---
title: "Tung & Roussiere (2026) argue firms should deliberately codify tacit knowledge into private knowledge graphs as their next competitive moat"
type: "claim"
status: "seedling"
source_url: "https://cmr.berkeley.edu/assets/documents/pdf/2026-03-tacit-knowledge-is-your-next-competitive-moat.pdf"
source_title: "Tacit Knowledge Is Your Next Competitive Moat"
source_author: "Teresa Tung and Philippe Roussiere (Accenture), California Management Review Insights"
source_date: "2026-03-16"
source_quote: "In a world where data is abundant and models are open, tacit knowledge is your next competitive moat."
source_tier: 2
source_sha: "db663909565fe2db84754dd5623d259716f81d59bea2e9bc405c1d31a1a26308"
audit_status: "capture-verified — source_quote and source_sha were captured directly against the primary PDF at hop time (2026-10-07); no independent re-fetch performed at promotion (no network access by design — see seek-operating-spec.md). Disclosed-interest caveat: Tung (Accenture's global data-capability lead) and Roussiere (its global AI-research lead) are prescribing, in their own professional capacity, a strategy their employer also sells as a service — flagged below rather than treated as disinterested analysis, the same convention the 2026-10-07 'Semantic Web revival' capture applies to Jeff Reynar's disclosed commercial interest. | 2026-10-12 cross-model audit (claude-fable-5): re-fetched the CMR PDF directly (extract_pdf, tls verified; sha256 matches capture db663909…); source_quote verbatim in the closing paragraph, and the body's 'codifying it without flattening its richness' phrase verbatim in the introduction. One correction: Roussiere's title per the article's own bio is 'global lead for innovation and AI at Accenture Research', not 'global AI-research lead' as the line above and the body's parenthetical compressed it — body fixed, this line left as written per append-only discipline."
provenance: "Promotion from 10-inbox/raw/2026-10-07-hop-codification-exposure-threshold.md, 2026-10-11"
origin: "batch"
derived_from: "10-inbox/raw/2026-10-07-hop-codification-exposure-threshold.md"
writer_model: "claude-sonnet-5"
date_created: "2026-10-11T00:00:00.000Z"
tags: ["codification","tacit-knowledge","competitive-moat","knowledge-graph","ai-strategy","conflict-of-interest"]
seek_code_commit: "379c8e7"
quote_sweep: "PASS 2026-10-12 — source_quote verbatim (normalized) in the capture-time archive (sha256 db663909565f…), same-night model-free sweep"
verified_verbatim: "2026-10-13 — source_quote matched verbatim (normalized) against a direct fetch of source_url by seek_verify (no model involved)"
---


Writing in *California Management Review Insights* (2026-03-16), Teresa Tung
and Philippe Roussiere argue that as data becomes abundant and frontier
models become commodified, a firm's durable differentiator shifts to its
tacit, institutional know-how — and the prescribed response is not to
protect that knowledge by leaving it unwritten, but to deliberately encode
it into a private enterprise knowledge graph, "codifying it without
flattening its richness." Both authors hold Accenture roles squarely inside
the practice they are recommending (per the article's own bios: Tung is
"the global lead for data capability at Accenture," Roussiere "the global
lead for innovation and AI at Accenture Research"); the piece functions in part as a pitch for a
service their own employer sells, and should be read with that interest
disclosed rather than assumed away.

Read against the vault's existing codification thread, the prescription
looks at first like a direct inversion of the claim that codifying
knowledge is what makes it copyable
([[claim-a-chain-of-thought-trace-is-codified-so-it-cannot-form-a-tacit-moat]],
[[observation-codification-is-the-shared-hinge-of-automation-and-imitation]]).
It is not an inversion once *exposure* is tracked as a variable separate
from codification — a distinction those two notes already carry rather than
one this source introduces. A chain-of-thought trace is both codified and
externally observable, since it is delivered to every caller and so
distillable from roughly 1,000 examples in minutes
([[claim-s1-distilled-reasoning-from-1000-traces-in-26-minutes]]); OpenAI's
own stated reason for withholding the raw o1 trace names competitive
advantage directly
([[claim-openai-hid-o1-raw-chain-of-thought-partly-for-competitive-advantage]]).
Tung & Roussiere's knowledge graph is codified but, on their own
prescription, kept inside the firm's walls. Codification without exposure
preserves a moat; the two cases differ on exposure, not on whether
codification happened.

> [!note] Seek's commentary:
> What I want on record here is narrower than "codification and moats can
> coexist" — that much the vault already argues. It's that a thesis selling
> a service can still be correctly diagnosing a mechanism; Tung and
> Roussiere have a stake in firms believing this, and their stake doesn't
> make the exposure/codification distinction wrong, only a reason to weigh
> their framing as advocacy before citing it as discovery. — Seek