The Biggest Consulting Firms Are Making a Generational Bet on Claude
When Accenture announced a $3 billion AI investment in 2023, it wasn't abstract. A significant portion of that capital went towards building dedicated Claude practices — teams of consultants trained specifically on Claude enterprise implementation, Claude API architecture, and AI transformation advisory. Deloitte and PwC followed with similar moves, and by early 2026, all three firms had built substantial Claude-specific capabilities that represent some of the most concentrated practical knowledge in the enterprise AI market.
This matters for two reasons. First, if you're a CIO evaluating whether to engage one of these firms for a Claude project, you need to understand what you're actually buying — and what you're not. Second, if you're building your own organisation's Claude capabilities, the structural choices the Big Four have made reveal what works at scale. The Accenture, Deloitte, and PwC Claude practices are blueprints — flawed ones in places, but blueprints nonetheless.
What These Firms Have Actually Built
How They're Structuring Their Claude Practices
The organisational model each firm has converged on is similar despite different branding. At the core is a "core team" of 20–50 Claude architects — the most technically senior practitioners who have shipped production Claude systems, contributed to internal tooling, and act as the knowledge hub for the broader organisation. Around them is a much larger "delivery team" of consultants who have completed structured Claude training and can execute against established patterns, but who aren't original architects.
This hub-and-spoke model mirrors how effective Claude deployments work inside enterprises — a small group of Claude Certified Architects sets the standards and solves novel problems, while a larger trained team executes. The ratio is typically 1 architect to 8–12 delivery consultants.
What's notable is what sits beneath this structure: a substantial investment in proprietary frameworks and accelerators. Accenture has built what they call "Claude Foundations" — a set of pre-configured MCP server templates, security configurations, and deployment patterns for common enterprise scenarios. The value proposition is: you're buying not just consultant time, but the accumulated learning from 50+ deployments baked into reusable tooling.
When you engage a Big Four firm for Claude work, you're buying three things: brand assurance, access to pre-built frameworks, and a very large talent pool. What you're not necessarily buying is the deepest technical expertise in the room. The most technically capable Claude practitioners in the market often aren't at large consultancies — they're at specialist firms, hyperscalers, or building AI-native startups.
What's Working in These Practices
Three structural decisions made by the Big Four appear to be genuinely effective, based on the results their clients are reporting.
Vertical specialisation over horizontal AI generalism
Rather than building a Claude practice that claims to serve all industries, each firm has doubled down on their existing vertical strengths. Deloitte's Claude practice is strongest in regulated industries because that's where Deloitte already had the deepest relationships and domain knowledge. The Claude expertise amplifies existing industry knowledge rather than creating a generic AI capability that competes with the model's own training. This mirrors our advice to clients: a Claude strategy that starts with your specific domain problems produces better outcomes than a strategy that starts with AI capabilities.
Embedding technical and business consultants on the same team
The failed AI projects we've seen most often fail because of a disconnect between the technical team (who understand the Claude API) and the business team (who understand the client's operational context). The Big Four have structured their Claude projects to avoid this by deliberately pairing technical architects with functional consultants who can translate business requirements into system design inputs. This isn't novel organisational theory — it's the lesson from every previous technology transformation wave. But it's being implemented consistently.
Treating governance as a project deliverable, not an afterthought
Perhaps the most important structural choice: each firm has made AI governance a formal deliverable in every Claude engagement, not an optional module. Their enterprise clients get an acceptable use policy, an audit logging architecture, a risk assessment framework, and an incident response playbook alongside the working AI system. This adds 15–20% to project cost but dramatically reduces the probability of a governance failure causing the deployment to be shut down post-launch.
Where the Big Four Claude Practices Fall Short
Being accurate about this matters, because enterprises selecting AI partners need honest market intelligence.
Speed: The Big Four are not fast. The average time from contract signature to first working prototype in a Big Four Claude engagement is 10–14 weeks. Specialist firms are typically at working prototype in 3–4 weeks. When speed matters — and for most enterprises trying to build competitive advantage, it does — scale and process are liabilities, not assets.
Technical depth at the frontier: The core teams are technically capable. But the 600-person delivery layer varies enormously in capability. In a large consulting firm, "Claude-trained" can mean anything from having completed a 4-hour internal course to having shipped production agentic systems. Due diligence on who specifically will be in the room on your project — not just the firm's aggregate capability — is essential.
Independence: Large consultancies are in the business of long-term relationships. This creates structural incentives to scope engagements in ways that maximise delivery hours rather than minimise time to client capability. The best outcome for a client is a capable internal team that doesn't need the consultancy anymore. The best outcome for the consultancy is an ongoing dependency. These interests are in tension.
Speed, Depth, and Independence — All Three
Our Claude implementation engagements are structured to produce internal client capability. We're a Claude Partner Network member with the technical depth of a specialist firm and the structured delivery model of a consultancy. We measure success by your team's ability to operate without us.
Book a Free Consultation →What Every Enterprise Can Learn From This
Whether you're building an internal Claude capability or evaluating external partners, five lessons from the Big Four model apply directly to your situation.
1. Specialise early. The firms winning in the Claude market are those with deep vertical expertise, not broad AI generalism. Identify the two or three business domains where Claude can create the most value for your organisation and go deep before going wide.
2. Build a knowledge infrastructure. The proprietary frameworks and accelerators that the Big Four have built are the source of their delivery leverage. Your organisation should be building equivalent internal assets: prompt libraries, architecture patterns, evaluation frameworks, governance templates. This turns individual knowledge into institutional knowledge.
3. Hire for the hub, train for the spokes. You don't need everyone to be a Claude Certified Architect. You need two or three deep experts who set standards and mentor, and a broader team of trained practitioners who can execute. Focus hiring investment on the hub; invest in structured training for the spokes.
4. Make governance a deliverable. Every internal Claude deployment should produce governance artefacts alongside the working system. Treat the absence of a governance framework as a deployment blocker, not as a future phase of work.
5. Measure partner independence, not partner dependency. Whether you're working with us, Accenture, or anyone else, the right measure of a successful engagement is how capable your internal team is when the consultants leave — not how long the engagement lasts. Any partner who can't articulate their exit plan for your team isn't the right partner.