Consulting is turning into agent delivery, not decks, as big firms chase “AI-native” work

KPMG’s US CEO Tim Walsh told Business Insider in January that the firm views itself as a technology company that delivers audit, tax, and advisory work. That line would have sounded like branding a decade ago. In 2026, it’s increasingly a procurement and operating-model signal: more consulting work is being sold, staffed, and governed as ongoing software delivery.

The near-term pressure is coming from two directions at once. First, enterprises are buying packaged AI agents from SaaS vendors because they’re easy to deploy and show quick productivity wins. Second, the value operators actually want, closed-loop automation that takes action inside systems of record, is pushing work toward custom and domain-specific agents that behave more like products than projects.

“AI-native” rhetoric is showing up as new delivery models and new talent math

Business Insider’s Polly Thompson reports that the center of gravity in consulting is moving away from generalist teams producing research and decks and toward building tools, systems, and multi-year transformation support. Rob Fisher, vice chair of advisory at KPMG, described the shift as embedding AI across traditional offerings while responding to client demand for “subscription-style products” alongside advice and managed services, according to Business Insider.

That demand for product-like delivery is changing staffing patterns. Business Insider notes that Accenture has added nearly 40,000 AI and data professionals over the last two years. The same report says EY has added 61,000 technologists since 2023. Those figures matter for operators because they’re a proxy for how much real implementation capacity sits behind the pitch deck, especially when programs move from pilots to production support.

PwC is also reshaping its workforce model. Business Insider reports the firm rewrote training around 30 core skills, split between 15 AI-centric and 15 human-centric capabilities, and introduced a new engineering career track in early 2026, its first such track in its 170-year history. In procurement terms, that’s a sign the Big Four expect more work to be evaluated like engineering delivery, with architecture, testing, and release cycles, rather than purely advisory staffing.

Forrester’s agent race frames what enterprises will buy first, and what will matter later

Forrester analyst Craig Le Clair describes four paths in the agentic AI market: SaaS and horizontal agents, edge agents, custom-built agents, and targeted agentic systems. In Forrester’s framing, the “hares” are the SaaS agents from vendors such as Microsoft, Salesforce, ServiceNow, SAP, and Google that are proliferating because they’re increasingly affordable and easy to deploy for tasks like summarizing, drafting, and information retrieval.

That speed is real. The operational catch, per Forrester, is that many packaged agents rely on similar foundation models and risk becoming interchangeable over time. When multiple vendors demo similar capabilities, buyers can end up competing on price while still carrying the integration and change-management burden.

Forrester’s “tortoises” are where the harder enterprise economics show up. Custom-built agents, built around a company’s processes, private data, and business logic and run behind corporate firewalls, are positioned to address what Forrester calls the “action gap,” the distance between an AI system’s recommendations and the actual execution of work in operational systems. Targeted agentic systems go further, coordinating multiple agents to run function- or industry-specific processes in areas like procurement, legal, healthcare, or finance, according to Forrester.

What changes in SOWs when consultants sell “agents as delivery”

Put Business Insider’s consulting shift next to Forrester’s agent taxonomy and a practical conclusion emerges: “AI-native” consulting is increasingly a bet on custom and targeted agentic systems, even if the first phase starts with SaaS copilots. The result is a different set of deliverables to specify and a different set of risks to govern.

For operators, the most immediate change is that more engagements will carry ongoing obligations. Business Insider’s reporting on subscription-style products and managed services at KPMG is a clue. If a consultancy is embedding AI in core audit, tax, advisory, or transformation work, the buyer is no longer only paying for analysis. They’re paying for a living system that needs monitoring, updates, and access management.

The second change is architectural. Forrester’s emphasis on edge agents points to a split deployment model for industrial, healthcare, and field-service environments. Where latency, connectivity, or data residency constraints apply, Forrester says smaller language models can enable fast, low-cost, secure AI at the edge, such as on factory floors or medical devices. That’s a different integration conversation than a cloud-only “turn on the feature” SaaS agent rollout.

The third change is procurement posture. When consultancies hire tens of thousands of AI and data professionals, as Business Insider documents at Accenture and EY, buyers should expect more vendor-led opinions about which platforms, orchestration layers, and model providers become standard. That can be helpful. It can also quietly harden into platform dependency unless the contract defines portability, ownership of agent logic, and boundaries around proprietary accelerators.

Contract and architecture questions to bring to the next AI-native consulting bid

• If the engagement includes “subscription-style products” or ongoing support, what is the operating model after go-live: SLAs for agent uptime, escalation paths, and a change-control process for prompts, tools, and workflow logic? Business Insider reports KPMG is responding to demand for subscription-like consumption, which belongs in the SOW as run-state obligations, not a handshake.
• Which category of agents are being proposed for each use case: packaged SaaS agents, edge agents, custom-built agents, or targeted agentic systems? Forrester’s four-path framing is a fast way to map cost, timeline, and governance expectations to the right delivery approach.
• Where does the agent run and where does data go: behind the firewall, at the edge, or in a vendor cloud? Forrester flags privacy, compliance, and ROI as drivers for custom agents, and that should translate into explicit data-boundary language and auditability requirements.
• What is the ownership model for agent logic and integration assets: are workflows, connectors, and evaluation harnesses portable if the managed-services provider changes? This matters more as consultancies shift toward software-like delivery, which Business Insider reports across the Big Four’s repositioning.