On this page
Most leaders, when pressed on what their organization is actually made of, will name one of two things. Either "our people" or "what we know." Both answers are right and both answers are incomplete. What they are pointing at, without quite naming it, is intelligence — and intelligence is not one thing. It is two.
There is the intelligence that lives in artifacts, data, frameworks, and language: the things an organization has gathered, written, recorded, decided, and learned. Call this informational intelligence — informational because it is, in the end, structured information that can be activated for transformation later. The word "knowledge" is too narrow. Most of what matters here is not formally known by anyone; it is captured, scattered, latent.
And there is the intelligence that lives in people, ties, history, trust, and reciprocity: the donor who funded the founding gift twenty years ago and still picks up the phone, the colleague who cosigned the early book, the former staffer who now sits on a peer board, the network of practitioners who quietly trust each other. Call this relational intelligence — the human graph that carries the work.
Every organization, every leader, every movement runs on both. And in almost every case, both are fragmented.
Fragmentation is not a content problem or a CRM problem. It is the condition in which an organization's two most valuable assets — what it knows and who it is connected to — exist as scattered fragments rather than integrated systems. The cost is rarely visible day to day. It compounds. It blocks formation. It makes multiplication impossible. And, increasingly, it makes AI useless.
This article does three things. First, it breaks intelligence fragmentation into its two halves and walks through what each one actually contains. Second, it traces the same four-stage path — integration, activation, formation, multiplication — through both. Third, it shows why AI becomes wise and useful only after this sequence has been followed, and why so many AI deployments are quietly failing for exactly this reason.
Part 1: The Two Halves of Intelligence
Informational Intelligence
If you opened up an organization or a leader's body of work and laid every piece of structured information on the floor, you would find at least ten distinct categories. Most organizations have all ten. Almost none have them integrated.
1. Documented knowledge. Books, articles, white papers, research reports, curricula, training manuals, published frameworks. The formal output. The thing that gets cited.
2. Tacit knowledge. Frameworks-in-progress, mental models, instincts, heuristics, the "how we think about this" that lives in senior heads and shows up in conversation but has never been written down. The most important category and the most consistently lost.
3. Operational data. CRM records, financial history, program metrics, attendance, donations, deliveries, transactions. The structured numerical record of what actually happened.
4. Communications archive. Emails, memos, meeting notes, board minutes, decision logs, internal Slack threads. The conversational record of how decisions actually got made.
5. Voice, style, and identity. The recurring rhetorical moves, the words an organization uses (and the ones it carefully does not), the way arguments are structured, the tone, the signature image patterns. The thing that makes a piece of writing recognizably theirs.
6. Decisions and precedent. Why this program was started, why that one was sunset, why we said no to that funder, why we changed methodology in 2019. Governance memory. The "why" behind the "what."
7. Stories and narrative. Case studies, testimonials, founding stories, formative incidents, the apocryphal anecdotes that everyone repeats at staff retreats. The narrative spine that carries the mission.
8. Media artifacts. Talks, podcast episodes, video, audio recordings, photography. Increasingly the dominant format for how leaders are encountered, and almost always the most fragmented category — scattered across YouTube channels, Vimeo accounts, podcast feeds, and personal hard drives.
9. Procedural knowledge. How things actually get done. SOPs, workflows, checklists, the unwritten sequence by which a major-gift solicitation, a program rollout, or a content production cycle actually works.
10. Theological, philosophical, or ethical foundations. What the organization or leader believes and why. The conceptual ground from which every other decision is made. For mission-driven organizations this is load-bearing; in fragmented form it is also frequently re-litigated, because no one can find the canonical statement of it.
This is a generous definition of "information." That is the point. The reason most AI implementations stall, and the reason most institutional memory walks out the door with departing staff, is that organizations have been treating only category 1 — documented knowledge — as if it were the whole picture. The other nine categories carry just as much intelligence and are almost always more fragmented.
Relational Intelligence
The second half is harder to digitize, which is part of why it is so often invisible. Relational intelligence is the structured (or unstructured) human graph that surrounds an organization. It also has at least ten categories.
1. Donor and funder relationships. Active givers, past givers, lapsed givers, prospects, foundation program officers. With each: a giving history, a relationship history, a set of stated and unstated motivations.
2. Staff and internal relationships. The team itself — including the dynamics between team members, the apprenticeship lines (who is being formed by whom), the coalitions, the unspoken hierarchies.
3. Network and peer relationships. The other leaders, organizations, and movements with whom there is mutual trust, shared work, occasional collaboration. The ecosystem.
4. Endorser and influence relationships. The voices that have publicly stood behind the work — blurbs, forewords, podcast hosts, conference platforms, social amplifiers. Reputation capital lives here.
5. Client, beneficiary, audience, or congregational relationships. The people the organization actually serves: students, congregants, readers, program participants, clients. The relationship with this group is usually the largest by count and the most weakly tracked.
6. Vendor and partner relationships. Production partners, technology providers, consultants, contractors, agencies. The operational supply chain.
7. Board and governance relationships. The fiduciary leadership. The history of board service, the tenure patterns, the personalities and dynamics that shape every strategic decision.
8. Alumni and former-staff relationships. Former employees, former board members, former cohort participants. The diaspora. Often the highest-leverage and least-managed category.
9. Generational and inheritance relationships. The emerging leaders being prepared to carry the work forward. The mentorship lines. The succession question.
10. Public and audience relationships. Readers, social followers, podcast listeners, conference attendees. The diffuse outer ring of relationship that is real even though it is mostly one-directional.
Underneath all ten of these categories sits a deeper layer of relational intelligence that is rarely captured anywhere: the history of interactions (what we have actually said and done with each other), the trust capital (what we can ask of each other and what we cannot), the mutual obligations (what each of us owes the other), the cultural and contextual fit (how this person needs to be approached, what they care about, what offends them), and the second-order graph (who knows whom, who introduced whom, who could open which door).
This is the intelligence that allows an experienced leader to walk into a room and immediately know who needs to be thanked, who needs to be asked, who needs to be deferred to, and who needs to be quietly avoided. It is enormous. It is real. And in most organizations, it lives entirely in the heads of three to five senior people. When they leave — or when they simply have a busy quarter — the organization functionally loses access to it.
Part 2: What Fragmentation Looks Like in Both Spheres
Informational fragmentation
The signs are easy to spot once you know what you are looking for.
- A book corpus exists across thirteen titles, but no one — including the author — can quickly find every passage where a given concept appears. The search tool is the author's memory.
- Talks live on five different platforms (YouTube, Vimeo, Apple Podcasts, a defunct conference site, the organization's own page) and are not cross-indexed. Most are not transcribed. None are tagged.
- A core framework exists in three different versions — one in a 2008 book, one in a 2017 book, and one in current speaking decks — and the differences between them have never been reconciled.
- Operational data sits in the CRM. Stories sit in a comms folder. Program data sits in a separate database. They do not talk to each other.
- A new staff member spends six weeks figuring out things that the most senior staff already know but have never written down.
- A board memo references a decision the board made in 2019 — and no one can find the original rationale, so the conversation reopens from scratch.
- A grant report is written from memory because the relevant program data is locked inside a tool that two people can use and one of them is on sabbatical.
- The organization's voice on the website does not match the voice in the donor newsletter does not match the voice in the founder's books. Each lives downstream of a different writer.
In aggregate: the organization knows a great deal, but it cannot reliably reach what it knows.
Relational fragmentation
The signs of this are subtler and tend to be socially papered over, but they are no less real.
- Donor history lives across the CRM, the development director's email inbox, the founder's phone, and three sticky notes.
- The connections graph — who knows whom — exists only as the implicit knowledge of two long-tenured staff. When they are unavailable, warm introductions become cold ones.
- A major prospect is approached without anyone realizing she sat on a board with the chair fifteen years ago. The opportunity for warm framing is missed.
- An alumnus rises to influence in a peer organization, and no one in the originating institution learns of it for three years.
- A foundation program officer changes seats and the relationship resets, because the relationship was carried by the prior officer's individual file rather than by an institutional record.
- An emerging leader has been informally mentored for seven years by the founder, but there is no shared inheritance map — no one else on the team knows what has been transmitted, what has been delegated, or what will need to be rebuilt if the founder steps back.
- Endorsers from a previous era exist on the back covers of books but have never been re-engaged. Their relational capital decays unused.
- The same donor receives three uncoordinated communications from three departments in the same week, each unaware of the others.
In aggregate: the organization is connected to a great many people, but it cannot reliably reach its own relationships.
What unites both forms of fragmentation is the same underlying fact. The intelligence exists. The integration does not. The cost of this gap is invisible until you try to do something that requires the whole picture at once — write a definitive book, prepare for a major gift conversation, onboard a successor, deploy AI — and then the cost becomes total.
Part 3: The Sequence — Integration, Activation, Formation, Multiplication
There is a single sequence that resolves fragmentation, and it is the same sequence whether the fragmentation is informational or relational. The four stages are integration, activation, formation, and multiplication. Each stage is necessary. None can be skipped. And each one looks meaningfully different in the two spheres, which is why both halves have to be worked simultaneously.
Stage 1: Integration
On the informational side, integration is the disciplined work of gathering everything across the ten categories and structuring it. Books, talks, frameworks, decisions, stories, voice, operational data, procedural knowledge — all of it pulled into a coherent layer where relationships between concepts are explicit, terminology is unified, canonical versions are designated, and a clear distinction is drawn between what is settled and what is still in progress. This is rarely visible work. It produces no new content. It is also the most load-bearing investment a fragmented organization can make.
On the relational side, integration means assembling the human graph. The CRM is reconciled with the inboxes, the personal contacts, the board files, the alumni records, the speaker bureaus, the historical files. Each relationship gains an actual record — interaction history, known motivations, mutual connections, prior commitments, current state. The implicit knowledge living in three senior heads becomes legible to the organization as a whole. The graph becomes queryable: who do we know at this foundation, who is connected to this prospect, which alumni are now in positions of influence, who should be at this gathering.
Integration is the stage where most organizations stall. It is unglamorous, it takes months, and it does not produce a deliverable that looks like progress. But every subsequent stage depends on it.
Stage 2: Activation
On the informational side, activation is the moment integrated intelligence becomes usable. Search works across the entire corpus. Frameworks become living artifacts that can be referenced, queried, and adapted. Internal staff can pull up the rationale behind a 2019 decision in seconds. The organization's voice exists as an actual style guide that can be applied to new writing. AI tools, grounded in this corpus, become reliable rather than improvisational.
On the relational side, activation is the moment the human graph becomes actionable. A development officer preparing for a meeting can see the donor's full history, the warm connections through the network, the prior interactions, the open commitments, the relevant story matches — in one place. A leader thinking about who should be invited to a private gathering can see the second-order graph and identify connectors who would otherwise be invisible. An emerging leader can see who in the network is positioned to mentor, sponsor, or open doors for them. The relationships do not change; the reachability of those relationships changes radically.
Activation is the stage where the integration work begins to pay back. The organization can finally do, in minutes, things that previously required hours of asking around.
Stage 3: Formation
This is the stage where most discussions of "AI strategy" and "knowledge management" go silent, because formation cannot be reduced to a tooling problem. Formation is what the integrated and activated system does to people.
On the informational side, formation is the moment when readers become practitioners and practitioners become leaders. A coherent corpus produces coherent pathways: structured curricula, sequential courses, deliberate progressions. The system stops merely informing and begins shaping. A reader who picks up the work today can be guided into deeper engagement, into practice, into adaptation in their own context, into eventual leadership of others — all within a system that maintains theological and methodological fidelity at every step. Without integration and activation this is impossible. With them it becomes nearly inevitable.
On the relational side, formation is the moment when transactional contacts become formed partners. Donors are not merely solicited; they are walked into the mission, given access to its substance, formed by its theology and practice until they begin to think of themselves as participants rather than funders. Staff are not merely managed; they are formed into operators who carry the organization's logic in their bones. Apprentices are not merely mentored ad hoc; they are shaped through a structured inheritance pathway. Network peers are not merely contacts; they are co-laborers whose work strengthens the whole because there is a shared frame.
The formation stage is where the moral seriousness of integration becomes visible. Without it, even an integrated organization is just an efficient one. With it, the organization actually changes the people who encounter it.
Stage 4: Multiplication
On the informational side, multiplication is the moment the work begins to reproduce itself beyond the originating leader and team. The frameworks travel. The curricula are translated. The corpus is grounded enough that AI can extend the leader's voice into contexts the leader has never visited. Practitioners adapt the work for their own settings without distorting its spine. The intellectual lineage becomes transmissible. The leader's calendar stops being the bottleneck.
On the relational side, multiplication is the moment the network begins to reproduce itself. Apprentices become senior leaders who in turn raise up apprentices. Donors become advocates who recruit other donors. Endorsers become co-laborers who extend the work into their own circles. Alumni become a living network rather than a list. The organization stops being the center of a hub-and-spoke and becomes the seed of a self-replicating mesh.
In both spheres, multiplication is not a marketing stage. It is the structural consequence of having done the prior three stages with integrity. Skip any of them and multiplication either does not happen or happens in distorted form.
Part 4: Why AI Becomes Wise Only at the End of This Sequence
There is a reason this article saves AI for last. Not because AI is unimportant — the opposite. AI is the most consequential activation layer most organizations have ever encountered. But it is exactly that: an activation layer. It cannot do for an organization what the organization has not yet done for itself.
A general-purpose AI tool, dropped into a fragmented organization, will produce three predictable failure modes.
On the informational side, it will hallucinate. With no grounded corpus to retrieve from, the model invents. It produces plausible-sounding answers that are subtly wrong about the leader's positions, the organization's history, and the framework's actual content. The more confident the output looks, the more dangerous the failure becomes — because the people downstream of it (readers, students, staff, donors) cannot tell which sentences are accurate and which are confabulated. Fragmentation does not just reduce AI's value; it actively weaponizes AI against the organization's integrity.
On the relational side, it will produce generic platitudes. Without a real graph of donors, alumni, network ties, and history, the model can only fall back on generic best-practice advice. It cannot prepare a meeting brief that recognizes a fifteen-year-old board overlap. It cannot surface a warm path through three intermediaries. It cannot flag the awkward history with a particular family foundation. It cannot tell a development officer that the prospect's brother went to graduate school with the program director. The relational fabric of the organization is exactly what the model cannot see.
In both spheres, the AI will fail to learn. The MIT Project NANDA report on enterprise AI deployments, published in mid-2025, found that 95% of generative AI pilots produced no measurable business return. The most common cited reason was that the tools "do not retain feedback, adapt to context, or improve over time." That diagnosis is precisely what fragmentation produces. A model with no integrated context cannot adapt to context, because there is no context to adapt to.
When the sequence has been followed, all three failure modes invert.
A model grounded in an integrated informational corpus retrieves before it generates. It cites. Its outputs are checkable. Its voice matches the leader's voice because the actual voice is in the system. Its frameworks are accurate because the canonical frameworks are in the system. The output is genuinely usable — for drafting, for translating, for teaching, for extending the leader's reach into contexts the leader will never personally encounter.
A model grounded in an integrated relational graph can see across the organization. It can answer questions no individual staff member could answer alone. It can compose a major-gift brief that carries the institution behind every line. It can flag at-risk relationships that no single dashboard would catch. It can suggest the second-order connector who would warm the prospect that nobody has yet figured out how to approach. The relational intelligence of the organization, previously locked in three senior heads, becomes available to the whole team.
And, critically, the model can learn. With an integrated context, every interaction, every correction, every new piece of content, every new relationship update feeds back into the system. The intelligence grows. The model adapts. The organization compounds.
This is what wise AI use actually looks like. Not the naked prompt that 90% of workers are quietly running on their personal devices. Not the corporate chatbot that produces generic output and quietly gets ignored. Wise AI use is what becomes possible when an organization has integrated its informational and relational intelligence, activated it through real systems, used it to form people over time, and is now ready to multiply the work.
Conclusion
Intelligence fragmentation is the most common, most expensive, and most invisible problem in mission-driven work. It is not one problem; it is two. The informational half — what the organization knows, has produced, has decided, has recorded, and has lived — is almost always scattered across categories that have never been integrated. The relational half — who the organization is connected to, what the history is, what the trust capital looks like, what the second-order graph reveals — is almost always locked in the heads of a few senior people.
Both halves can be integrated. Both can be activated. Both can be used to form people rather than merely inform them. Both can multiply when the prior three stages are honored. And only when both halves have been worked through this sequence does AI move from a novelty (or, more often, a liability) into the role it is actually capable of playing: a faithful, queryable, scalable extension of a coherent organization.
The organizations and leaders who skip the sequence and try to deploy AI on top of fragmented intelligence are not getting AI. They are getting an expensive amplifier of their existing disorder.
The organizations and leaders who do the integration work first — patiently, across both halves — get something else entirely. They get a body of work that compounds, a network that reproduces, a mission that scales without distortion, and an AI layer that finally makes the institution available to itself.
Intelligence is not one thing. It is two. And both have to be integrated before either can be activated, formed, or multiplied. Everything downstream of that work — including the AI question that everyone is currently asking — depends on it.