit trail exists by default.
The right commentary, share class and disclosures for each client and market, assembled automatically. Not a mail merge. A different document.
The rework tax collapses. Teams stop rebuilding the same content in four platforms and author it once. Serving a new market stops scaling linearly with headcount.
Agents can only be trusted when they draw from structured, approved, current content. This is the prerequisite for AI in a regulated firm.
02 · The architecture
No single solution does all of this. Distinct platforms sit in capability zones that mirror the lifecycle: components get authored, managed in the platform at the center, assembled into deliverables in the production zone, and handed to channel surfaces for delivery. Text and structured content moves via API at runtime. Digital assets replicate to channel libraries under sync control. Click any block to light up its connections and see its role, inputs, outputs, example tools and five-year trajectory.
Flow lines render on larger screens. Tap any block for its inputs, outputs, example tools and trajectory.
Content gets asked for in two ways. Sometimes you know the exact thing. A named asset. A specific fragment. The approved version of a disclosure. The repository hands that straight back. Other times you only know the situation. This client. This channel. This market. Nobody can name the right answer up front. It has to be worked out from relationships, entitlements and rules. That is governed retrieval's job. The repository owns the content. Governed retrieval decides who gets what. And the things firms want next all run on that second path. RFP responses. Personalization. AI agents.
Websites. Portals rendering known content. Client reporting and books production, where automation pulls named fragments and the rules engine pairs disclosures per run. These work against the DAM directly from day one, and never stop working that way.
RFP responses. Personalized marketing automation. Digital self service. Agentic content production. Seller search that beats a shared drive. None of these arrive until governed retrieval exists, which is why Horizon 2 is where the model starts paying for itself beyond production savings.
An integrated stack run by siloed teams delivers siloed outcomes at a higher license cost. The architecture pays when ownership is wired the same way the platforms are. Five teams, one supply chain. Authors answer for what the content says. Operations answers for whether it stays findable and current. Keep those two jobs separate.
Investment writers, product marketing and creative. Accountable for the accuracy and currency of what they author, by content domain: the commentary, the bios, the product copy. The author is the owner, the same way a data product owner answers for their data.
The librarians. Responsible for library health: tagging quality, expiration enforcement, taxonomy stewardship. Authors are accountable for the content. Operations keeps it findable, current and clean. Without a named owner the taxonomy decays within a few quarters.
Run the MRM, the repository and every integration as internal products with roadmaps, not projects that disband at go-live. The agent workflow tier lives here too. Someone has to own the gates, the logging and the rollout sequence.
Review and approval, and ownership of every disclosure component. The emerging capability inside this team is compliance engineering: translating regulatory requirements into deterministic, testable rules.
Web, sales enablement and client reporting teams consume from the shared source instead of maintaining private libraries. Feed usage back. Their willingness to give up local copies is a leading indicator of the whole program.
03 · Vision & principles
Every architecture decision either honors these or quietly breaks them. They are written to be argued with.
01
Author the commentary, the bio, the chart once. The factsheet, the pitchbook and the web page are assemblies of those components, not separate creations.
02
Compliance state is a property of the component, not the document. Re-approval happens when the component changes, not when a new document uses it.
03
Every component carries structured metadata. Taxonomy is the contract that lets systems and agents find, assemble and govern content without a human in the middle.
04
Where content is made has nothing to do with where it lands. Authoring experiences are purpose built for the maker's workflow. Delivery surfaces consume from the same source. Adding a channel never forces re-authoring.
05
No single platform holds every component, by design. Text and assets in the repository, disclosures in their own library, data in APIs, web components native. What makes it one source of truth is one taxonomy, one approval model, one provenance registry, and one retrieval layer spanning them all.
06
Disclosure logic, jurisdiction rules and usage restrictions live in rules engines with deterministic outcomes and full audit trails. Human review concentrates on judgment.
07
Stale content is a regulatory and commercial liability. Every component has an owner, a review trigger and an expiration date. The system enforces this. People do not have to remember.
08
An integrated stack run by siloed teams is shelfware with better plumbing. Ownership, decision rights and incentives have to be wired the same way the platforms are. The operating model is half the architecture.
09
Vendors rotate. The taxonomy, the ontology and the component model are the portable assets. Design them platform neutral and every future migration gets cheaper instead of harder.
10
Structured content, clean taxonomy, approval APIs and audit logging are the preconditions for AI agents doing real work. Every decision moves the firm toward that or away from it.
04 · The evolution path
Nobody rips and replaces. The stack evolves in three horizons, and point solutions are not decommissioned by ambition. They are descoped when specific things become real. Interfaces and workflow layers erode first. Delivery and record keeping erode last. The compliance spine never does.
Horizon 1 · Years 0–1.5
The firm still ships documents while the plumbing changes underneath. Most of the organization never notices this horizon. It is not supposed to.
Nothing is displaced yet. The work is making displacement possible.
What has to become real
Agents in this horizon
Horizon 2 · Years 1.5–3
Content starts finding people instead of people hunting for content. Agents begin doing real work behind real gates.
First descoped: the MRM interface. Agents take briefs, tag, route and report. The MRM survives as a state machine and reporting database behind them.
What has to be true first
Erosion begins: sales enablement library functions, as the semantic tier serves approved, entitled content directly. Last-mile delivery persists.
What has to be true first
Horizon 3 · Years 3–5
Documents stop being made and start being assembled. Production becomes a runtime event, not a team's quarter.
Narrowed: automated document production. Content storage and template copy migrate to the stack. The production runtime of scheduling, batch, exceptions and validation moves last, if it moves at all.
What has to be true first
Descoped: the CMS as a content repository. It returns to experience delivery. Channel-specific web components stay native permanently.
05 · Where to start
None of this requires the five-year stack. It requires decisions, one proven component, and evidence.
Name one platform as the master record for approved components and manage approval state there and nowhere else. Give content operations named ownership of the taxonomy and component model as living products. Decision rights explicit, not implied. Everything downstream depends on these two calls.
Separate legal-owned disclosure content from the general component library and document the rules requirements: jurisdiction logic, versioning, pairing rules. Do not build the engine yet. Scope it. Disclosures behave differently from everything else and pretending otherwise poisons the whole library.
Pick a component with real complexity. Attribution commentary is a strong candidate. Trace it from ideation through intake, tagging, authoring, review, storage, distribution and measurement across the actual stack. Table top first, then live. And run the sizing exercise alongside it: fragment volume projection against repository API throughput and pricing, so the platform's ceiling becomes a number the test produces rather than an argument. Where it breaks is the roadmap.
Stand up component usage reporting alongside the pilot, not after it. Which components get used, where, how often, and what sits stale. Without the evidence base the model is an act of faith and the business case never graduates from a slide.
Approval workflows exposed as APIs. Entitlement scope defined. Audit logging designed in. Build for agents before you deploy them. Firms that do this in horizon one enter horizon two with a stack agents can safely operate in. Firms that skip it get demos.
The firms that get this right will not look faster at making documents. They will stop making documents as a primary activity. Content becomes infrastructure. Production becomes assembly. And the marginal cost of serving one more client, one more channel or one more market falls to the cost of review, not the cost of rebuild.
Interactive · Ten questions, two minutes Where does your firm stand on the three-horizon path? Take the readiness check →06 · See it in practice
Two ways into the same machinery. Follow a component and watch its passport fill in as it travels the supply chain. Or follow a person and watch what their day looks like when the stack does the mechanics: what they do, where they do it, and the work that stops arriving.
07 · The readiness check
Ten questions, two minutes. Answer for how things work today, not for the roadmap. The score places you on the three-horizon path and points at the two things to fix next. Nothing leaves your browser and nobody is collecting your email.
0 of 10 answered
Where to focus next
Get in touch