Building AI Agents For The Enterprise

Anthropic (vendor-published; Reader recorded the byline as 'Claude') · pdf · 2025 source ↗ Reader 🖥️

practical · mixed — Predominantly practical and prescriptive (a three-pillar transformation framework plus a four-phase deployment playbook), with an argumentative undercurrent ('the agentic thinking divide' — organisations that treat AI as point solutions vs. those that embed it as operational substrate) and an expository scaffold that defines a vocabulary (agent, chatbot, Claude, Claude Code, Claude Cowork, MCP, plugins, skills) before prescribing against it. Like the Founder's Playbook, this is a sales artefact: every prescribed action maps to an Anthropic surface, every case study links to a claude.com customer page, and the closing CTA is a sales contact. Crucially, this PDF puts governance and security up front as design constraints rather than checklist items — a real concession to the enterprise audience.

Structure

Key terms

• Agent (Anthropic's enterprise definition) Agentic AI
  An AI system that goes beyond Q&A: 'independently plan, make decisions, and use tools across multiple steps to complete a task'. Defined explicitly in opposition to chatbot. The contrast is load-bearing because the entire document is an argument for moving from chatbots to agents, framed as the 'agentic thinking divide'.
• Enterprise (implicit)
  Never defined explicitly, but operationally: a multi-thousand-employee, multi-country, multi-business-unit organisation with a centralised IT/CISO function, existing trust and compliance infrastructure, and procurement authority. The L'Oreal (150 countries), Lyft (global), Rakuten (70+ businesses), Novo Nordisk and RBC case studies establish the implicit reference class.
• Agentic thinking divide Enterprise AI Adoption & Readiness
  Coined framing: the strategic gap between organisations that deploy AI as isolated point solutions (chatbots, summarisers, single-team pilots) and those that embrace agentic AI as core operational infrastructure. The divide compounds: the latter pulls further ahead over time.
• Three pillars: employees, processes, products Enterprise AI Adoption & Readiness
  The taxonomy the document uses to organise enterprise AI transformation. Upskilling employees (Pillar 1) targets time-saved and adoption; accelerating processes (Pillar 2) targets cycle-time compression and quality; transforming products (Pillar 3) targets revenue and competitive moat.
• Plugins
  'Packages of skills, context, and connectors that give Claude role-specific expertise'. Built once, shared across the organisation; turn tribal knowledge into organisational infrastructure. Anthropic has open-sourced 11 (productivity, sales, finance, data, legal, marketing, customer support, product management, enterprise search, biology research, plugin management itself).
• Skills Agent Skills
  'Reusable, predefined workflows that teach Claude how to perform specific tasks consistently'. The unit of codified expertise inside a plugin.
• MCP (Model Context Protocol) MCP (Model Context Protocol)
  'A universal standard that allows Claude to connect securely to external services'. Framed here as a protocol Claude uses to talk to CRMs, databases, internal platforms.
• Claude Cowork Non-Engineers Building
  'Anthropic's collaborative AI interface that brings the capabilities of Claude Code to everyone' — i.e., a non-developer-targeted desktop/knowledge-work application that runs tasks locally, integrates with Excel/PowerPoint/etc, and supports the plugin/marketplace model.
• Claude Managed Agents Long-Running & Managed Agents
  'Prebuilt, configurable agent harness available via the Claude Platform' — the execution layer (compute, memory, storage, sandbox) for agents, offered as managed infrastructure so engineering teams don't build it themselves. Cited as what unlocked Rakuten's bi-weekly release cadence.
• Trust boundary
  'In any industry where customers trust you with sensitive data, the trust boundary is the first design constraint for any AI-powered product.' Operationally: data residency, access controls, audit trails, regulatory perimeters. Whether an AI product can ship is determined by whether it operates inside the existing trust architecture.
• Plugin marketplace (admin-controlled) Agent Governance (Internal)
  Internal Anthropic-provided marketplace where admins decide which plugins are available, pre-installed, or require approval. Framed as the cure for 'shadow AI sprawl' and 'uncontrolled proliferation'.
• Compounding returns / compounding AI advantage
  Used in two distinct senses: (a) within a process, expert reviews feed back into the AI's knowledge base, so every approved output makes future outputs better; (b) across the organisation, every new team's rollout is cheaper because of accumulated context, plugins, and governance from earlier waves. Both senses imply a first-mover advantage.
• Local execution (Claude Cowork) Local AI
  'Claude processes your files, your data, and your documents on the employee's machine. Nothing is uploaded to a cloud service for processing.' Positioned as removing the most common security objection.

Propositions

• explicit40% of US employees report using AI at work as of September 2025, up from 20% in 2023.conf: high
• explicitOrganisations that treat AI as a point solution get point-solution results; transformation requires simultaneously rethinking employees, processes, and products.conf: high
• explicitA general-purpose AI system that knows nothing about your organisation produces general-purpose output that requires costly human edit before it's useful.conf: high
• explicitcontestedEncoded organisational knowledge (plugins, skills) is the binding scarce input: two companies using the same model get dramatically different results based on context provided.
  extendsMeta Deploys Unified AI Agents to Automate Performance Optimization at Hyperscale · p6 — Generalizes Meta's encode-expertise-as-skills pattern into a cross-enterprise claim: encoded organisational knowledge is the binding scarce input that differentiates outcomes on the same model.
  refinesAgents Over Bubbles · p5 — Relocates the point of differentiation from model-harness integration to the encoded organisational knowledge fed into the harness.
  contradictsThis Week in AI: Production Viability · p11 — Urges encoding organisational knowledge into the model context as the binding scarce input, while the older piece warns that piping institutional memory into foundation models erodes organizational IP.
• explicitSelf-educating, compounding AI systems give the earliest movers the largest cumulative advantage; every month of expert approvals makes the next month's output better.
  supportsAgents Over Bubbles · p10 — Compounding self-educating AI systems give earliest movers a cumulative advantage, reinforcing the structural disadvantage of non-adopters.
• explicitL'Oreal's Claude-based multi-agent platform serves 44,000 monthly users / 2.5M messages/month / 15,000 DAU and improved conversational analytics accuracy from 90% to 99.9%.
• explicitLyft's Claude-powered support reduced resolution time by 87% and improved decision-making accuracy by 30%.
• explicitRakuten using Claude Managed Agents now ships major releases every two weeks (was: quarterly), reduced initial critical errors by 97% in pilot, and reduced agent cost/latency by >30%.
• explicitdevelopedThe trust boundary (data residency, audit, regulatory perimeter) determines whether an AI product can ship; AI products that require client data to leave the firm's perimeter trigger compliance reviews that often kill them.conf: high
  extends ←Podcast: Will Agentic AI Bring Fantasia's Sorcerer's Apprentice to Life?: A Conversation with Tracy Bannon · p1 — p9 frames the trust boundary narrowly as a regulatory/data-residency perimeter that determines shippability; Bannon extends this to a general claim that crossing between ANY distinct software ecosystems -- not only regulated data leaving a firm -- sharply raises agentic risk.
• explicitcorroboratedDefensible AI products combine a frontier model with proprietary data, existing trust relationships, and deep domain expertise; the model is the enabler, the defensibility lives in everything around it.
  supports ←Building an Autonomous Enterprise for Real-World Services with Netic Founder Melisa Tokmak · p7 — Near-identical framing: defensibility lives in everything around the model (data/trust/domain expertise there; harness/orchestration/product here), with the model as enabler rather than moat.
• explicitCustom-built agent platforms are the high-friction path; Claude Cowork (plugins + local execution + admin controls) is the low-friction path that delivers comparable outcomes without engineering investment.
• explicitAdmin-controlled plugin marketplaces are the cure for 'shadow AI sprawl' — uncontrolled team-level adoption of disconnected AI tools.
  extendsIncident 1368: Malicious OpenClaw Skills Reportedly Delivered AMOS Stealer and Exfiltrated Credentials via ClawHub · p5 — Moves from the incident's diagnosis (open skill ecosystems lack vetting) to a governance remedy: admin-controlled plugin marketplaces for skill distribution.
• explicitcontestedClaude Cowork runs tasks locally on the employee's machine — 'nothing is uploaded to a cloud service for processing'.conf: medium
  contradictsGoogle Brings MCP Support to Colab, Enabling Cloud Execution for AI Agents · p2 — Opposing security prescriptions: Colab argues delegating execution to managed cloud solves local agents' security constraints; Anthropic pitches local-only task execution as the enterprise trust guarantee.
• explicitdevelopedAuditability is available via OpenTelemetry-compatible observability tools, providing the evidence chain regulators may ask for.
  extends ←The Meter Was Always Running · p7 — That piece asserts auditability is 'available via OpenTelemetry-compatible observability tools' in general terms; Haelen supplies the concrete schema and granularity (turn-level signal categories, example fields) that claim needs to be actionable rather than aspirational.
• explicitcorroboratedStart with specificity, not scale: give Claude organisational context from the beginning; generic first interactions kill adoption.conf: high
  supports ←Everyone Is Talking About Context for AI. Here's What Most Companies Still Miss. · p3 — Both claim enterprise-specific/organizational context (not generic AI) is the precondition for AI to be useful/trustworthy — this doc frames it as 'operational context', the other as giving Claude organizational context from the start to avoid generic first interactions.
• explicitPilots must have measurable finish lines set upfront; vague goals ('improve productivity') produce dismissable results.conf: high
• explicitBuild plugins for reuse from day one, not as one-team solutions retrofitted later.conf: high
• explicitcorroboratedGovernance is a prerequisite, not a feature to add later; organisations that skip it spend more on cleanup than they saved on speed.conf: high
  supports ←AI Model Context Protocol Adds Centralised Auth for Enterprise · p3 — EMA operationalizes 'governance is a prerequisite, not a feature to add later' by shipping enterprise-managed, admin-defined authorization as the access-control layer before scale.
• implicitdevelopedAnthropic's three product surfaces (Claude/Claude Code/Claude Cowork) plus Claude Managed Agents plus open-sourced plugins are the natural and right substrate for enterprise AI transformation.conf: high
  extends ←How building software is changing at Anthropic · p1 — Supplies the origin story, internal build cost, and ~6-month timeline behind the Claude Managed Agents product that the other extraction credits as a strategically central part of Anthropic's enterprise substrate.
• explicitcorroboratedWaiting for a comprehensive strategy is the most common transformation mistake; the right move is to start narrow and learn fast.conf: high
  supports ←Zero to Agent in 30 Minutes: Build a Hermes Social Media Agent with Craig Hewitt · p2 — Both independently converge on 'start narrow, prove it, then generalize' — the enterprise-transformation principle applied here at the scale of a single agent workflow rather than an organizational rollout.
• explicitTeams moving to centralise AI within operations now (vs treating it as an add-on layer) have a compounding head start that later movers cannot easily close.

Arguments

Narratives

• L'Oreal: 150-country skincare/haircare/makeup/fragrance company built a Claude-based multi-agent internal platform with 15+ specialised agents; 44,000 MAU, 2.5M messages/month, 15,000 DAU; conversational analytics accuracy improved from 90% to 99.9%.
  → Anchor case study for Pillar 1 (Upskilling employees). Demonstrates scale + measurable outcome. Implicitly supports the 'plugins/skills as institutional infrastructure' argument because L'Oreal's platform is a custom analogue to what Anthropic now offers off-the-shelf.
• Lyft: global mobility platform facing 30-40 minute support wait times rolled out Claude support assistant starting with driver onboarding (region-specific compliance), expanded to riders. Resolution time dropped 87%; decision accuracy improved >30%; reinvested savings into agents and the Lyft Silver program for older riders.
  → Pillar 2 anchor (Accelerating processes). Demonstrates compounding accuracy via expert feedback and role redefinition (agents go from volume-juggling to handling complex cases requiring empathy). Lands sympathetically (older-rider program).
• Rakuten: 70+-business conglomerate built agent infrastructure from scratch as part of 'AI-nization' strategy; switched to Claude Managed Agents, freeing engineering talent. Now ships major releases every 2 weeks (was: quarterly), 97% reduction in initial critical errors in pilot, >30% cost/latency reduction. Power-user 'Galileo' tier of employees builds across domains beyond their primary role.
  → Pillar 3 anchor (Transforming products). Demonstrates the build-vs-buy tradeoff Anthropic wants you to make: Rakuten chose to offload the execution layer to Anthropic so they could focus on differentiated agentic experiences.
• Novo Nordisk built 'NovoScribe', reducing clinical study documentation from 10+ weeks to 10 minutes while maintaining regulatory quality (mentioned in passing, not full case study).
  → Existence proof that AI works in heavily-regulated life sciences. Single-line use; not elaborated.
• RBC delivered an agentic solution serving 2,200 advisors managing $689 billion in assets, giving advisors more time with each customer (mentioned in passing).
  → Existence proof for financial services. Single-line use; not elaborated.

Questions addressed

• ✅ Why do most enterprise AI initiatives plateau?
  Because they treat AI as a point solution rather than as a substrate for cross-pillar transformation (employees, processes, products).
• ✅ What's the difference between AI adoption and AI transformation?
  Adoption puts a tool in front of employees; transformation changes the baseline of what every employee can accomplish via encoded organisational context.
• ✅ How should an enterprise structure its first 6 months of AI rollout?
  Phase 1 (eval and success criteria, 2-3 teams with clear workflows, specific metrics); Phase 2 (champion pilot, 2 months, production workflows, qualitative + quantitative feedback); Phase 3 (scale with admin marketplace and approval workflows).
• ✅ Should we build a custom agent platform or buy?
  The implicit answer is 'buy Claude Cowork / Claude Managed Agents'. Rakuten case study is the explicit version: they built, then switched to managed. The document does not seriously engage the case for building (vendor concentration risk, capability lock-in, future-proofing) — only flags it as engineering opportunity cost.
• ✅ How do we prevent shadow AI sprawl?
  Admin-controlled plugin marketplace with approval workflows, role-based access, spend controls, usage analytics.
• ✅ How do we make this work in regulated industries?
  Trust-boundary-first design; local execution; OpenTelemetry observability; admin controls. The framing is correct, but the detail beneath it (data residency for model inference, sovereign-cloud requirements, sector-specific obligations) is not engaged.
• ✅ What metrics signal AI transformation is working?
  Pillar-specific: employees (time saved, adoption rates); processes (cycle time compression, quality scores); products (revenue impact, speed to market, customer capability shift).
• ⭕ What happens if Anthropic stops being the right vendor (capability regress, price change, strategy shift)?
  Never addressed. The entire prescription presumes Anthropic surfaces (Claude, Claude Code, Claude Cowork, Claude Managed Agents, plugin marketplace, MCP-via-Anthropic). MCP is mentioned but not framed as portability infrastructure.
• ⭕ What does 'local execution' precisely mean — does the model inference also happen on-premise?
  Almost certainly NOT: the local execution refers to file/data handling and tool use, while the model itself runs on Anthropic's cloud. The document does not clarify this critical distinction, which is the one Norwegian/EU regulators and CISOs will ask first.
• ⭕ What about EU AI Act, GDPR, NIS2, DORA, Norwegian public-sector procurement, sovereign cloud, data residency in the EEA?
  Not addressed at all. 'Regulated industries' in the document means US financial services and life sciences. EU-specific obligations, Nordic public-sector procurement constraints, and AI Act compliance are entirely absent.
• ⭕ What about multi-vendor strategy / portability?
  Never raised. MCP is positioned as Claude's connector protocol, not as a cross-vendor abstraction layer (which it actually is). Plugin marketplace is single-vendor. The structural answer is 'no'.
• ⭕ What's the cost structure at full enterprise scale?
  Token economics, per-user cost, model-price exposure, COGS sensitivity are not discussed. Spend controls are mentioned as a marketplace feature but no scenarios are given.
• ⭕ What about enterprises that built this way and failed?
  No failure stories. The case studies are uniformly positive (L'Oreal, Lyft, Rakuten, Novo Nordisk, RBC). Stranded plugin investments, transformation-program failure, rollback case studies are absent.
• ⭕ How do plugin politics work — whose process becomes the canonical plugin?
  Not addressed. 'The best practices of one team become the default for every team' is asserted as benign; the political economy of standardisation is not.
• ⭕ What does this look like for non-multinational enterprises (mid-market, public sector, single-country regulated incumbents)?
  All five named customers (L'Oreal, Lyft, Rakuten, Novo Nordisk, RBC) are global multinationals. The implicit reference class is not the typical Norwegian regulated incumbent.
• ✅ Where does the agent harness sit relative to the application — and is Claude Managed Agents the right harness for me?
  Rakuten case study makes the build-vs-buy argument for the harness layer specifically, and lands on 'buy ours'. The deeper question — what makes a good harness, what are the failure modes of relying on a managed one — is not engaged. The Anthropic harness-engineering paper (separately extracted) goes much deeper on this.

Gaps

• incompleteVendor concentration risk is structurally absent. 'Enterprise AI transformation' as defined here means 'Anthropic-substrate transformation': every pillar runs on Anthropic surfaces (Claude, Claude Code, Claude Cowork, Claude Managed Agents), every plugin lives in an Anthropic-defined marketplace, and the 11 open-source plugins are Anthropic-authored. The compounding lead the document urges is also compounding lock-in. This is exactly the workflow-lock-in pattern the Founder's Playbook prescribes against competitors — applied here to the customer.
  Read next: Cloudflare 'Mythos' retro (already extracted) on dependence dynamics; MCP as cross-vendor portability layer (read the actual MCP spec, not Anthropic's framing of it); any 'multi-model enterprise architecture' practitioner writing.
• incomplete'Local execution' is doing heavy work in the security pitch but is never defined precisely. For a CISO or DPO evaluating Claude Cowork in EEA/Norway, the deciding question is whether the model inference (not just file handling) happens on-premise or in Anthropic's cloud — and what data crosses the boundary in API calls. The document explicitly says 'Nothing is uploaded to a cloud service for processing', which is misleading if the prompt/context still goes to api.anthropic.com (almost certainly the case). This is the single most important precision the document elides.
  Read next: Anthropic's own technical documentation on Claude Cowork architecture; Norwegian Datatilsynet guidance on US cloud + EEA data; EU AI Act provider obligations; any sovereign-cloud / partner-cloud Anthropic offerings (Amazon Bedrock, Vertex AI deployments).
• incompleteEU / Norwegian / regulated-incumbent context is entirely absent. The document name-checks 'regulated industries' (financial services, healthcare) but operationalises them as US contexts. EU AI Act obligations on providers and deployers, GDPR data-residency, NIS2 cybersecurity, DORA operational resilience, Norwegian sector-specific guidance (NSM, DigDir), public-sector procurement constraints, sovereign-cloud requirements — none appear. For the reader (Norwegian senior engineering leader, regulated context relevant), this is a substantial blind spot.
  Read next: EU AI Act practitioner guides; NSM's AI guidance; DigDir's references on third-party AI; any Norwegian public-sector AI procurement writing (NAV, Skatteetaten, Helsedirektoratet have all published); Healey's AI-advisory-board piece (already extracted) on incumbent AI governance.
• incompleteMid-market and non-multinational enterprises are out of frame. All five named customers are global multinationals (L'Oreal 150 countries, Lyft global, Rakuten 70+ businesses, RBC, Novo Nordisk). A 1,000-3,000 person Norwegian/Nordic incumbent — closer to the reader's likely organisational shape — would face different unit economics, different governance ratios, different plugin-investment payback periods. The document's prescriptions are calibrated for the upper end of 'enterprise'.
  Read next: Mid-market AI adoption case studies; Nordic-context enterprise AI writing; Entur AI adoption piece; NAV fast-flow piece.
• incompleteThe cost structure at scale is not discussed. The document celebrates compounding returns but is silent on token economics: model price exposure, per-user spend distributions, plugin-driven multi-step agents that 10x token consumption per task, what 'spend controls' actually do when a single team's plugin runs amok. Margin scenarios, COGS exposure to model price changes, and elasticity of demand are absent.
  Read next: Public AI-startup tear-downs and inference-cost analyses; enterprise pricing teardowns; any first-hand writing on 'we deployed Cowork to 10,000 employees and the bill was...'.
  closed byTravel Through the Lens of AI with Booking.com CEO Glenn Fogel — Fogel brings the omitted cost dimension g5 flags—per-contact cost (down ~10%), model-price exposure and 'which model for which purpose and when', token consumption, and ROI/LTV—though by his own words the ROI/LTV itself remains unresolved ('things we don't know yet'), so he surfaces rather than settles the economics.
• incompleteThe build-vs-buy reasoning is one-sided. The case studies (L'Oreal built, Lyft built, Rakuten built then switched) implicitly argue for build, while the prescriptive framing argues for buy. The document does not seriously confront when building remains the right choice (sovereign cloud requirements, sector-specific regulatory chains, vendor concentration risk, unique workflow architectures, or simply that bespoke is what the case studies actually did to achieve the cited numbers). This is the most important business decision and it gets implicit-only treatment.
  Read next: Anthropic harness-engineering paper (already extracted) for the technical 'what is a harness' substrate; Bockeler harness-engineering talk (already extracted); Heis ep 11 AI-special on insourcing logic; small-code / Cloudflare-Mythos contrarian harness aesthetic.
• incompleteFailure cases and stranded investments are absent. Every named case study is positive. Enterprises that built pillar-level plugin investments and then watched the underlying model change behaviour, or hit token-cost walls at scale, or saw a competitor leapfrog with a different stack, do not appear. 'Compounding lead' is presented without a 'compounding sunk cost' counter-scenario.
  Read next: Post-mortems of enterprise AI programs; 'killed by GPT-5' / vendor-strategy-change narratives; gartner-style 'why N% of AI initiatives fail' research.
• incompleteThe plugin-standardisation political economy is not discussed. 'The best practices of one team become the default for every team' sounds frictionless but ignores the contested-knowledge problem: whose sales process becomes the plugin, who reviews plugin updates, what happens when business units disagree, how plugin versions are deprecated. Cutler-style organisational reality is absent. The document's frame is 'institutional knowledge becomes infrastructure'; the real question is 'whose institutional knowledge'.
  Read next: Cutler on overload and institutional dynamics; Heis ep 11 AI-special on enterprise AI politics; anything on internal developer-platform governance (which is structurally similar — plugin marketplaces and IDP marketplaces share dynamics).
• incompleteThe MCP framing is muted because it cuts against the single-vendor positioning. MCP is genuinely a cross-vendor protocol Anthropic authored that other providers are adopting. Framed honestly, it would support a multi-model enterprise architecture; framed as in this document ('a universal standard that allows Claude to connect securely'), it is reduced to 'Claude's USB-C'. The portability story is undersold.
  Read next: MCP specification itself; AWS MCP GA announcement (already extracted); Genkit middleware piece (already extracted) for cross-vendor abstraction patterns.
• incompleteAuditability claims are credible but narrow. OpenTelemetry-compatible observability is real and useful. But for the EU AI Act regulatory chain (where you must demonstrate human oversight, risk assessment, conformity assessment, post-market monitoring), traces of 'what Claude did' are necessary but not sufficient. The document does not engage what additional governance instrumentation is needed beyond OTEL traces.
  Read next: EU AI Act Articles on high-risk systems documentation; ISO/IEC 42001 AI management system standard; sector-specific AI assurance frameworks.
• incompleteAttention and orchestrator cognitive-load concerns are absent. The document moves employees from 'producing' to 'reviewing and judgement', as if review is purely value-additive. But review fatigue, automation complacency, deskilling of junior staff who never learn to draft from scratch — these are real failure modes the document does not engage. This is the hidden-layer concern (Bailey, Stolen Focus, Cutler on overload) the reader carries.
  Read next: Cutler TBM on institutionalised overload + AI; Bailey on attention infrastructure; deskilling literature; Heis ep 9 Brede Moe on autonomy/learning culture in AI-augmented organisations.
• illogicalInternal contradiction between case studies and product pitch. The case studies (L'Oreal, Lyft, Rakuten in its original architecture) succeeded by building custom platforms — exactly what the Chapter 4 pitch tells you not to do. The document handles this by saying 'Claude Cowork changes the equation' going forward, but does not show evidence that Cowork-deployed customers achieve comparable outcomes. The credibility is borrowed; the inference is forward-projected.
  Read next: Look for Cowork-specific case studies (none yet provided in this document) when they exist; treat the L'Oreal/Lyft/Rakuten numbers as evidence for transformation worth pursuing, not as evidence for Cowork specifically.

Closes gaps elsewhere

• Agents Over Bubbles · g1 — Supplies the missing quantitative enterprise-agent evidence the gap asks for: Anthropic customer case studies with numbers (L'Oreal 44k users/99.9% accuracy, Lyft 87% faster resolution, Rakuten biweekly releases and 97% fewer critical errors).
• Agents Over Bubbles · g4 — Documents agent effectiveness outside coding — L'Oreal conversational analytics and Lyft customer support with quantified results — addressing whether agents generalize beyond software engineering.
• AWS MCP Server Reaches GA with Full API Coverage and IAM-Based Governance · g6 — Is the agent-runtime market commentary the gap sought: Anthropic explicitly positions the runtime layer (Claude Cowork, Managed Agents, plugin marketplaces) as where enterprise AI consolidates, with cloud endpoints reduced to connected backends.
• Why Enterprise AI Has a Leadership Problem · g1 — Provides the positive organizational-transformation case studies the gap found absent (L'Oreal, Lyft, Rakuten with measured outcomes), though vendor-sourced.

Related documents

• Google Brings MCP Support to Colab, Enabling Cloud Execution for AI Agents
• This Week in AI: Production Viability
• Incident 1368: Malicious OpenClaw Skills Reportedly Delivered AMOS Stealer and Exfiltrated Credentials via ClawHub
• You Outsourced the AI—but You Still Own the Risk
  Agent Governance (Internal) Enterprise AI Adoption & Readiness MCP (Model Context Protocol) Moats & Defensibility
• AWS Launches Amazon GuardDuty Investigation Agent to Automate Threat Triage
  Agent Governance (Internal) Agentic AI MCP (Model Context Protocol)
• BREAKING: Block (Jack Dorsey's company) just open-sourced a local AI agent
  Agentic AI Local AI MCP (Model Context Protocol)
• How Claude Code works in large codebases: Best practices and where to start
  Agent Governance (Internal) Agent Skills MCP (Model Context Protocol)
• An Evolutionary Architecture Pattern for Managing AI's Pace of Change
  Agent Governance (Internal) Enterprise AI Adoption & Readiness MCP (Model Context Protocol)
• The Founder's Playbook: Building An AI-Native Startup
  Agent Skills Moats & Defensibility Non-Engineers Building
• GitLab Duo Expands Self-Hosted AI Options Through Microsoft Foundry
  Data Sovereignty Enterprise AI Adoption & Readiness Local AI